Gateways to the Internet of Things (IoT) are typically servers that communicate with IoT devices, providing them with low-latency services, and connecting them to the internet and other backbone networks. Since IoT devices are often simple and have limited storage and computational capabilities, gateways can be equipped with Attack Detection (AD) software to analyze incoming traffic, detect potential cyberattacks, and protect both the gateway and connected devices from threats that could overwhelm the system as a whole. This paper presents an enhanced gateway system that combines a traffic shaping technique with an attack detection module and an opti-mum attack mitigation scheme aimed at protecting the gateway and the overall system from cyberattacks. The optimum mitigation approach selects a sampling interval for the AD, that minimizes the total overhead of AD and mitigation. The proposed approach is implemented in a practical test-bed, so that the performance of the mitigation scheme may be evaluated in the presence of flood attacks. The experiments show its practical value and illustrate the agreement obtained between the analysis and the measurements obtained from several experiments

Signalling storms have occurred in several generations of mobile networks due to malevolent end users, as well as a consequence of service disruptions and system failures. Due to the massive increase in the volume of traffic and the diversity of end users and applications in $6 G$, including the massive access of the Internet of Things, signalling storms and other cyberattacks will require attention with research on adequate detection and mitigation schemes, as we move towards the next generation of mobile networks. Thus, in the present paper, we conduct discrete event simulations of a signalling storm, measure its effects in a simulated environment, and observe that it results in deviations from the well-known and naturally observed statistical property known as “Taylor’s Law”. We then discuss a mathematical model of storms, linking the fraction of malevolent or malfunctioning users to the rate of call establishment requests, and show that the mathematical model also exhibits a violation of Taylor’s Law. Since the detection of cyberattacks from network behaviour parameters is a challenging problem, we suggest that the violation of Taylor’s Law may be a useful tool for designing detectors for signalling attacks, or other network attacks or malfunctions.

This paper introduces a new dynamic networked trust model, the Random Neural Network Trust Model (RNNTM), which incorporates the dynamics of trust formation in a network through a sequence of “votes” from each entity regarding all other entities. The model assures fairness among entities through a fixed replenishment rate of “voting rights” for each entity, whose voting rights are reduced each time the entity votes. A positive vote received by an entity increases its voting rights and its trustworthiness, while a negative vote reduces its voting rights and also its trustworthiness, and a non-negative integer represents each entity’s instantaneous “trust value”. An important property of the RNNTM is that an entity cannot express its trust or distrust of the other entities, and hence affect their trust values, when its trust level is down to zero, so that untrustworthy entities are not allowed to express trust or distrust. After developing the theoretical characteristics for the RNNTM model, this paper details its use to evaluate the trust value of multiple entities in a network of Internet of Things (IoT) devices and gateways, where cyberattacks against the gateways, and messages that should be received from IoT devices at regular intervals, modify the parameters that express the trust or distrust between entities. To illustrate its use for a network of interaction gateways, servers and user equipment, several detailed time-dependent simulations of the RNNTM are conducted in the presence of cyberattacks.

The growing proliferation of Internet of Things (IoT) devices has intensified the need for sustainable energy solutions, particularly in resource-constrained deployments where non-rechargeable batteries and supercapacitors are the primary energy sources. Green IoT (G-IoT) frameworks address this challenge by combining energy-saving techniques with energy harvesting from ambient sources such as solar power. However, the intermittent nature of renewable energy and the non-ideal behavior of energy storage systems-such as energy leakage and capacity degradation-complicate reliable energy provisioning. This paper presents a novel Markovian framework for modelling the coupled dynamics of time-varying solar energy harvesting, time-dependent energy consumption, and state-dependent energy leakage in G-IoT systems. Unlike traditional steady-state models, our approach uses Discrete-Time Markov Chains (DTMCs) to capture the stochastic variability in both energy harvesting and consumption processes. We also introduce a refined leakage model in which the leakage rate is dynamically dependent on the stored energy level, enabling a more realistic characterization of energy losses due to energy leakage. Through extensive analytical evaluation, we examine how key parameters-such as storage capacity, leakage rate coefficient, and energy harvesting and consumption patterns-affect critical performance metrics, including the mean stored energy and energy-related service outage probability. Furthermore, we propose a parameter tuning strategy to optimize energy reliability and storage efficiency. The proposed model provides valuable insights for the design and optimization of robust, energy-aware IoT systems powered by renewable energy sources.

Gateway Servers for the Internet of Vehicles (IoV) must meet stringent Security and Quality-of-Service (QoS) requirements, including cyberattack protection, low delays, and minimal packet loss, to offer secure real-time data exchange for human and vehicle safety and efficient road traffic management. Therefore, it is vital to protect these systems from cyberattacks with adequate attack detection (AD) and Mitigation mechanisms. Such attacks often include packet Floods that impair the QoS of the networks and Gateways and even impede the Gateways’ capability to carry out AD. Thus, this article first evaluates these effects using system measurements during Flood attacks. It then demonstrates how a smart quasi-deterministic policy forwarder (SQF) at the entrance of the Gateway can regulate the incoming traffic to ensure that the Gateway supports the AD to operate promptly during an attack. Since Flood attacks create substantial packet backlogs, we propose a novel adaptive attack mitigation (AAM) system that is activated after an attack is detected to dynamically sample the incoming packet stream, determine whether the attack is continuing, and also drop batches of packets at the input to reduce the effects of the attack. The AAM is designed to minimize a cost function that includes the sampling overhead and the cost of lost benign packets. We show experimentally that the Optimum AAM approach is effective in mitigating attacks and present theoretical and experimental results that validate the proposed approach.

The Massive Access Problem arises due to devices that forward packets simultaneously to servers in rapid succession, or by malevolent software in devices that flood network nodes with high-intensity traffic. To protect servers from such events, attack detection (AD) software is installed on servers, and the Quasi-Deterministic Transmission Policy (QDTP) has been proposed to “shape traffic” and protect servers, allowing attack detection to proceed in a timely fashion by delaying some of the incoming packets individually based on their arrival times. QDTP does not cause packet loss, and can be designed so that it does not increase end-to-end packet delay. Starting with measurements taken on an experimental test-bed where the QDPT algorithm is installed on a dedicated processor, which precedes the server itself, we show that QDPT protects the server from attacks by accumulating arriving packets at the input of the QDTP processor, then forwarding them at regular intervals to the server. We compare the behaviour of the server, with and without the use of QDTP, showing the improvement it achieves, provided that its “delay” parameter is correctly selected. We analyze the sample paths associated with QDTP and prove that when its delay parameter is chosen in a specific manner, the end-to-end delay of each packet remains unchanged as compared to an ordinary First-In-First-Out system. An approach based on stationary ergodic processes is developed for the stability conditions. Assuming mutually independent and identically distributed inter-arrival times, service times and QDTP delays, we exhibit the positive recurrent structure of a two-dimensional Markov process and its regeneration points.

Pipelines are the most convenient ways to transport fluids (e.g., water, oil, and gas). However, leakage of fluids into the environment results in resource wastage (primarily water, which is becoming a scarce resource) and environmental pollution (in the case of leakage of toxic fluids like oil and gas). Emerging technologies like the Internet of Things (IoT), Wireless Sensor Networks (WSNs), Artificial Intelligence (AI), distributed computing, and cloud computing enable continuous monitoring of pipelines to detect leakages and corrosion on the pipeline. The main challenge with using battery-powered sensor nodes to monitor pipelines is the energy constraint, necessitating frequent battery replacement. Thus, there is a need to develop energy-saving mechanisms to prolong the lifetime of these sensor nodes. In this paper, we use the diffusion approximation modelling framework in which the data from the experimental testbed are used to model the dynamics of the battery’s energy content and to estimate the mean and variance of the device’s lifetime. The novelty in the proposed diffusion model of the battery of an IoT node is the introduction of multiple energy thresholds that split the energy state-space of the battery into multiple energy-saving regimes. As the battery discharges, the node gradually transitions into energy-saving regimes by reconfiguring some of its parameters to reduce energy consumption (sometimes at the cost of trading off some performance metrics). We investigate the impact of energy-saving regimes or the number of thresholds on the node’s lifetime.

We analyze, further and deeper, a recently proposed technique for addressing the Massive Access Problem (MAP), an issue in telecommunications which arises when too many devices transmit packets to a gateway in quick succession. This technique, the Adaptive-Quasi-Deterministic Transmission Policy (AQDTP) is a special case of “traffic shaping” which involves delaying some packets at the points of origin to alleviate congestion at the routers. One nice feature of AQDTP is that it loses no packets and allows an infinite buffer. In this work, to clarify the approach in a general queueing theory framework, and to move beyond the original telecommunications application, we frame these potential delays as time spent at a café by customers before proceeding to a service facility. We first present some sample-path results that significantly refine and expand upon what was shown in previous work, and then present further results under a general stationary ergodic stochastic framework. In the sample-path realm, we give conditions that ensure AQDTP will not change the total delay and sojourn time of any customer as compared to what that customer would have experienced if there was no café; but we also prove that AQDTP can never reduce the total delay. The difference is that, under AQDTP, some of that delay is spent at the café instead of in the queue/line at the service facility. In a stochastic framework, our focus is on stability and constructing proper stationary versions of the model. Under i.i.d. assumptions we dig deeper by proving Harris recurrence of an underlying two-dimensional Markov process, and explicitly find positive recurrent regeneration points.

In the present paper, we investigate the impact of imperfections (non-idealities) of the energy storage system (e.g., batteries, capacitors, or supercapacitors) on the energy performance of green IoT nodes. In particular, we investigate the impact of energy leakage from the Energy Storage System (ESS) on important energy performance metrics, such as service outage probability, the density of the lifetime of the node, and the time-dependent mean number of energy packets (EPs) in the ESS. Also, we explore various strategies that can be employed to compensate for the impact of energy losses due to energy leakage on the energy performance metrics. Specifically, we examine two potential strategies for improving the energy performance of the IoT nodes: (i) through increasing the energy generation rate of the energy harvesters (e.g., by adding additional solar panels or replacing the existing solar panels with more efficient ones that can produce more energy), and (ii) through reducing the energy consumption rate of the IoT node (e.g., by configuring ESS energy thresholds below which the nodes are forced to operate in low energy consumption states).

Cyberattacks, especially Botnet Distributed Denial of Service (DDoS), increasingly target networked systems, compromise interconnected nodes by constantly spreading malware. In order to prevent these attacks in their early stages, which includes stopping the spread of malware, it is vital to identify compromised nodes and successfully predict potential attack paths. To this end, this paper proposes a novel system based on an Associated Random Neural Network (ARNN) that simultaneously detects intrusion at the network-level and estimates the network attack graph. In this system, ARNN is trained online to minimize problem-specific multi-task loss so that it identifies compromised network nodes, while the neural network connection weights also estimate the attack path. The performance of the method is calculated using the Kitsune attack dataset, showing that the method achieves a recall rate above 0.95 in estimating the network attack graph, and provides a near-perfect classification of compromised nodes. The ARNN-based system for dynamic and continuous estimation of compromised nodes and network attack graphs, can pave the way for enhancing security measures, and stopping Botnet DDoS attacks from spreading in networked systems.

Internet of Things (IoT) networks are highly vulnerable to common network DoS and DDoS attacks, which flood limited system resources or IoT devices, overwhelming them with large numbers of attack packets. In order to mitigate such attacks, this paper develops a lightweight yet effective Intrusion Detection and Prevention System (IDPS), that sequentially detects and mitigates the attack via a Deep Random Neural Network (DRNN) and a Drop-Idle-Repeat process. The IDPS is evaluated for UDP Floods, attacks on an experimental test-bed. The results show that UDP Flood attacks can be mitigated with the proposed IDPS, allowing the system to continue routine operations, and resume communications when the attack ends.

Vulnerability prediction is an important mechanism for secure software development, as it enables the early identification and mitigation of software vulnerabilities. Vulnerability Prediction Models (VPMs) are Machine Learning (ML) models able to detect potentially vulnerable software components based on information retrieved from their source code. Despite the notable advancements in the field of vulnerability prediction, especially with the utilization of Deep Learning (DL) and text mining techniques, current literature still lacks a highly accurate, reliable, and practical VPM. Recently, the Large Language Models (LLMs), which have demonstrated remarkable capabilities in text understanding and processing, have started being utilized for vulnerability prediction, demonstrating highly promising results. The purpose of the present paper is to explore the utilization of LLMs in the field of vulnerability detection, identify challenges and open issues that still need to be addressed, and potentially propose directions for future research. Our analysis suggests that while LLM-based VPMs have outperformed traditional DL approaches in vulnerability prediction, significant challenges still need to be addressed to be considered sufficiently accurate, reliable, and practical.

Smartphones are widely used as portable data collectors for wearable and healthcare sensors that can passively collect data streams related to the environment, health status, and behaviors. Recent research shows that the collected data can be used to monitor not only the physical states but also the mental health of individuals. However, extracting the features of digital phenotypes that characterize major depressive disorder (MDD) is technically challenging and may raise significant privacy concerns. Addressing such challenges has become the focus of many researchers. This article provides a comprehensive analysis of several key issues related to ubiquitous sensing to aid in detecting MDD. Specifically, this article analyzes existing methodologies and feature extraction algorithms used to detect possible MDD through digital phenotyping from smartphone data. In particular, five types of features are summarized and explained, namely, location, movement, rhythm, sleep, and social and device usage. Finally, related limitations and challenges are discussed to provide paths for further research and engineering.

This paper proposes a novel Self-Supervised Intrusion Detection (SSID) framework, which enables a fully online Deep Learning (DL) based Intrusion Detection System (IDS) that requires no human intervention or prior off-line learning. The proposed framework analyzes and labels incoming traffic packets based only on the decisions of the IDS itself using an Auto-Associative Deep Random Neural Network, and on an online estimate of its statistically measured trustworthiness. The SSID framework enables IDS to adapt rapidly to time-varying characteristics of the network traffic, and eliminates the need for offline data collection. This approach avoids human errors in data labeling, and human labor and computational costs of model training and data collection. The approach is experimentally evaluated on public datasets and compared with well-known machine learning and deep learning models, showing that this SSID framework is very useful and advantageous as an accurate and online learning DL-based IDS for IoT systems.

This paper presents several novel algorithms for real-time cyberattack detection using the Auto-Associative Deep Random Neural Network. Some of these algorithms require offline learning, while others allow the algorithm to learn during its normal operation while it is also testing the flow of incoming traffic to detect possible attacks. Most of the methods we present are designed to be used at a single node, while one specific method collects data from multiple network ports to detect and monitor the spread of a Botnet. The evaluation of the accuracy of all these methods is carried out with real attack traces. The novel methods presented here are compared with other state-of-the-art approaches, showing that they offer better or equal performance, with lower learning times and shorter detection times, as compared to the existing state-of-the-art approaches.

The IoT is vulnerable to network attacks, and Intrusion Detection Systems (IDS) can provide high attack detection accuracy and are easily installed in IoT Servers. However, IDS are seldom evaluated in operational conditions which are seriously impaired by attack overload. Thus a Local Area Network testbed is used to evaluate the impact of UDP Flood Attacks on an IoT Server, whose first line of defence is an accurate IDS. We show that attacks overload the multi-core Server and paralyze its IDS. Thus a mitigation scheme that detects attacks rapidly, and drops packets within milli-seconds after the attack begins, is proposed and experimentally evaluated.

There is a growing awareness of the need to reduce carbon emissions from the operation of mobile networks. The massive deployment of ultra-dense 5G and IoT networks will significantly increase energy demand and put the electricity grid under stress while also driving up operational costs. In this paper, we model the energy performance of an off-grid sustainable green cellular base station site which consists of a solar power system, Battery Energy Storage (BESS) and Hydrogen Energy Storage (HESS) system, and various types of macrocells, microcells, picocells, or femtocells, with broadband optical or microwave transmission systems, and other electrical and electronic systems (air conditioner, power converters, and controllers. We propose diffusion-based models of the charging and discharging processes of the energy storage systems, and obtain the probability of charging them to their full capacities during the day and completely discharging them at the end of each day. We also investigate the impact of design parameters such as the mean charging rate and the mean discharging rate on the probability densities of charging BESS and HESS to their full capacities during the day and of completely discharging them before the end of each night period.

Cyberattacks are increasingly threatening net-worked systems, often with the emergence of new types of unknown (zero-day) attacks and the rise of vulnerable devices. uch attacks can also target multiple components of a Supply Chain, which can be protected via Machine Learning (ML)-based Intrusion Detection Systems (IDSs). However, the need to learn large amounts of labelled data often limits the applicability of ML-based IDSs to cybersystems that only have access to private local data, while distributed systems such as Supply Chains have multiple components, each of which must preserve its private data while being targeted by the same attack To address this issue, this paper proposes a novel Decentralized and Online Federated Learning Intrusion Detection (DOF-ID) architecture based on the G-Network model with collaborative learning, that allows each IDS used by a specific component to learn from the experience gained in other components, in addition to its own local data, without violating the data privacy of other components. The performance evaluation results using public Kitsune and Bot-loT datasets show that DOF -ID significantly improves the intrusion detection performance in all of the collaborating components, with acceptable computation time for online learning.

Servers at Supply Chains and othet Cyber-physical systems that receive packets from IoT devices should meet the QoS needs of incoming packets, and protect the system from Cyberattacks. UDP Floods are often included in attacks to overwhelm Supply Chains and the IoT through congestion that paralyzes their ability for timely Attack Detection and Mitigation. Thus this paper proposes an architecture that protects a connected Server using a Smart Quasi-Deterministic Transmission Policy Forwarder at its input. This Forwarder shapes the incoming traffic, sends it to the Server without increasing the overall packet delay, and avoids Server congestion. The relevant theoretical background is reviewed, and measurements during a UDP Flood Attack are provided to compare the Server performance, with and without the Forwarder. It is seen that during a UDP Flood Attack, the Forwarder protects the Server from congestion, allowing it to effectively identify Attack Packets. Congestion at the Forwarder is rapidly eliminated with "drop" commands generated by the Forwarder, or sent by the Server to the Forwarder.

Developing modern secure and low-energy applications in a short time imposes new challenges and creates the need of designing new software tools to assist developers in all phases of application development. The design of such tools cannot be considered a trivial task, as they should be able to provide optimization of multiple quality requirements. In this paper, we introduce the SDK4ED platform, which incorporates advanced methods and tools for measuring and optimizing maintainability, dependability and energy. The presented solution offers a com-plete tool-flow for providing indicators and optimization meth-ods with emphasis on embedded software. Effective forecasting models and decision-making solutions are also implemented to improve the quality of the software, respecting the constraints imposed on maintenance standards, energy consumption limits and security vulnerabilities. The use of the SDK4ED platform is demonstrated in a healthcare embedded application.

IoT gateways aim to meet the deadlines and QoS needs of packets from as many IoT devices as possible, though this can lead to a form of congestion known as the Massive Access Problem (MAP). While much work was conducted on predictive or reactive scheduling schemes to match the arrival process of packets to the service capabilities of IoT gateways, such schemes may use substantial computation and communication between gateways and IoT devices. This paper proves that the recently proposed "Quasi-Deterministic-Transmission-Policy (QDTP)" traffic shaping approach which delays packets at IoT devices, substantially alleviates the MAP: QDTP does not increase overall end-to-end delay and reduces gateway queue length. We then introduce the Adaptive Non-Deterministic Transmission Pol-icy (ANTP) that requires only one packet buffer at the gateway, offering substantial QoS improvement over FIFO scheduling.

The paper considers important issues surrounding the energy consumption by Information and Communication Technologies (ICT) which has been steadily growing and is now attaining approximately 10% of the worldwide electricity consumption with a significant impact on greenhouse gas emissions. The perimeter of ICT systems is discussed, and the role of the sub-systems that compose ICT is considered. Data from recent years is used to consider how each of these subsystems contribute to ICT’s energy consumption. The positive correlation between the penetration of ICT in some of the world’s different economies and the same economies’ contributions to undesirable greenhouse gas emissions is also discussed. We also examine how emerging technologies such as 5G, edge computing and cryptocurrencies are contributing to the worldwide increase in electricity consumption by ICT, despite the ever increase in efficiency, in energy per bit processed, stored or transmitted by ICT systems. The measurement of specific ICT systems’ electricity consumption is considered, and the manner in which this consumption can be minimized in two specific technical contexts is also discussed in some detail.

The complexity of battery-powered autonomous devices such as Internet of Things (IoT) nodes or Unmanned Aerial Vehicles (UAV) and the necessity to ensure an acceptable quality of service, reliability, and security, have significantly increased their energy demand. In this paper, we discuss using a diffusion approximation process to approximate the dynamic changes in the energy content of a battery. We consider the case when energy harvesting sources are constantly charging the battery. The model assumes a probabilistic consumption and delivery of energy, giving the time-dependent distributions of the energy at the battery, of the time remaining until it becomes empty, the time required to charge the battery to its total capacity, or the time it is operational between two moments of complete depletion. When possible, we compare the diffusion approximation results with corresponding models based on continuous-time Markov chains.

Malicious network attacks are a serious source of concern, and machine learning techniques are widely used to build Attack Detectors with off-line training with real attack and non-attack data, and used online to monitor system entry points connected to networks. Many machine learning based Attack Detectors are typically trained to identify specific types attacks, and the training of such algorithms to cover several types of attacks may be excessively time consuming. This paper shows that G-Networks, which are queueing networks with product form solution and special customers such as negative customers and triggers, can be trained just with “non-attack” traffic, can accurately detect several different attack types. This is established with a special case of G-Networks with triggerred customer movement. A DARPA attack and non-attack traffic repository is used to train and test the the G-Network, yielding comparable or clearly better accuracy than most known attack detection techniques.

Random Neural Networks (RNNs) are recurrent network models with strong machine learning capabilities that have been exploited in many applications. We will first recall their theoretical properties, including the product form solution, their capability as approximations for continuous and bounded functions and their efficient deep learning algorithms. Then we will focus on the Cognitive Packet Network (CPN) which uses RNNs to introduce AI into network algorithms to route traffic intelligently with the objective of optimizing Quality of Service (QoS), reducing energy consumption and mitigating security threats and attacks. Then we will provide experimental results showing that CPN can be implemented in various ways: through additional software resident in conventional routers, also through Software Defined Networks (SDN), or also as overlay networks.

IoT networks handle incoming packets from large numbers of IoT Devices (IoTDs) to IoT Gateways. This can lead to the IoT Massive Access Problem that causes buffer overflow, large end-to-end delays and missed deadlines. This paper analyzes a novel traffic shaping method named the Quasi-Deterministic Traffic Policy (QDTP) that mitigates this problem by shaping the incoming traffic without increasing the end-to-end delay or dropping packets. Using queueing theoretic techniques and extensive data driven simulations with real IoT datasets, the value of QDTP is shown as a means to considerably reduce congestion at the Gateway, and significantly improve the IoT network’s overall performance.

Internet connected IoT devices have often been particularly vulnerable to Botnet attacks of the Mirai family in recent years. Thus we develop an attack detection scheme for Mirai Botnets, using the Auto-Associative Dense Random Neural Network that has recently been successful for other attacks such as the SYN attack. The resulting method is trained with normal traffic and tested with attack traffic, and shown to result in high accuracy detection of attacks with low false alarms. The approach is compared on the same data set with two other common Machine learning methods (Lasso and KNN) and shown to have higher accuracy, and much lower computation times than KNN and slightly higher (but comparable) computation times with respect to Lasso.

A significant challenge of IoT networks is to offer Quality of Service (QoS) and meet deadline requirements when packets from a massive number of IoT devices are forwarded to an IoT gateway. Many IoT devices tend to report their data to their wired or wireless network gateways at closely correlated instants of time, leading to congestion known as the Massive Access Problem (MAP), which increases the probability that the IoT data will not meet its required deadlines. Since IoT data loses much of its value if it arrives to destination beyond a required deadline, MAP has been extensively studied in the literature. Thus we first take a queueing theoretic view of the problem, and also use a Diffusion Approximation to gain insight into the IoT traffic statistics that affect MAP. Then we introduce the Quasi-Deterministic Transmission Policy (QDTP) which significantly alleviates MAP when the average traffic rate grows beyond a given level and substantially reduces the probability that IoT data deadlines are missed. The results are validated using real IoT data which has been placed in IP packets for transmission.

With the expansion of the IoT, it is important to optimize available bandwidth to reliably support edge to device communications. Thus we propose a wireless network where each edge server communicates with its end devices using its wireless band as a primary channel, assisted by a secondary edge server that can relay communications via its own wireless band as a secondary channel. The network can optimize capacity by balancing load between primary and secondary wireless bands, and we analyze the geometry of achievable rate regions, depending on the state of bands modeled as Rayleigh fading channels. The allocation of a connection to the primary or secondary band is formulated as an optimization problem which is then solved, and illustrated with numerical examples.

Although the abilities of human beings as participants in urban traffic, when they take decisions and interact with the transportation infrastructure and other vehicles, have been greatly amplified by powerful portable devices and efficient human-machine interfaces, the intelligence of vehicle drivers and pedestrians and their possible pro-social behaviour such as helpfulness and sense of duty, have been excluded in previous studies of Intelligent Transportation Systems (ITS). Thus the robustness of an ITS has not been evaluated as a function of the likelihood that participants follow instructions. Moreover, much effort has been dedicated to the use of Artificial Intelligence, while in fact many tasks can be easily accomplished by road users in the system who use ordinary human intelligence. Hence, in this paper, we propose a reward mechanism to integrate the intelligence of human road users into a large-scale transportation system to improve the effectiveness and robustness of the system by introducing a transportation-related task publishing system which is assisted by a queueing network model. The experimental results show that the use of a reward mechanism can significantly improve the performance of the transportation system in terms of average travel time of vehicles and the average response time to various tasks.

We study programs which operate in the presence of possible failures and which must be restarted from the beginning after each failure. In such systems checkpointsare introduced to reduce the large costs of program restarts when failures occur. Here we suggest that checkpoints should be introduced in a manner which assures effective reliability, while reducing both the computational overhead as much as possible, but also to save energy. We compute the total average program execution time in the presence of checkoints so as to limit the re-execution time of the program from the most recent checkpoint. We also study the total energy cnsumption of the program under the same conditions, and formulate an optimization problem to minimize a wighted sum of both average computation time and energy. This approach is placed in the context of Application Level Checkpointing and Restart (ALCR). We then focus on checkpoints placed at the beginning of a loop, and derive the optimum placement of checkpoints to minimize a weighted combination of the program's execution time and energy consumption. Numerical results are presented to illustrate the analysis. Finally we describe a software tool with a graphical interface that has been designed to assist a system designer in choosing the optimum checkpoint for a given program as a function of different failure rates and other parameters.

The need to adaptively manage computer systems and networks so as to offer good Quality of Service (QoS) and Quality of Experience (QoE) with secure operation at relatively low levels of energy consumption is challenged by their sheer complexity and the wide variability of the workloads. A possible way forward is through self-awareness, whereby self-measurement and self-observation, together with on-line control mechanisms, operate adaptively to attain the required performance and QoE. We survey the premises for these ideas arising from cognitive science and active networks and review recent work on self-aware computer systems and networks, including those that propose the use of software-defined networks as a means to implement these concepts. Then we provide some examples from the literature on self-aware systems to illustrate the performance gains that they can provide. Finally, we detail an example system and its working algorithms to allow the reader to understand how such a system may be implemented. Measurements showing how it can react rapidly to changing network conditions regarding QoS and security are presented. Some conclusions and suggestions for further work are listed.

Mobile data traffic is experiencing unprecedented increases due to the proliferation of highly capable smartphones, laptops and tablets, and mobile data offloading can be used to move traffic from cellular networks to other wireless infrastructures such as small-cell base stations. This work addresses the related issue of data allocation, by proposing a novel infrastructure independent method based on the hotspot function of smartphones. In the proposed scheme, smartphones transfer data allowances among mobile users, so that users with excess data allowances act as accessible Wi-Fi hotspots, selling their data allowance to other users who need extra data allowances. To achieve this objective, we propose to use auctions with single and multiple data sellers. Efficient schemes based on auction models are discussed to sell the data allowances over successive days in a month, and over different time slots during a single day. Overall system performance is considered based on the behavior of mobile users, such as changing demands for the sale or purchase of data allowances. Together with the analytical results presented, our simulation experiments also indicate that knowledge of user behavior can significantly improve the performance of data allowance transactions, leading to highly efficient allocations among users.

Over the last decade, the increased automation in the management of computer systems and networks, the rising costs of energy, and the sensitivity about the environmental impact of energy consumption, has resulted in an increase of the importance of energy in the overall running of ICT (Information and Communication) Systems. The deregulation in telecommunication services has also had an impact in this direction because of the increased duplication in telecommunication infrastructures. Furthermore, most cyberattacks, also increase the energy consumption of ICT systems. Thus, today the overall electricity consumed by ICT worldwide is comparable to the total electricity consumption of two major industrial powers: Japan plus Germany. Clearly, of one wishes to save energy in ICT, this may come at the expense of a reduction in the measured Quality of Service (QoS) of ICT systems. Thus we will survey a number of our own research results, covering wireless communications, compute servers and network routing, to offer optimal trade-offs between energy consumption and QoS. The Energy Packet Network (EPN) that we have introduced to optimise the Energy-QoS trade-offs will be described and applied to a number of examples. We will also describe some results on the potential and limitations for communicating and computing with particle spins as a means to achieve energy savings.

A new shallow multi-layer auto-encoder that combines the spiking Random Neural Network (RNN) with the network architecture typically used in deep-learning, is proposed with a learning algorithm inspired by non-negative matrix factorization which satisfies the non-negative probability constraints of the RNN. Auto-encoders equipped with this learning algorithm are tested on typical images including the MNIST, Yale face and CIFAR-10 datasets, and also using 16 real-world datasets from different areas, exhibiting the desired high learning and recognition accuracy. Monte carlo simulations of the stochastic spiking behaviour of this RNN auto-encoder have also been carried out, showing that it can be implemented in a highly parallel manner to achieve substantial speed improvements.

Health data exchange is a major challenge due to the sensitive information and the privacy issues entailed. Considering the European context, in which health data must be exchanged between different European Union (EU) Member States, each having a different national regulatory framework as well as different national healthcare structures, the challenge appears even greater. Europe has tried to address this challenge via the epSOS ("Smart Open Services for European Patients") project in 2008, a European large-scale pilot on cross-border sharing of specific health data and services. The adoption of the framework is an ongoing activity, with most Member States planning its implementation by 2020. Yet, this framework is quite generic and leaves a wide space to each EU Member State regarding the definition of roles, processes, workflows and especially the specific integration with the National Infrastructures for eHealth. The aim of this paper is to present the current landscape of the evolving eHealth infrastructure for cross-border health data exchange in Europe, as a result of past and ongoing initiatives, and illustrate challenges, open issues and limitations through a specific case study describing how Italy is approaching its adoption and accommodates the identified barriers. To this end, the paper discusses ethical, regulatory and organizational issues, also focusing on technical aspects, such as interoperability and cybersecurity. Regarding cybersecurity aspects per se, we present the approach of the KONFIDO EU-funded project, which aims to reinforce trust and security in European cross-border health data exchange by leveraging novel approaches and cutting-edge technologies, such as homomorphic encryption, photonic Physical Unclonable Functions (p-PUF), a Security Information and Event Management (SIEM) system, and blockchain-based auditing. In particular, we explain how KONFIDO will test its outcomes through a dedicated pilot based on a realistic scenario, in which Italy is involved in health data exchange with other European countries.

With the development of online applications based on social networks, many different approaches have emerged to evaluate the service that these applications provide. Reports made by end users regarding the consumer's experience or opinion are commonly used to rate the quality of different online services. Therefore, ensuring the authenticity of the users' reports, and the detection of malicious users' dishonest reports, have both become important issues to achieve accuracy in the rating of such services. In this paper, we propose and evaluate a private-prior peer prediction-based trustworthy service rating system, which requires users to report their prior and posterior beliefs regarding whether their peers will report a high-quality opinion of the service. The reports are made to a data processing center which evaluates the users' trustworthiness by applying a strictly proper scoring rule, and removes reports received from users whose trustworthiness rating is low. This peer prediction method is compatible with incentives to motivate users to report honestly. In addition, an unreliability index is proposed to identify malicious users, and malfunctioning or unreliable users who have a high error rate in making judgments about quality. Thus, reports with high unreliability values will also be excluded from the service rating system. By combining trustworthiness and unreliability, malicious users face the dilemma that they cannot receive both a high trustworthiness and low unreliability rating simultaneously when their reports are false. Simulation results indicate that the proposed peer prediction-based trustworthy service rating can identify malicious and unreliable behaviors effectively and motivate users to report truthfully, and that a relatively high service rating accuracy is achieved by the proposed system.

Recently, wireless streaming of on-demand videos of mobile users (MUs) has become the major form of data traffic over cellular networks. As a response, caching popular videos in the storage of small base stations (SBSs) has been regarded as an efficient approach to reduce the transmission latency and alleviate the data traffic loaded over backhaul channels. This paper considers a small-cell based caching market composed of one mobile network operator (MNO) and multiple video service providers (VSPs). In this system, the MNO manages and operates its SBSs, and assigns these SBSs' storage to different VSPs, who have caching requirements. However, videos have different popularities and MUs present different preferences to these VSPs when they request videos. In addition, the caching service brings different utilities to different VSPs as well as that providing caching service to different VSPs causes distinct costs to the MNO. Such privacy information cannot be aware of among VSPs and the MNO. Therefore, to elicit this hidden information, this paper designs a double auction-based caching mechanism, which ensures the efficient operation of the market by maximizing the social welfare, i.e., the gap between VSPs' caching utilities and MNO's caching costs. Moreover, this paper demonstrates the economic properties of the designed caching mechanism, which are also validated by the simulation results.

Attacks on the content and quality of service of IoT platforms have economic and physical consequences well beyond the Internet's lack of security. This paper describes a new research project on “Secure and Safe Internet of Things” (SerIoT) to improve both the information and physical security of IoT applications platforms in a holistic and cross-layered manner. The purpose is to be able to create secure operational IoT platfnrms for diverse applieations.

The unprecedented growth of the volume of mobile data calls for novel approaches that improve the sharing of data allowances among mobile users with diverse needs. Specifically, the Wi-Fi hotspot function of current smartphones allows mobile-to-mobile offloading, but requires fast and efficient transactions between mobile users. Thus we propose an auction-based approach to allow the transfer of data allowances between mobile users with excess and deficits of data allowances, together with a cognitive approach to access the needed information about the system. The objective is to optimize the income of “sellers” and satisfy the needs of the other mobile users. Analytical and simulation results are presented, showing that by taking advantage of mobile users’ behaviors, and of varying demands of data allowance selling and buying, the cognitive auction and data allocation mechanism can significant improve the overall performance of the mobile data allowance transaction system.

Currently, The unprecedented increasing of mobile data traffic challenges the performance of current cellular networks. To meet this explosive demands of mobile traffic, the mobile data offloading technology has been proposed to alleviate the traffic load by moving traffic load of cellular networks to other wireless networks provided by infrastructures such as small-cell base stations. In this work, an infrastructure-free offloading method is proposed, which realizes the data transaction among mobile users by applying the hotspot function of smartphones. In this transaction, mobile users with redundant data perform as accessible Wi-Fi hotspots, and sell their mobile data to users with data requirements. Considering the scenarios with multiple data sellers, a networked auction model is introduced to model the process of data transaction. Additionally, high efficient data allocation mechanisms are designed in this work, which decide how to schedule the data transaction in different time slots, based on the establish edauction model. Simulation results indicate that introducing the prediction information of user behaviors can effectively improve the performance of data allocation, and achieve a high efficient data transaction operation.

The increasingly wide application of Cloud Computing enables the consolidation of tens of thousands of applications in shared infrastructures. Thus, meeting the QoS requirements of so many diverse applications in such shared resource environments has become a real challenge, especially since the characteristics and workload of applications differ widely and may change over time. This paper presents an experimental system that can exploit a variety of online QoS aware adaptive task allocation schemes, and three such schemes are designed and compared. These are a measurement driven algorithm that uses reinforcement learning, secondly a “sensible” allocation algorithm that assigns tasks to sub-systems that are observed to provide a lower response time, and then an algorithm that splits the task arrival stream into sub-streams at rates computed from the hosts' processing capabilities. All of these schemes are compared via measurements among themselves and with a simple round-robin scheduler, on two experimental test-beds with homogenous and heterogenous hosts having different processing capacities.

We consider auction mechanism design and performance analysis for data transactions in mobile social networks. Existing mobile network plans can result in some users ending a monthly plan with excess data, while others may have to pay a costly fee to buy more data. Thus we suggest data auctions with a single seller, or a multiple-seller networked data auction, that operate in mobile social networks, to deal with the asymmetry between extra unused data resources and urgent data demands. Based on earlier work on the analysis of auctions, we design the data transaction mechanism, and summarise the analysis on state transmission, stationary probabilities of the system, and the expected income for data sellers. To improve the efficiency and performance of the system, socially- aware mobility models are also proposed. The proposed data auction mechanisms and friendship-based mobility model are then simulated as operating on Flickr, a real-world online social network database. Results show that the number of data bidders in different auctions can be balanced through the proposed mobility model, and also increase the income per unit time of sellers in the networked data auction.

Recent work demonstrated the value of multi clusters of spiking Random Neural Networks (RNN) with dense soma-to-soma interactions in deep learning. In this paper we go back to the original simpler structure and we investigate the power of single RNN cells for deep learning. First, we consider three approaches with the single cells, twin cells and multi-cell clusters. This first part shows that RNNs with only positive parameter can conduct convolution operations similar to those of the convolutional neural network. We then develop a multi-layer architecture of single cell RNNs (MLSRNN), and show that this architecture achieves comparable or better classification at lower computation cost than conventional deep-learning methods.

Task allocation systems in the Cloud have been recently proposed so that their performance is optimised in real-time based on reinforcement learning with spiking Random Neural Networks (RNN). In this paper, rather than reinforcement learning, we suggest the use of multi-layer neural network architectures to infer the state of servers in a dynamic networked Cloud environment, and propose to select the most adequate server based on the task that optimises Quality of Service. First, a procedure is presented to construct datasets for state classification by collecting time-varying data from Cloud servers that have different resource configurations, so that the identification of server states is carried out with supervised classification. We test four distinct multi-layer neural network architectures to this effect: multi-layer dense clusters of RNNs (MLRNN), the hierarchical extreme learning machine (H-ELM), the multi-layer perceptron, and convolutional neural networks. Our experimental results indicate that server-state identification can be carried out efficiently and with the best accuracy using the MLRNN and H-ELM.

In heterogeneous ultra-dense networks (HetUDNs), the software-defined wireless network (SDWN) separates resource management from geo-distributed resources belonging to different service providers. A centralized SDWN controller can manage the entire network globally. In this paper, we focus on mobile traffic offloading and resource allocation in SDWN-based HetUDNs, constituted of different macro base stations and small-cell base stations (SBSs). We explore a scenario where SBSs' capacities are available, but their offloading performance is unknown to the SDWN controller: this is the information asymmetric case. To address this asymmetry, incentivized traffic offloading contracts are designed to encourage each SBS to select the contract that achieves its own maximum utility. The characteristics of large numbers of SBSs in HetUDNs are aggregated in an analytical model, allowing us to select the SBS types that provide the off-loading, based on different contracts which offer rationality and incentive compatibility to different SBS types. This leads to a closed-form expression for selecting the SBS types involved, and we prove the monotonicity and incentive compatibility of the resulting contracts. The effectiveness and efficiency of the proposed contract-based traffic offloading mechanism, and its overall system performance, are validated using simulations.

In Software Defined Networks (SDN), intensive traffic monitoring is used to optimize the Quality-of-Service (QoS) of the network paths which are selected. Thus, we introduce the use of the Cognitive Packet Network (CPN) algorithm to SDN in order to optimize the search for new high-QoS paths. We install the CPN algorithm in the Cognitive Routing Engine (CRE), a new application software for SDN, and show that with limited monitoring overhead we are able to determine the near-optimal paths for given QoS metrics that may be proposed by the end users. Measurements that we have conducted on an experimental replica of the GEANT network that our approach uses close to 10 times less monitoring data than conventional SDN, but that we are able to approach the optimal paths within 2%.

This paper introduces techniques for Deep Learning in conjunction with spiked random neural networks that closely resemble the stochastic behaviour of biological neurons in mammalian brains. The paper introduces clusters of such random neural networks and obtains the characteristics of their collective behaviour. Combining this model with previous work on extreme learning machines, we develop multilayer architectures which structure Deep Learning Architectures a a “front end” of one or two layers of random neural networks, followed by an extreme learning machine. The approach is evaluated on a standard - and large - visual character recognition database, showing that the proposed approach can attain and exceed the performance of techniques that were previously reported in the literature.

Energy harvesting has recently attracted much interest due to the emergence of the Internet of Things, and the increasing need to operate wireless sensing devices in challenging environments without much human intervention and maintenance. This paper presents a novel approach for modeling the performance of an energy harvesting wireless sensor node, which takes into account fluctuations in the amount of energy extracted from the environment, energy loss due to battery leakage, as well as the energy cost of sensing, data processing and communication. The proposed approach departs from the common queueing-theoretic framework used in the literature, and instead uses Brownian motion to represent more accurately the time evolution of the distribution of the node's battery level. The paper derives some performance measures of interest along with the stationary solution of the system, and discusses possible directions for reducing the number of parameters and states of the model without compromising accuracy.

More and more businesses are deploying their application(s) with different cloud providers which are close to their customers in order to provide better Quality of Service (QoS) to their end customers. In this work, an optimized and secure software-defined overlay network is proposed as an efficient mechanism to interconnect these geographically-dispersed applications compared to using only plain IP routing. A logically centralized Cognitive Routing Engine (CRE), based on Random Neural Networks with Reinforcement Learning, was developed to find with minimal monitoring overhead the optimal overlay paths when the public Internet is used as the communication means between the overlay nodes. CRE was evaluated by using an overlay network composed of hosts from 5 different public clouds where it was shown that the latency of CRE paths is most of the time within 5% of the latency of the optimal IP paths. Furthermore, it was also demonstrated that CRE is able to do asymmetric path optimization where the forward path is different from the reverse path for a given data centre pair in order to further improve QoS.

Cloud systems include both locally based servers at user premises and remote servers and multiple Clouds that can be reached over the Internet. This paper describes a smart distributed system that combines local and remote Cloud facilities. It operates with a task allocation system that takes decisions to allocate tasks dynamically to the service that offers the best overall Quality of Service and a routing overlay which optimizes network delay for data transfer between clouds. Internet-scale experiments exhibit the effectiveness of our approach in adaptively distributing workload across multiple clouds.

This paper uses big data and machine learning for the real-time management of Internet scale quality-of-service (QoS) route optimisation with an overlay network. Based on the collection of data sampled every 2 min over a large number of source-destinations pairs, we show that intercontinental Internet protocol (IP) paths are far from optimal with respect to QoS metrics such as end-to-end round-trip delay. We, therefore, develop a machine learning-based scheme that exploits large scale data collected from communicating node pairs in a multihop overlay network that uses IP between the overlay nodes, and selects paths that provide substantially better QoS than IP. Inspired from cognitive packet network protocol, it uses random neural networks with reinforcement learning based on the massive data that is collected, to select intermediate overlay hops. The routing scheme is illustrated on a 20-node intercontinental overlay network that collects some 2 × 106measurements per week, and makes scalable distributed routing decisions. Experimental results show that this approach improves QoS significantly and efficiently.

The increase of the number of smart devices using mobile networks' services is followed by the increase of the number of security threats for mobile devices, generating new challenges for mobile network operators. Signalling attacks and storms represent an emerging type of distributed denial of service (DDoS) attacks and happen because of special malware installed on smart devices. These attacks are performed in the control plane of the network, rather than the data plane, and their goal is to overload the Signalling servers which leads to service degradation and even network failures. This paper proposes a detection and mitigation mechanism of such attacks which is based on counting repetitive bandwidth allocations by mobile terminals and blocking the misbehaving ones. The mechanism is implemented in our simulation environment for security in mobile networks SECSIM. The detector is evaluated calculating the probabilities of false positive and false negative detection and is characterised by very low negative impact on un-attacked terminals. Simulation results using joint work of both detector and mitigator, are shown for: the number of allowed attacking bandwidth allocations, end-to-end delay for normal users, wasted bandwidth and load on the Signalling server. Results suggest that for some particular settings of the mechanism, the impact of the attack is successfully lowered, keeping the network in stable condition and protecting the normal users from service degradations.

We review the characteristics of signalling storms that have been caused by certain common apps and recently observed in cellular networks, leading to system outages. We then develop a mathematical model of a mobile user's signalling behaviour which focuses on the potential of causing such storms, and represent it by a large Markov chain. The analysis of this model allows us to determine the key parameters of mobile user device behaviour that can lead to signalling storms. We then identify the parameter values that will lead to worst case load for the network itself in the presence of such storms. This leads to explicit results regarding the manner in which individual mobile behaviour can cause overload conditions on the network and its signalling servers, and provides insight into how this may be avoided.

Mobile telephony based on UMTS uses finite-state control schemes for wireless channels and for signaling across the network. These schemes are used systematically in various phases of the communication and are vulnerable to attacks that can bring down the network through unjustified bandwidth allocation and excessive signaling across the control plane. In this paper we identify those system parameters which are critical to the success of such attacks, and propose changes that can limit the effect of the attack. The approach is based on establishing a mathematical model of a UMTS system that is undergoing attacks, andon showing how parameters can be optimally modified to minimise the effect of the attack as experienced by he mobile device and the network.

Storage nodes are expected to be placed as an intermediate tier of large scale sensor networks for caching the collected sensor readings and responding to queries with benefits of power and storage saving for ordinary sensors. Nevertheless, an important issue is that the compromised storage node may not only cause the privacy problem, but also return fake/incomplete query results. We propose a simple yet effective dummy reading-based anonymization framework, under which the query result integrity can be guaranteed by our proposed verifiable top-$k$query (VQ) schemes. Compared with existing works, the VQ schemes have a fundamentally different design philosophy and achieve the lower communication complexity at the cost of slight detection capability degradation. Analytical studies, numerical simulations, and prototype implementations are conducted to demonstrate the practicality of our proposed methods.

Network testbed experimentation is useful to evaluate new protocols, since it offers realism and repeatability under controllable conditions. Thus, in this paper, we make use of a software network platform, the cognitive packet network (CPN), to offer best-effort quality of service (QoS) to end users and to develop a new bilateral QoS differentiation between pairs of communicating nodes. In the proposed approach, each CPN edge or user node is a source and a destination at the same time, managing uplink user-originated traffic and downlink traffic sent back in response to the uplink. The bilateral communication is implemented with four distinct QoS objectives that can be met between sender nodes (original source or destination). Traffic volume asymmetry between the received and the sent data is used to trigger changes in QoS. The lower traffic rate requires short-delay QoS, whereas the higher traffic rate requires loss minimization. The effectiveness of the approach is evaluated by several measurements.

We introduce two congestion metrics to guide emergency evacuations using sensing and local area networks, that use current congestion and congestion forecasts, together with the Cognitive Packet Network (CPN), a routing algorithm which intelligently discovers paths. Using simulations we find that CPN performs well for emergency evacuation with such metrics, and also reveal the dynamics that these schemes can create.

We present two spatially distributed computing systems that operate in a building and provide intelligent navigation services to people for evacuation purposes. These systems adapt to changing conditions by monitoring the building and using local communication and computation for determining the best evacuation paths. The first system, called distributed evacuation system (DES), comprises a network of decision nodes (DNs) positioned at specific locations inside the building. DNs provide people with directions regarding the best available exit. The second system, called opportunistic emergency support system (OESS), consists of mobile communication nodes (CNs) carried by people. CNs form an opportunistic network in order to exchange information regarding the hazard and to direct the evacuees towards the safest exit. Both DES and OESS employ sensor nodes deployed at fixed locations for monitoring the hazard. We evaluate the spatial systems using simulation experiments with a purpose-built emergency simulator called DBES. We show how parameters such as the frequency of information exchange and communication range affect the system performance and evacuation outcome.

We have developed RNNI, a global alignment method for protein-protein interaction networks between species, using a random neural network model (RNN) tailored for the alignment problem. The benchmark of the method in comparison with other available alignment approaches was performed using a range of measurements. The alignment results of the human and yeast pair showed that RNNI is capable of generating alignments with large conserved networks with functionally-related protein pairs while maintaining the closeness to the naive- sequence homology approach (BLAST).

This paper briefly discusses cyper-physical systems that include human beings and vehicles in a built environment such as a building or a city, together with Sensor Networks, Communications and Decision Support Systems, with the purpose of optimising the human outcome in the case of an emergency.

Emergency management systems (EMS) are important and complex examples of Cyber-Physical-Human systems that are deployed so as to optimise the outcome of an emergency from a human perspective. They use sensor networks, networked decision nodes and communications with evacuees and first responders to optimise the overall Quality of Service to benefit primarily human beings in terms of survival, health and safety, and the the protection of nature, property and valuable infrastructures. The use of technology for emergency management also has side effects in terms of failures and malicious attacks of the ICT system, so that the outcome will be affected by how well the ICT system operates under stress. Thus this paper surveys research on wireless sensor- assisted EMS, including networking, distributed control, and knowledge discovery. An evaluation of increased effectiveness and liabilities that wireless communications introduce is conducted when adversaries exacerbate the emergency by malicious attacks through the wireless system.

This paper studies whether a packet will ultimately succeed in reaching a given destination, how long this will take and how much energy may be expended, in the context of a network with imperfect routing tables and non-homogeneous network characteristics. It also investigates the effect of non-cooperative routers that may actually choose to drop certain packets if they view them to be dangerous for destination nodes, as when packets may be carrying worms, viruses or malware, and when certain packets have been identified as being part of a Denial of Service attack. The approach we take is to construct a probability model for packet travel from a source to destination node in a large non-homogeneous multiple hop network. The randomness models the lack of precise routing information at each of the network hops, and randomness in routing can also be used to model networks where one wishes to explore alternate paths in a network to discover the more reliable paths, or those that may have other desirable characteristics such as lower delay or lower packet loss. We assume that each packet has the same time out: when the time-out elapses, the packet is dropped if it has not yet reached the destination, and some time later the source will retransmit a duplicate packet. A numerical–analytical solution is developed to compute the average travel time of the packet from source to destination and to estimate its total energy consumption. Two applications of these results are then presented. In the first one, the packet is an ‘attack’ packet (e.g. a Denial of Service packet, or some malware) and as it approaches the destination node it is being frequently inspected by routers that may decide to drop it if they correctly detect that it is a threat. The second example considers a wireless network where areas which are remote from the source and destination nodes have poorer wireless coverage so that packet losses become more frequent as the packet ‘unknowingly’ (due to poor routing tables errors) meanders away from the main coverage area. Other applications in wireless networks are also provided and a simulation study is performed to validate the analytical model.

This paper reviews computation in natural systems, focusing mainly on biology and citing examples of the computation that is inherent in chemistry, natural selection, gene regulatory networks, and neuronal systems.

Gene expression models play a key role to understand the mechanisms of gene regulation whose aspects are grade and switch-like responses. Though many stochastic approaches attempt to explain the gene expression mechanisms, the Gillespie algorithm which is commonly used to simulate the stochastic models requires additional gene cascade to explain the switch-like behaviors of gene responses. In this study, we propose a stochastic gene expression model describing the switch-like behaviors of a gene by employing Hill functions to the conventional Gillespie algorithm. We assume eight processes of gene expression and their biologically appropriate reaction rates are estimated based on published literatures. We observed that the state of the system of the toggled switch model is rarely changed since the Hill function prevents the activation of involved proteins when their concentrations stay below a criterion. In ScbA-ScbR system, which can control the antibiotic metabolite production of microorganisms, our modified Gillespie algorithm successfully describes the switch-like behaviors of gene responses and oscillatory expressions which are consistent with the published experimental study.

Gene regulatory networks can provide the systematic view of a complex living system. However, constructing large-scale gene regulatory networks is one of the challenging problems in Systems biology. Also a burst of various biological data demands a proper integration technique for reliable gene regulatory network construction. Here we propose a new re- verse engineering approach based on Bayesian model averaging technique which ensembles all appropriate regression models. This Bayesian approach with a prior having Gibbs distribution provides a convenient way to integrate multiple sources of biological data. In the simulation study with maximum 2000 genes, our method shows better sensitivities than elastic-net and Gaussian graphical models with a fixed specificity 0.99 though the computation time is not as good as the others. Three large-scale networks consisting of 4257 genes are built using the gene expression data of non-tumor, grade III tumor, and grade IV tumor samples, respectively. We found that genes having a large variance of degree distribution among the three tumor networks are mostly involved in regulatory and immunology of biological processes, which can provide different information from that of a differentially expressed gene analysis.

This paper considers a probability model for travel of a packet from a source node to a destination node in a large non-homogeneous multiple hop network with unreliable routing tables. Use of a random model is justified by the lack of precise information that can be used in each step of the packet's travel, and randomness can also be useful in exploring alternate paths when a long sequence of hops has not resulted in the packet's arrival to the destination. The packet's travel may also be impeded if certain routers on its path prove to be unreliable, or the packet may be dropped from a buffer or destroyed due to packet loss. The packet also has a limited time-out that allows the source to retransmit a dropped or lost packet. Because the network itself may be extremely large, we consider packet travel in an infinite random non-homogeneous medium, with events that may interrupt, destroy or stop the packet from moving towards its destination. We derive a numerical-analytical solution allowing us to compute the average travel time of the packet from source to destination, as well as to estimate its energy consumption. Two interesting applications are then presented. In the first one a wireless network where areas which are remote from the source and destination nodes may have poor wireless coverage so that the packet losses become more frequent as the packet "unknowingly" (due to poor routing tables for instance) meanders away from the source and destination node. The second application is related to defending a destination node against attacks that take the form of packets that carry a virus or a worm that can be detected via deep packet inspection at intermediate nodes, and as the packet approaches the destination node it is more frequently inspected and dropped if it is a threat.

Although there is great interest in reducing energy consumption for all areas of human activity, many of the proposed approaches such as the smart home or the smart grid are actually prone to an increase in the use of Information and Communication Technologies (ICT) which itself is a big consumer of energy. It is remarkable that ICT's carbon imprint is of the order of 2% of the world total, comparable to the carbon imprint of air travel. Thus, it is imperative to address energy savings in ICT and, in particular, in data centres and networks. This paper follows up on our previous work that seeks novel ways to reduce the energy consumption in packet networks which constitute the backbone of the Internet and of the information society as a whole. Here we discuss the use of routing control as a means to reduce energy consumption while remaining aware of QoS considerations, and propose a method that uses a queueing theoretic analysis and optimization technique to distribute traffic so as to reduce a cost function that comprises both energy and QoS. A method using G-networks is developed to incorporate both the effect of user traffic and the overhead in QoS and energy consumption introduced by the control traffic that will be needed to carry out the re-routing decisions. For an N-node network, we show that this approach results in an algorithm which has O(N3) time complexity. Because this approach may be too costly in computational overhead and delays, we also propose another approach that uses load balancing and which would be much simpler to implement.

Gene expression models play a key role to understand the mechanisms of gene regulation whose aspects are grade and switch-like responses. Though many stochastic approaches attempt to explain the gene expression mechanisms, the Gillespie algorithm which is commonly used to simulate the stochastic models hardly explain the switch-like behaviors of gene responses. In this study, we propose a stochastic gene expression model which can describe the switch-like behaviors of gene responses by employing Hill functions to the conventional Gillespie algorithm. We assume eight processes of gene expression and their biologically appropriate reaction rates are estimated based on published literatures. Our negative regulatory model shows that the modified Gillespie algorithm successfully describes the switch-like behaviors of gene responses, which is consistent with a published experimental study. We observe that the state of the system of the toggled switch model is rarely changed since the Hill function prevents the activation of involved proteins when their concentrations stay at low level. In ScbA/ScbR system which can control the antibiotic metabolite production of microorganisms, our proposed stochastic approach successfully models its switch-like gene response and oscillatory expressions.

The need for network stability and reliability has led to the growth of autonomic networks that can provide more stable and more reliable communications via on-line measurement, learning and adaptation. A promising architecture is the Cognitive Packet Network (CPN) that rapidly adapts to varying network conditions and user requirements using QoS driven reinforcement learning algorithms that drive the routing control. Contrary to conventional mechanisms, the users rather than the nodes, control the routing by specifying their desired QoS requirements (QoS Goals), such as Minimum Delay, Maximum Bandwidth, Minimum Cost, etc., and the network then routes each user's traffic individually based on their specific needs and on a "glocal" view. In CPN the user has the ability to explore the network for its own needs, and evaluate its own impact on the network as a whole and vice-versa, and then take appropriate decisions. CPN routing has been evaluated extensively under normal operating conditions and has proven to be very adaptive to network changes such as congestion. Here we show how CPN can respond and survive to catastrophic node failures caused by the spread of network worms. This survival is based on two complementary approaches that are run concurrently: one the one hand, each user attempts to concurrently and adaptively avoid paths which are infected, and secondly patching algorithms are continuously run to repair the network. Experiments show that this approach assures the stability of network communications throughout the course of an attack.

We consider the problem of optimizing the performance of a network coding router with two stochastic flows. We develop a queueing model which accounts for the fact that coding is not performed when packets are transmitted, but is done by a separate program or hardware which operates independently of the hardware that sends packets out over links. We formulate and solve a constrained optimization problem which provides the optimal time that the router should wait before sending the information that it has uncoded, so that the average response time of the system is minimized. The trade-offs between delay and bandwidth or energy associated with the choice of the waiting time are also investigated, and the results indicate that network coding offers significant performance gains in a moderate to heavily loaded system.

Energy efficiency is increasingly important for future information and communication technologies (ICT), because the increased usage of ICT, together with increasing energy costs and the need to reduce green house gas emissions call for energy-efficient technologies that decrease the overall energy consumption of computation, storage and communications. Cloud computing has recently received considerable attention, as a promising approach for delivering ICT services by improving the utilization of data centre resources. In principle, cloud computing can be an inherently energy-efficient technology for ICT provided that its potential for significant energy savings that have so far focused on hardware aspects, can be fully explored with respect to system operation and networking aspects. Thus this paper, in the context of cloud computing, reviews the usage of methods and technologies currently used for energy-efficient operation of computer hardware and network infrastructure. After surveying some of the current best practice and relevant literature in this area, this paper identifies some of the remaining key research challenges that arise when such energy-saving techniques are extended for use in cloud computing environments.

We describe a distributed simulation tool which addresses the unique needs for the simulation of emergency response scenarios. The simulation tool adopts the multi-agent paradigm, so as to facilitate the modelling of diverse and autonomous agents, and it provides mechanisms for the interaction of the entities that are being simulated. It operates in a distributed fashion to reduce the simulation time required for such large-scale systems. The simulation tool represents the individuals that need to be evacuated, the resources that contribute to the evacuation including human rescuers, and other active resources and entities which may include robots and which can autonomously interact with the environment and with each other and take individual or collaborative decisions. We illustrate the tool with an application and compare the results for both centralized and distributed execution. Our results also show the significant reduction in execution time that is achieved for different degrees of distribution of the simulator on multiple servers.

The Computer Journal was established over 50 years ago to record and make publicly available the most significant scientific and technical developments that were occurring in the field that was later to be called Computer Science.It is now published through a joint effort of two prestigious organizations, the British Computer Society and Oxford University Press.Up to the end of 2007, when I assumed the duties of Editor-in-Chief, our Journal published 60 or so papers per year in the form of six bimonthly issues, selected from some 200 submissions per year covering all areas of Computer Science.Our Journal now receives over 430 paper submissions per year and has undergone major growth and change; thus the purpose of this editorial is to inform our readers and our more than 4000 subscribers worldwide about these changes and developments.Since 2007, usage of the The Computer Journal as measured by full text online downloads has increased by nearly one-third.In the same period, the Impact Factor has risen by more than two-thirds.It is heartening to see that with the dedication of capable staff at the supporting organizations, and with better international coverage and a broader choice of topics, the Journal's subscription income has increased (rather than been merely maintained) despite difficult economic conditions.Computer Science, or 'Informatics' as it is often called in Europe, deals with the design, evaluation and optimization of information processing systems that are built out of microelectronic, optical and magnetically based hardware units for processing and storage, and of the complex software systems that are designed to manage these hardware components.The aim is to offer human users and virtual applications the facilities and services that are needed, with high levels of reliability and Quality of Service.Our Journal serves the computer science research, academic and advanced engineering communities, which now total hundreds of thousands of researchers and educators worldwide.Indeed, some 90% of our paper submissions come from outside the UK, and over 97% of our institutional subscribers are also abroad.Thus, although The Computer Journal is based in the UK, through its authors, readers and editorial board it is essentially an international enterprise.Yet ideas and developments that have originated in the UK have had a major impact in Computer Science worldwide, and not just via the initial genius of individuals such as Alan

Erol Gelenbe, Stephen Gilmore; Discussants’ Comments on the Computer Journal Lecture by Peter Harrison presented at the British Computer Society on 24th Fe

Sensor networks (SNs) consist of spatially distributed sensors which monitor an environment, and which are connected to some sinks or backbone system to which the sensor data is being forwarded. In many cases, the sensor nodes themselves can serve as intermediate nodes for data coming from other nodes, on the way to the sinks. Much of the traffic carried by SNs will originate from routine measurements or observations by sensors that monitor a particular situation, such as the temperature and humidity in a room or the infrared observation of the perimeter of a house, so that the volume of routine traffic resulting from such observations may be quite high. When important and unusual events occur, such as a sudden fire breaking out or the arrival of an intruder, it will be necessary to convey this new information very urgently through the network to a designated set of sink nodes where this information can be processed and dealt with. This paper addresses the important challenge by avoiding the routine background traffic from creating delays or bottlenecks that impede the rapid delivery of high priority traffic resulting from the unusual events. Specifically we propose a novel technique, the ‘Randomized Re-Routing Algorithm (RRR)’, which detects the presence of novel events in a distributed manner, and dynamically disperses the background traffic towards secondary paths in the network, while creating a ‘fast track path’ which provides better delay and better quality of service (QoS) for the high priority traffic which is carrying the new information. When the surge of new information has subsided, this is again detected by the nodes and the nodes progressively revert to best QoS or shortest-path routing for all the ongoing traffic. The proposed technique is evaluated using a mathematical model as well as simulations, and is also compared with a standard node by a node priority scheduling technique.

We investigate the assignment of assets to tasks where each asset can potentially execute any of the tasks, but assets execute tasks with a probabilistic outcome of success. There is a cost associated with each possible assignment of an asset to a task, and if a task is not executed, there is also a cost associated with the non-execution of the task. Thus, any assignment of assets to tasks will result in an expected overall cost which we wish to minimize. We formulate the allocation of assets to tasks in order to minimize this expected cost, as a nonlinear combinatorial optimization problem. A neural network approach for its approximate solution is proposed based on selecting parameters of a random neural network (RNN), solving the network in equilibrium, and then identifying the assignment by selecting the neurons whose probability of being active is the highest. Evaluations of the proposed approach are conducted by comparison with the optimum (enumerative) solution as well as with a greedy approach over a large number of randomly generated test cases. The evaluation indicates that the proposed RNN-based algorithm is better in terms of performance than the greedy heuristic, consistently achieving on average results within 5% of the cost obtained by the optimal solution for all problem cases considered. The RNN-based approach is fast and is of low polynomial complexity in the size of the problem, while it can be used for decentralized decision making.

Prof. Erol Gelenbe PhD DSc FIEEE FACM FIET is a Member of the French National Academy of Engineering and of the Turkish Academy of Sciences. He holds the “Dennis Gabor Chair” and is Head of the Intelligent Systems and Networks Group in the Electrical and Electronic Engineering Department at Imperial College, where he conducts research on computer systems and networks. Author of over 140 journal papers, and several books published in English, French, Japanese and Korean, he has won numerous major international awards and honours for his work, including the ACM SIGMETRICS LifeTime Achievement Award, the Honoris Causa Doctorate from three universities: Liege in Belgium (2006), Bogazici in Istanbul (2004) and Rome in Italy (1996). His recent work includes path finding algorithms in noisy and uncertain conditions, the use of neural networks to control routing in computer networks, the analysis of decision making based on market based techniques, and modeling problems from the basic sciences including neural networks, gene regulatory networks, and models of chemical reactions. His research is currently funded by EPSRC and MoD with significant industry participation (BT, BAE Systems and QinetiQ), and by the EU FP7 programme. Appointed to his first chair at the age of 27 at the University of Liege in Belgium, he served as a research director at INRIA (France), with successive professorial posts at the University of Paris, Duke University where he was Department Head, and the University of Central Florida where he held a distinguished chair and was Associate Dean of Engineering. He is the Editor in Chief of The Computer Journal, and editor of the Proc. of the Royal Society A, Performance Evaluation, and Acta Informatica. He has graduated over 50 PhDs. 2009 International Conference on Advanced Information Networking and Applications

We develop analytical and simulation models to evaluate the additional delay at intermediate nodes of a store and forward packet network when Network Coding (NC) is applied. The approach is based on the analysis of queueing systems with specific service processes that capture the effect of NC. The analytical results are compared with simulations.

Environmental monitoring is one of the most popular applications in wireless sensor networks. Although it is important to obtain a continuous record of the environment, users may gain enough information without receiving every routine sensor measurement. Reducing unnecessary traffic allows better utilization of the network resources. However, it is uneasy to decide on when and what kind of traffic to reduce in a dynamically changing environment. In this paper, we study the problem of information-aware traffic reduction for wireless sensor networks. We propose a two-step information-aware traffic reduction algorithm to address this problem. First, we provide a distributed and real-time algorithm for sensors to classify their measurements and report them selectively based on the importance of information. Then, we propose a bandwidth allocation algorithm to assign different forwarding probabilities to packets considering both the information quality and the network load. Our algorithm can be implemented and integrated easily with existing routing protocols to maximize the quality of information to the users, while reducing the amount of network traffic. We evaluate our approach based on the real sensing measurements to demonstrate the quality of information achieved. Simulations are also conducted to evaluate the performance of our proposed scheme in a larger network.

Web-based computerized auctions are increasingly present in the Internet. We can imagine that in the future this trend will actually be extended to situations where virtual buyer and seller agents will conduct automated transactions across the network, and that large sectors of the economy may be strucured in this manner. The purpose of this article is to model automated bidders and sellers which interact through a network. We model the bidding process as a random arrival process while the price attained by a good is modeled as a discrete random variable. We obtain analytical solutions allowing us to compute the income from a single auction, or the income per unit time from a repeated sequence of auctions. A variety of single-auction models are studied, including English and Vickrey auctions, and the income per unit time is derived as a function of other parameters, including the rate of arrival of bids, the seller's decision time, the value of the good, and the “rest time” of the seller between successive auctions. We illustrate the results via numerical examples. We also introduce a model for networked auctions where bidders can circulate among a set of interconnected auctions which we call the Mobile Bidder Model (MBM). We obtain an analytical solution for the MBM under the assumption,which we call the “active bidders assumption,” that activities that are internal to an auction (bids and sales) are much more frequent than changes that occur in the number of bidders at each auction.

In this paper we examine the quality of information (QoI) at the output of a real wireless sensor network by considering the difference between the monitored environment and the interpreted data produced by the network. Using practical examples in an experimental setting, we hope to shed light on the concept of QoI and on the manner of estimating and evaluating it. We use a real wireless network in combination with simulated events, to help us formulate and understand the concept of QoI and its associated technical questions. Using algorithms such as trilateration and clustering to interpret the outputs of the sensor network, we explore several definitions of QoI, including the peak signal to noise ratio. Furthermore we investigate the impact that different packet transmission approaches have on the QoI. We show that QoI is time-varying, and that in-network processing allows QoI levels to be maintained while reducing network load.

Wireless sensor networks (WSN) can report large volumes of slowly varying routine data, while important or significant events can be relatively rare. An important challenge is then to offer the significant or unusual data an adequate routing policy that will allow it to rapidly reach the sink nodes, despite the large volume of routine packets in the network. In this paper we introduce randomized re-routing (RRR), to detect the unusual events in a distributed manner, and dynamically transfer routine data packets to secondary paths in the network, while offering a fast track path with better QoS for the packets carrying unusual data. In this paper we describe the RRR algorithm and evaluate it with extensive simulations.

We describe a prototype of trading system platform populated by agents who autonomously decide to buy and/or sell items according to a set of local needs which arise dynamically (also by possibly accessing information provided by pervasive devices) by in the process of fulfilling a given overall utility. The market has combinatorial nature in a way that items to be traded are combined into packages, in accordance with a principle that drives the nature of many current markets. However, differently from these, items belong to a number of distinct sellers distributed in the platform, and are chosen singularly on the basis of buyers preferences and needs. Agents are thus situation-aware, with sellers coming acquainted of the market demand, and buyers price offers, through a Knowledge Network. This latter drives the way market offers balance the demand by gathering the needed information in an autonomous way and taking advantage of pervasive devices. Packaging is realized by agent aggregation into Virtual Sellers, in an autonomous fashion, and we propose an opportunistic policy whereby aggregation is governed by a Combinatorial Auction. The market is studied through proof-of-concept simulation, where the efficiency deriving from the opportunistic aggregation based on Combinatorial Auctions and the influence of contextual self-awareness are studied.

The study of artificial neural networks has originally been inspired by neurophysiology and cognitive science. It has resulted in a rich and diverse methodology and in numerous applications to machine intelligence, computer vision, pattern recognition and other applications. The random neural network (RNN) is a probabilistic model which was inspired by the spiking behaviour of neurons, and which has an elegant mathematical treatment that provides both its steady-state behaviour and offers efficient learning algorithms for recurrent networks. Second-order interactions, where more than one neuron jointly act upon other cells, have been observed in nature; they generalize the binary (excitatory–inhibitory) interaction between pairs of cells and give rise to synchronous firing (SF) by many cells. In this paper, we develop an extension of the RNN to the case of synchronous interactions, which are based on two cells that jointly excite a third cell; this local behaviour is in fact sufficient to create SF by large ensembles of cells. We describe the system state and derive its stationary solution as well as a O(N3) gradient descent learning algorithm for a recurrent network with N cells when both standard excitatory–inhibitory interactions, as well as SF, are present.

Department of Computing, Imperial College London, London, UK. Email: [email protected] A resurgence in product-forms Interest in the stochastic behaviour of queueing networks began in the 1960s with the work by Jackson, Gordon, Newell and others [1–3]. These authors derived the so-called product-form solutions for the equilibrium probabilities of the joint state in such a continuous time Markov chain (CTMC). From the preceding lecture, of course, these networks are special cases of G-networks, where there are no negative customers, triggers, signals, etc. After a number of generalizations of Jackson networks in the 1970s and 1980s, it was thought by many that a product-form for a stochastic network would only exist if that network satisfied a condition called partial or local balance [4, 5]. Essentially, this specifies a specific way by which the global balance equations of probability fluxes into, and out of, a given state (the steady-state theorem for CTMCs) are constructed from balanced subsets of these equations, each subset relating to a component of the network. Gelenbe's G-network model [6–14], and the related RNN model [15], with its non-linear traffic equations in particular, provided counter-examples to this widely held view. In turn, this sparked a resurgence of interest in the quest for product-forms, which continues today.

Large-scale distributed systems, such as natural neuronal and artificial systems, have many local interconnections, but they often also have the ability to propagate information very fast over relatively large distances. Mechanisms that enable such behavior include very long physical signaling paths and possibly saccades of synchronous behavior that may propagate across a network. This letter studies the modeling of such behaviors in neuronal networks and develops a related learning algorithm. This is done in the context of the random neural network (RNN), a probabilistic model with a well-developed mathematical theory, which was inspired by the apparently stochastic spiking behavior of certain natural neuronal systems. Thus, we develop an extension of the RNN to the case when synchronous interactions can occur, leading to synchronous firing by large ensembles of cells. We also present an O(N3) gradient descent learning algorithm for an N-cell recurrent network having both conventional excitatory-inhibitory interactions and synchronous interactions. Finally, the model and its learning algorithm are applied to a resource allocation problem that is NP-hard and requires fast approximate decisions.

Smart networks have grown out of the need for stable, reliable, and predictable networks that will guarantee packet delivery under Quality of Service (QoS) constraints. In this article we present a measurement-based admission control algorithm that helps control traffic congestion and guarantee QoS throughout the lifetime of a connection. When a new user requests to enter the network, probe packets are sent from the source to the destination to estimate the impact that the new connection will have on the QoS of both the new and the existing users. The algorithm uses a novel algebra of QoS metrics, inspired by Warshall's algorithm, to look for a path with acceptable QoS values to accommodate the new flow. We describe the underlying mathematical principles and present experimental results obtained by evaluating the method in a large laboratory test-bed operating the Cognitive Packet Network (CPN) protocol.

We study gene regulatory networks that include the effect of multiple binary interactions, when various genes can activate or inhibit other genes, as well as of ternary interactions where two genes jointly affect the level of activity of a third gene. We also prove that Boolean identities for the activation state of a gene in terms of the activation level of another set of genes, can be represented, so that the overall probability of activation of any gene in the network can be computed in the presence of logical dependencies, and excitatory-inhibitory interactions between genes.

Denial of service (DoS) is a prevalent threat in today's networks. While such an attack is not difficult to launch, defending a network resource against it is disproportionately difficult, and despite the extensive research in recent years, DoS attacks continue to harm. The first goal of any protection scheme against DoS is the detection of its existence, ideally long before the destructive traffic build-up. In this paper we propose a generic approach which uses multiple Bayesian classifiers, and we present and compare four different implementations of it, combining likelihood estimation and the random neural network (RNN). The RNNs are biologically inspired structures which represent the true functioning of a biophysical neural network, where the signals travel as spikes rather than analog signals. We use such an RNN structure to fuse real-time networking statistical data and distinguish between normal and attack traffic during a DoS attack. We present experimental results obtained for different traffic data in a large networking testbed.

Adaptive routing is once again becoming of interest because of the possibility to couple on-line probing in networks with real-time dynamic and distributed control of paths and flows. Wireless networks, with their rapidly changing network conditions also create a need to revisit this issue. This paper uses measurements in a wired adaptive network test-bed, the cognitive packet network (CPN), to investigate the pros and cons of adaptive routing. CPN routes packet flows through a store and forward network according to their quality of service (QoS) needs through an on-line, distributed reinforcement learning mechanism. This paper investigates routing oscillations which occur due to the interaction of multiple flows and studies their effect on QoS in the context of CPN. Our results indicate that routing oscillations can be easily controlled by randomising the route switching, and that from an overall QoS viewpoint increased switching can also lead to improved performance.

Adaptive routing is once again becoming of interest because of the possibility to couple on-line probing in networks with real-time dynamic and distributed control of paths and flows using Reinforcement Learning. Wireless networks, with their rapidly changing network conditions also create a need to revisit this issue. This paper uses measurements in a wired bio-inspired adaptive network test-bed, the Cognitive Packet Network (CPN), to investigate the pros and cons of adaptive routing. CPN routes packet flows through a store and forward network according to their Quality of Service (QoS) needs through an on-line, distributed reinforcement learning mechanism that incorporates a biologically-inspired Neural Network model for making routing decisions. This paper investigates routing oscillations which occur due to the interaction of multiple flows and studies their effect on QoS in the context of CPN. Our results indicate that routing oscillations can be easily controlled by randomising the route switching, and that from an overall QoS viewpoint increased switching can also lead to improved performance.

The demand for stable and reliable networks which can offer packet delivery under Quality of Service constraints led to the development of autonomous networks that use adaptive packet routing in order to provide the best possible QoS. A mechanism which takes networks like these a step further in guaranteeing packet delivery even under strict QoS constraints, is Admission Control (AC). In this paper we describe a measurement-based admission control algorithm which tries to control the ingress traffic of a network by not allowing the entrance of connections which would negatively affect the QoS of the existing users of the network. It is a multiple QoS mechanism in which the users are the ones that specify the QoS levels they need in order to function properly. The impact that this new call will have on the QoS of the existing users, is based on measurements of probe traffic and monitoring of the network . The decision of whether to accept a new call or not is made using a novel algebra of QoS metrics, inspired by Warshall's algorithm, which searches whether there is a feasible path with enough resources to accommodate the new flow, without affecting the ongoing traffic.

This special issue includes a total of seven papers that were selected after a rigorous refereeing process from submissions solicited from the research community comprising the Defence Technology Centre on Data and Information Fusion (DIF-DTC) that includes teams from several UK companies and eight UK universities. These papers illustrate two complementary aspects of the research conducted within the DIF-DTC. On the one hand, this national programme brings together teams of researchers and industrial participants so as to address problem areas and systems of practical importance through a coordinated research effort, and through the construction of demonstrators, which serve to illustrate and evaluate a coordinated set of research activities. The second important aspect of the DIF-DTC is the existence within this framework, of many deep research contributions, which contribute significant technical input to the broader agenda. Thus, this special issue includes papers that describe the objectives and progress made with two of the DIF-DTC's broad-based but well-focused ‘cluster projects’, followed by five papers that describe specific research contributions.

The future digital battlespace will be a fast-paced and frenetic environment that stresses information communication technology systems to the limit. The challenges are most acute in the tactical and operational domains where bandwidth is severely limited, security of information is paramount, the network is under physical and cyber attack and administrative support is minimal. Hyperion is a cluster of research projects designed to provide an automated and adaptive information management capability embedded in defence networks. The overall system architecture is designed to improve the situational awareness of field commanders by providing the ability to fuse and compose information services in real time. The key technologies adopted to enable this include: autonomous software agents, self-organizing middleware, a smart data filtering system and a 3-D battlespace simulation environment. This paper reviews some of the specific techniques under development within the Hyperion sub-projects and the results achieved to date.

We consider a wireless network in which packets are forwarded opportunistically from the source towards the destination, without accurate knowledge of the direction that they should take. A Brownian motion model that includes the effect of packet losses, and subsequent retransmission after a time-out, is used to compute the average travel time of the packet. The results indicate that the average travel time is always finite provided that a time-out is used, and that there is an element of randomness in the manner in which successive nodes are being chosen. We show that the average packet travel time can be minimized by a judicious choice of the time-out, and its optimum value in turn depends on other system parameters such as packet-loss probabilities. We present simulations that illustrate the analytical results.

Many large scale autonomous systems based on a large number of interacting agents in a structured physical environment have emerged in diverse areas such as biology, ecology or finance. Inspired by the desire to better understand and make the best out of such systems, we model them in order to gain insight, predict the future and control it partially if not fully. In this paper, we present a stochastic approach to modeling such systems based on G-networks. We propose two methods which deal with cases where complete or incomplete world knowledge is available. We use strategic military planning in urban scenarios as an example to demonstrate our approach. Our results suggest that this approach tackles the problem of modeling autonomous systems at low computational cost. Apart from offering numerical estimates of various outcomes, the approach helps us identify the parameters or characteristics that have the greatest impact on the system most and allows us to compare alternative strategies

The worldwide growth in broadband access and multimedia traffic has led to an increasing need for Quality- of-Service (QoS) in networks. Real time network applications require a stable, reliable, and predictable network that will guarantee packet delivery under QoS constraints. Network self- awareness through on-line measurement and adaptivity in response to user needs is one way to advance user QoS when overall network conditions can change, while admission control (AC) is an approach that has been commonly used to reduce traffic congestion and to satisfy users' QoS requests. The purpose of this paper is to describe a novel measurement-based admission control algorithm which bases its decision on different QoS metrics that users can specify. The self-observation and self- awareness capabilities of the network are exploited to collect data that allows an AC algorithm to decide whether to admit users based on their QoS needs, and the QoS impact they will have on other users. The approach we propose finds whether feasible paths exist for the projected incoming traffic, and estimates the impact that the newly accepted traffic will have on the QoS of pre-existing connections. The AC decision is then taken based on the outcome of this analysis.

The number of nodes that our global Internet supports will rapidly move from the hundreds of millions to the hundreds of billions. Sensor networks, smart dust, on-board networks in personal vehicles, networked letters, packages and clothes, will all contribute to this growth. While elements of top-down and well organised design will still be an important element of these future complex computer systems, many systems will have to interact in an ad hoc manner in order to offer fast access and services when fully organised and completely reliable information is not available. This collection of articles, which are broadly in the emerging area of Autonomic Networks, address some issues that are relevant to this broad agenda, going from the ‘upper and user’ end of networks all the way down to the hardware of routers. The paper by G. Rubino's group at IRISA Rennes addresses how Quality of Service (QoS) can be measured in a manner which is compatible with a human user's perception. Clearly one cannot have human users acting as ‘quality control monitors’ across the Internet. On the other hand, if we can dispose of an algorithm that inputs objectively measurable QoS (such as packet loss) indicators and provide a resulting estimate about how a human user might react to the resulting, for instance, video sequence that is received from the network, one could then evaluate whether the network is doing its job properly. The IRISA group describes an approach that uses the Random Neural Network trained to provide user-oriented quality predictions from measured QoS metrics. The results are tested in a variety of practical settings. The work by J. Pitt and his colleagues considers voting based protocols in ad hoc networks. Voting can be used in many contexts where some form of distributed decision making is needed, both for the management of resources, or for allocating degrees of trustworthiness to certain agents, or to take a decision in a distributed electronic environment. The paper by R. Lent considers how highly distributed ad hoc networks may be evaluated during the design and development phase; he suggests techniques that combine the physical characteristics of the network, as well as its software and the users' characteristics within one unified and flexible simulation environment. The paper written by A. Gyorgi and G. Ottusak considers the question of routing, when pre-existing information does not exist. This theoretical paper considers methods, similar to the ones that have been implemented in Imperial College's Cognitive Packet Network test-bed [1,2,3], to make routing decisions based on the success or failure of past routing decisions. They show that such decisions, if properly defined, can converge to the optimal decisions in some precise mathematical manner. Finally, the work by T. Koçak considers how the Random Neural Network can be imbedded in a hardware router to obtain fast and smart routing decisions. This paper discusses routing hardware which is being developed for the Cognitive Packet Network.

Autonomic communications seek to improve the ability of network and services to cope with unpredicted change, including changes in topology, load, task, the physical and logical characteristics of the networks that can be accessed, and so forth. Broad-ranging autonomic solutions require designers to account for a range of end-to-end issues affecting programming models, network and contextual modeling and reasoning, decentralised algorithms, trust acquisition and maintenance---issues whose solutions may draw on approaches and results from a surprisingly broad range of disciplines. We survey the current state of autonomic communications research and identify significant emerging trends and techniques.

Packet routing in networks requires knowledge about available paths, which can be either acquired dynamically while the traffic is being forwarded, or statically (in advance) based on prior information of a network's topology. This paper describes an experimental investigation of path discovery using genetic algorithms (GAs). We start with the quality-of-service (QoS)-driven routing protocol called "cognitive packet network" (CPN), which uses smart packets (SPs) to dynamically select routes in a distributed autonomic manner based on a user's QoS requirements. We extend it by introducing a GA at the source routers, which modifies and filters the paths discovered by the CPN. The GA can combine the paths that were previously discovered to create new untested but valid source-to-destination paths, which are then selected on the basis of their "fitness." We present an implementation of this approach, where the GA runs in background mode so as not to overload the ingress routers. Measurements conducted on a network test bed indicate that when the background-traffic load of the network is light to medium, the GA can result in improved QoS. When the background-traffic load is high, it appears that the use of the GA may be detrimental to the QoS experienced by users as compared to CPN routing because the GA uses less timely state information in its decision making.

Denial of service attacks, viruses and worms are common tools for malicious adversarial behaviour in networks. We propose the use of our autonomic routing protocol, the cognitive packet network (CPN), as a means to defend nodes from distributed denial of service (DDoS) attacks, where one or more attackers generate flooding traffic from multiple sources towards selected nodes or IP addresses. We use both analytical and simulation modelling, and experiments on our CPN testbed, to evaluate the advantages and disadvantages of our approach in the presence of imperfect detection of DDoS attacks, and of false alarms.

We present experimental results on an autonomic network test-bed, the cognitive packet network (CPN), designed for research in adaptive quality-of-service (QoS) management. CPN is fully compatible at its edges with the Internet Protocol, while internally it offers dynamic routing based on on-line sensing and monitoring. CPN can implement distributed adaptive shortest-path routing, and we compare it with minimum-delay based routing and a composite approach.

Novel user-oriented networked systems will simultaneously exploit a variety of wired and wireless communication modalities to offer different levels of quality of service (QoS), including reliability and security to users, low economic cost, and performance. Within a single such user-oriented network, different connections themselves may differ from each other with respect to QoS needs. Similarly, the communication infrastructure used by such a network will, in general, be shared among many different networks and users so that the resources available will fluctuate over time, both on the long and short term. Such a user-oriented network will not usually have precise information about the infrastructure it is using at any given instant of time, so that its knowledge should be acquired from online observations. Thus, we suggest that user-oriented networks should exploit self-adaptiveness to try to obtain the best possible QoS for all their connections. In this paper we review experiments which illustrate how "self-awareness," through online self-monitoring and measurement, coupled with intelligent adaptive behavior in response to observations, can be used to offer user-oriented QoS. Our presentation is based on ongoing experimental work with several "cognitive packet network" testbeds that we have developed.

Network self-awareness is the ability of a network to observe its own behavior using internal probing and measurement mechanisms, and to make effective autonomous use of these observations for self-management. Experiments are conducted to evaluate the goal's impact on observed QoS for the user's payload. In addition to packet loss due to congestion, we also introduce an artificial packet loss at certain nodes to represent failures or other undesirable events. We see that just using delay in the QoS goal is a good way to reduce delay and loss if losses are only the result of congestion. However, as one would expect, using loss in the user's QoS goal is seen to be useful if the paths, which are selected by SPs, are to avoid nodes where packet losses are occurring for reasons other than congestion. In general we see good correlation between the QoS goal that the SPs use to find paths, and the resulting QoS observed by DPs.

Spiked recurrent neural networks with "multiple classes" of signals have been recently introduced by Gelenbe and Fourneau (1999), as an extension of the recurrent spiked random neural network introduced by Gelenbe (1989). These new networks can represent interconnected neurons, which simultaneously process multiple streams of data such as the color information of images, or networks which simultaneously process streams of data from multiple sensors. This paper introduces a learning algorithm which applies both to recurrent and feedforward multiple signal class random neural networks (MCRNNs). It is based on gradient descent optimization of a cost function. The algorithm exploits the analytical properties of the MCRNN and requires the solution of a system of nC linear and nC nonlinear equations (where C is the number of signal classes and n is the number of neurons) each time the network learns a new input-output pair. Thus, the algorithm is of O([nC]/sup 3/) complexity for the recurrent case, and O([nC]/sup 2/) for a feedforward MCRNN. Finally, we apply this learning algorithm to color texture modeling (learning), based on learning the weights of a recurrent network directly from the color texture image. The same trained recurrent network is then used to generate a synthetic texture that imitates the original. This approach is illustrated with various synthetic and natural textures.

We propose that learning agents (LAs) be incorporated into simulation environments in order to model the adaptive behavior of humans. These LAs adapt to specific circumstances and events during the simulation run. They would select tasks to be accomplished among a given set of tasks as the simulation progresses, or synthesize tasks for themselves based on their observations of the environment and on information they may receive from other agents. We investigate an approach in which agents are assigned goals when the simulation starts and then pursue these goals autonomously and adaptively. During the simulation, agents progressively improve their ability to accomplish their goals effectively and safely. Agents learn from their own observations and from the experience of other agents with whom they exchange information. Each LA starts with a given representation of the simulation environment from which it progressively constructs its own internal representation and uses it to make decisions. The paper describes how learning neural networks can support this approach and shows that goal based learning may be used effectively used in this context. An example simulation is presented in which agents represent manned vehicles; they are assigned the goal of traversing a dangerous metropolitan grid safely and rapidly using goal based reinforcement learning with neural networks and compared to three other algorithms.

Presents a simple continuous analog hardware realization of the random neural network (RNN) model. The proposed circuit uses the general principles resulting from the understanding of the basic properties of the firing neuron. The circuit for the neuron model consists only of operational amplifiers, transistors, and resistors, which makes it candidate for VLSI implementation of random neural networks with feedforward or recurrent structures. Although the literature is rich with various methods for implementing the different neural networks structures, the proposed implementation is very simple and can be built using discrete integrated circuits for problems that need a small number of neurons. A software package, RNNSIM, has been developed to train the RNN model and supply the network parameters which can be mapped to the hardware structure. As an assessment on the proposed circuit, a simple neural network mapping function has been designed and simulated using PSpice.

We introduce several novel random neural network based techniques to address a difficult "inverse problem" in semiconductor fabrication metrology. The problem is that of deducing a chip's vertical cross-section from two-dimensional top-down scanning electron microscope images of the chip surface. Our results are illustrated with a variety of real data sets. In semiconductor chip fabrication, photo resistive material is used as an overlay which will protect substrate areas (typically metal) which must remain on the chip after other unprotected substrate areas are etched off. The shape and size of the photo-resist material, at the submicron level, is therefore largely responsible for the shape and quality of the protected substrate. Critical dimension scanning electron microscopy (SEM) is used to determine this shape, and the research addressed in the paper proposes methods using learning neural networks, combined with physical modelling, to accurately obtain surface shape information from SEM imaging.

Based on our earlier work ("Towards networks with intelligent packets", Proc. 14th Int. Symp. on Computer and Information Sciences, p. 1-11, Oct. 1999), we discuss packet networks in which intelligent capabilities for routing and flow control are concentrated in the packets, rather than in the nodes and protocols. This paper describes a possible testbed to test and evaluate their capabilities, and presents an analytical model for the worst and best case performance of such systems.

Sources of real-time traffic are generally highly unpredictable with respect to the instantaneous and average load which they create. Yet such sources will provide a significant portion of traffic in future networks, and will significantly affect the overall performance of and quality of service. Clearly high levels of compression are desirable as long as video quality remains satisfactory, and our research addresses this key issue with a novel learning-based approach. We propose the use of neural networks (NNs) as post-processors for any existing video compression scheme. The approach is to interpolate video sequences and compensate for frames which may have been lost or deliberately dropped. We show that deliberately dropping frames will significantly reduce the amount of offered traffic in the network, and hence the cell loss probability and network congestion, while the NN post-processor will preserve most of the desired video quality. Dropping frames at the sender or in the network is also a fast way to react to network overload and reduce congestion. Our interpolation techniques at the receiver, including neural network-based algorithms, provide output frame rates which are identical to (or possibly higher than) the original video sequence's frame rate. The resulting video quality is essentially equivalent to the sequence without frame drops, despite the loss of a significant fraction of the frames. Experimental evaluation using real video sequences is provided or interpolation with a connectionist NN using the backpropagation learning algorithm, the random NN (RNN) in a feed-forward configuration with its associated learning algorithm, and cubic spline interpolation. The experiments show that when more frames are being dropped or lost, the RNN performs generally better than the other techniques in terms of resulting video quality and overall performance. When the fraction of dropped frames is small, cubic splines offer better performance. Experimental data shows that this receiver-reconstructed subsampling technique significantly reduces the cell loss rates in an asynchronous transfer mode switch for different buffer sizes and service rates.

The cost and the closely related length of time spent in searching for mines or unexploded ordnance (UXO) may well be largely determined by the number of false alarms. False alarms can result in time consuming digging of soil or in additional multisensory tests in the minefield. The authors consider two area-based methods for reducing false alarms. These are: (a) the previously known "declaration" technique and (b) the new /spl delta/ technique, which they introduce. They first derive expressions and lower bounds for false-alarm probabilities as a function of declaration area and discuss their impact on receiver operation characteristic (ROC) curves. Second, they exploit characteristics of the statistical distribution of sensory energy in the immediate neighborhood of targets and of false alarms from available calibrated data, to propose the /spl delta/ technique, which significantly improves discrimination between targets and false alarms. The results are abundantly illustrated with statistical data and ROC curves using electromagnetic-induction sensor data made available through DARPA from measurements at various calibrated sites.

We propose cognitive packet networks (CPN) in which intelligent capabilities for routing and flow control are concentrated in the packets, rather than in the nodes and protocols. Cognitive packets within a CPN route themselves. They are assigned goals before entering the network and pursue these goals adaptively. Cognitive packets learn from their own observations about the network and from the experience of other packets with whom they exchange information via mailboxes. Cognitive packets rely minimally on routers. This paper describes CPN and shows how learning can support intelligent behavior of cognitive packets.

This paper presents a novel random neural network (RNN) based soft decision decoder for block codes. One advantage of the proposed decoder over conventional serial algebraic decoders is that noisy codewords arriving in non-binary form can be corrected without first rounding them to binary form. Another advantage is that the RNN, after being trained, has a simple hardware realization that is ideal for implementation as a VLSI chip. The proposed decoder is tested on Hamming linear codes and the results are compared with that of the optimum soft decision decoder and the conventional hard decision decoder. Extensive simulations show that the RNN based decoder reduces the error probability to zero in the range of the error correcting capacity of the used code. On the other hand, it is much better than the hard decision decoder for codewords corrupted with more errors.

Remedial mine detection and the detection of unexploded ordnance (UXO) have become very important for humanitarian reasons. This paper addresses mine detection using commonly used electromagnetic induction sensors. We propose and evaluate two neural network approaches to mine detection which provide a robust nonparametric technique, based on training the networks using data from a previously calibrated portion of the minefield, or from a similar minefield. In the first approach, we combine a novel statistic, the S-statistic (which is a real valued variable related to the relative energy difference measured around a point in the minefield) with the /spl delta/-technique in a random neural network (RNN) design. In the second approach, a RNN is trained using a 3/spl times/3 block measurement window, and then applied as a postprocessor for the /spl delta/-technique. This RNN has an unconventional feedforward structure which realizes a matched filter to discriminate between nonmine patterns and mines. Experimental results for both approaches show that the RNN reduces false alarms substantially over the /spl delta/-technique and the energy detector.

By extending the pulsed recurrent random neural network (RNN) discussed in Gelenbe (1989, 1990, 1991), we propose a recurrent random neural network model in which each neuron processes several distinctly characterized streams of "signals" or data. The idea that neurons may be able to distinguish between the pulses they receive and use them in a distinct manner is biologically plausible. In engineering applications, the need to process different streams of information simultaneously is commonplace (e.g., in image processing, sensor fusion, or parallel processing systems). In the model we propose, each distinct stream is a class of signals in the form of spikes. Signals may arrive to a neuron from either the outside world (exogenous signals) or other neurons (endogenous signals). As a function of the signals it has received, a neuron can fire and then send signals of some class to another neuron or to the outside world. We show that the multiple signal class random model with exponential interfiring times, Poisson external signal arrivals, and Markovian signal movements between neurons has product form; this implies that the distribution of its state (i.e., the probability that each neuron of the network is excited) can be computed simply from the solution of a system of 2Cn simultaneous nonlinear equations where C is the number of signal classes and n is the number of neurons. Here we derive the stationary solution for the multiple class model and establish necessary and sufficient conditions for the existence of the stationary solution. The recurrent random neural network model with multiple classes has already been successfully applied to image texture generation (Atalay & Gelenbe, 1992), where multiple signal classes are used to model different colors in the image.

This paper examines the function approximation properties of the "random neural-network model" or GNN. The output of the GNN can be computed from the firing probabilities of selected neurons. We consider a feedforward Bipolar GNN (BGNN) model which has both "positive and negative neurons" in the output layer, and prove that the BGNN is a universal function approximator. Specifically, for any f is an element of C([0, 1]s) and any epsilon>0, we show that there exists a feedforward BGNN which approximates f uniformly with error less than epsilon. We also show that after some appropriate clamping operation on its output, the feedforward GNN is also a universal function approximator.

Future networks must be adequately equipped to handle multipoint communication in a fast and economical manner. Services requiring such support include desktop video conferencing, tele-classrooms, distributed database applications, etc. In networks employing the asynchronous transfer mode (ATM) technology, routing a multicast is achieved by constructing a minimum cost tree that spans the source and all the destinations. When the network is modeled as a weighted, undirected graph, the problem is that of finding a minimal Steiner tree for the graph, given a set of destinations. The problem is known to be NP-complete. Consequently, several heuristics exist which provide approximate solutions to the Steiner problem in networks, We show how the random neural network (RNN) can be used to significantly improve the quality of the Steiner trees delivered by the best available heuristics which are the minimum spanning tree heuristic and the average distance heuristic. We provide an empirical comparison and find that the heuristics which are modified using the neural network yield significantly improved trees.

In this paper we describe a novel neural network technique for video compression, using a "point-process" type neural network model we have developed which is closer to biophysical reality and is mathematically much more tractable than standard models. Our algorithm uses an adaptive approach based upon the users' desired video quality Q, and achieves compression ratios of up to 500:1 for moving gray-scale images, based on a combination of motion detection, compression, and temporal subsampling of frames. This leads to a compression ratio of over 1000:1 for full-color video sequences with the addition of the standard 4:1:1 spatial subsampling ratios in the chrominance images. The signal-to-noise ratio ranges from 29 dB to over 34 dB. Compression is performed using a combination of motion detection, neural networks, and temporal subsampling of frames. A set of neural networks is used to adaptively select the desired compression of each picture block as a function of the reconstruction quality. The motion detection process separates out regions of the frame which need to be retransmitted. Temporal subsampling of frames, along with reconstruction techniques, lead to the high compression ratios.

Brain magnetic resonance (MR) images contain massive information requiring lengthy and complex interpretation (as in the identification of significant portions of the image), quantitative evaluation (as in the determination of the size of certain significant regions), and sophisticated interpretation (as in determining any image portions which indicate signs of lesions or of disease). In this paper we first survey the clinical and research needs for brain imaging. We present the state-of-the-art in relevant image analysis techniques. We then discuss our recent work on the use of novel artificial neural networks which have a recurrent structure to extract precise morphometric information from MRI scans of the human brain. Finally, experimental data using our novel approach is presented and suggestions are made for future research.

Traditional approaches to guaranteeing quality-of-service (QoS) in ATM networks have focused on performance metrics such as cell loss probability, end-to-end cell delay and delay jitter. However, the block loss rate is a more meaningful metric for applications such as medical imaging and real-time video, and for high-level protocols such as IP which will use ATM as the transport mechanism. A block is defined as a group of consecutive ATM cells. A block loss occurs when a single cell from a block is lost. We propose and evaluate, via extensive simulation, a technique for reducing block loss in ATM networks. Our method combines two well-studied approaches for minimizing the impact of information loss during transport over ATM networks which have been considered independently of each other in the past: priority-based cell discarding and forward error correction (FEC). We also study an enhanced buffer management algorithm which we call adaptive pushout (ADP) that accounts for correlations between cell losses. The ADP algorithm reduces block loss rates to near-optimal levels. We consider further performance enhancements to the ADP policy that deliver excellent performance even under very heavy load.

This Hot-Topics Session on G-Networks aims at bringing these relatively new models which we introduced for the first time in 1989 and 1990, to the attention of the performance evaluation and modeling community. The session includes presentations by Peter Harrison, Onno Boxma, Jean-Michel Fourneau and myself. We will cover the basic concepts, some examples of potential applications, as well as recent research efforts in this area.

We report on the design and simulation of novel algorithms which will ensure that application software runs correctly on a MIMD system in which processing units (PU) can fail. The effect of these algorithms is evaluated for random task graphs using simulation as failure rates increase. An example of a specific application is also examined (the Fast Fourier Transform) for which we construct the task graph and then simulate its execution under various values of the failure rates of processors.

The capacity to learn from examples is one of the most desirable features of neural network models. We present a learning algorithm for the recurrent random network model (Gelenbe 1989, 1990) using gradient descent of a quadratic error function. The analytical properties of the model lead to a "backpropagation" type algorithm that requires the solution of a system of n linear and n nonlinear equations each time the n-neuron network "learns" a new input-output pair.

For a transaction oriented system, the authors propose that in addition to the conventional recovery techniques, such as dumps and roll-back recovery, system availability be enhanced by the introduction of 'failure tests'. They present a model to analyze the effect of the failure rate, the failure tests, and the periodic dumps, on global system availability. They then compute the optimum value of the interval between dumps, and also the best time interval between failure tests for this system. Numerical examples are presented for various failure models.>

We propose a parallel algorithm for the generation of colour textures based upon the non-linear equations of the "multiple class random neural network model". A neuron is used to obtain the texture value of each pixel in the bit-map plane. Each neuron interacts with its immediate planar neighbours in order to obtain the texture for the whole plane. A model which uses at most 4(C2 + C) parameters for the whole network, where C is the number of colours, is proposed. Numerical iterations of the non-linear field equations of the neural network model, starting with a randomly generated image, are shown to produce textures having different desirable features such as granularity, inclination and randomness. The experimental evaluation shows that the random network provides good results, at a computational cost which is considerably less than that of other approaches such as Markov random fields.

The generation of artifical textures is a useful function in image synthesis systems. The purpose of this paper is to describe the use of the random neural network (RN) model developed by Gelenbe to generate various textures having different characteristics. An eight parameter model, based on a choice of the local interaction parameters between neighbouring neurons in the plane, is proposed. Numerical iterations of the field equations of the neural network model, starting with a randomly generated gray-level image, are shown to produce textures having different desirable features such as granularity, inclination, and randomness. The experimental evaluation shows that the random network provides good results, at a computational cost less than that of other approaches such as Markov random fields. Various examples of textures generated by our method are presented.

This paper presents an analysis of the performance of the Connection Machine, with special emphasis on estimating the effect of its interprocessor communication architecture. A queueing model of the network architecture, including the NEWS and ROUTER networks, is used to compute the slow-down induced by message exchange between processors. Locality of the message exchanges is modelled by message sending probabilities which depend on whether a message is sent by a processor to another processor placed on the same NEWS network, or on the same ROUTER, or at a “remote” location which is only accessible via the ROUTER network. The specific slotted TDMA structure of the ROUTER Network communications is taken into account. The performance degradation of the Connection Machine as a function of the communication and architecture parameters is derived.

In a recent paper (Gelenbe 1989) we introduced a new neural network model, called the Random Network, in which “negative” or “positive” signals circulate, modeling inhibitory and excitatory signals. These signals can arrive either from other neurons or from the outside world: they are summed at the input of each neuron and constitute its signal potential. The state of each neuron in this model is its signal potential, while the network state is the vector of signal potentials at each neuron. If its potential is positive, a neuron fires, and sends out signals to the other neurons of the network or to the outside world. As it does so its signal potential is depleted. We have shown (Gelenbe 1989) that in the Markovian case, this model has product form, that is, the steady-state probability distribution of its potential vector is the product of the marginal probabilities of the potential at each neuron. The signal flow equations of the network, which describe the rate at which positive or negative signals arrive at each neuron, are nonlinear, so that their existence and uniqueness are not easily established except for the case of feedforward (or backpropagation) networks (Gelenbe 1989). In this paper we show that whenever the solution to these signal flow equations exists, it is unique. We then examine two subclasses of networks — balanced and damped networks — and obtain stability conditions in each case. In practical terms, these stability conditions guarantee that the unique solution can be found to the signal flow equations and therefore that the network has a well-defined steady-state behavior.

We introduce a new class of random “neural” networks in which signals are either negative or positive. A positive signal arriving at a neuron increases its total signal count or potential by one; a negative signal reduces it by one if the potential is positive, and has no effect if it is zero. When its potential is positive, a neuron “fires,” sending positive or negative signals at random intervals to neurons or to the outside. Positive signals represent excitatory signals and negative signals represent inhibition. We show that this model, with exponential signal emission intervals, Poisson external signal arrivals, and Markovian signal movements between neurons, has a product form leading to simple analytical expressions for the system state.

Uncertainty in the contents of a data base can be due to several reasons: errors in the data which is entered, changes in the real data which have not been introduced into the data base in the form of updates, errors in the data collection process, unreliable operation of the computer system, "don't care" conditions which are purposely left open by the data base designer, etc. The purpose of this paper is to present a formal model of uncertainty in terms of a probabilistic representation of the data base, and to evaluate the effect of this uncertainty on query processing and on the aggregate or summary information which may suffice in many applications. Our model leads to precise quantifiable engineering estimates and to theorems on the robustness of answers to queries as a function of the uncertainty in the data.

In a flexible manufacturing system stations are arranged along a common conveyor that brings items for processing to the stations and also carries away the processed items. At each station specialized robots automatically load and unload items on and off the conveyor. We examine here a single station in such a system. A new kind of queueing problem arises, with input-output dependencies that result because the same conveyor transports items both to and from the station. The paper analyzes two models of a station. Model 1 has one robot that cannot return a processed item to the conveyor while unloading a new item for processing. Model 2 has two robots to allow simultaneous loading and unloading of the conveyor. A principal goal of the analysis is the proper choice of the distance separating the two points at which items leave and rejoin the conveyor.

We present a model for distributed systems with failing components. Each node may fail and during its recovery the load is distributed to other nodes that are operational. The model assumes periodic checkpointing for error recovery and testing of the status of other nodes for the distribution of load.

The resequencing or serialization problem is of basic interest in distributed systems and computer communication systems.This is because a flow of packets, messages, or updates entering a communication system in chronological order from the same port or from different ports may be disordered.The receiving port must then ensure that these objects are resequenced in the appropriate order before they are fed to the output of the system.In this paper we analyze the end-to-end delay recurred by objects traversing such a system, including the d~sordering delay, the delay introduced by the resequencing algorithm, and the delay due to the output server at the receiving port.The analysis is carded out via factorization methods.

It is of interest to determine whether loosely coupled multiprocessors can be profitably used for the solution of larger numerical problems. We present here a performance evaluation of the gain obtained by solving partial differential equation systems on such an architecture. The experimental setting is an LSI 11 based multimicroprocessor system using a fiber optics local area network designed and implemented at Laboratoire de Recherche en Informatique, Université Paris-Sud. The paper includes a discussion of the numerical methods and of their implementation, a performance model of the parallel processing system, and measurements taken on the experimental system. The experimentally validated theoretical results confirm the interest of our approach based on performance models.

We consider the effect on system performance of the distribution of a data base in the form of multiple copies at distinct sites. The purpose of our analysis is to determine the gain in READ throughput that can be obtained in the presence of consistency preserving algorithms that have to be implemented when UPDATE operations are carried out on each copy. We show that READ throughput diminishes if the number of copies exceeds an optimal value. The theoretical model we develop is applied to a system in which consistency is preserved through the use of Ellis' ring algorithm.

One of the basic problems related to the efficient and secure operation of a transaction oriented file or database system is the choice of the checkpoint interval In this paper we show that the optimum checkpoint interval (i e the time interval between successive checkpoints which maximizes system avadabihty) is a function of the load of the system We also prove that the total operating time of the system (and not the total real time) between successive checkpoints should be a deterministic quantity in order to maximize the availability An explicit expression for this time interval Is obtained These results are a significant departure from previous work where load independent results have been obtained We also present a rigorous analysis of the queuelng process related to the requests for transaction processing arriving at the system, and prove the ergodiclty conditions for the system

A formal model allowing a precise definition of coherence and promptness in a multiple copy information system is presented and used to analyze a class of update synchronization techniques.

The subject of this paper is an adaptive control procedure for stabilizing and optimizing the performance of ALOHA-like slotted packet switching channels. We prove sufficient conditions, via a mathematical model of the adaptive procedure,under which the procedure stabilizes the channel. This procedure is based on "listening" to the silent periods of the channel and modifying the retransmission probabilities of the blocked terminals as a function of this measurement. The optimal choice of the parameters of the adaptive procedure is analysed and simulation results are given in order to illustrate its practical effect.

In this study, we are concerned with a simple error control protocol, the "send and wait" protocol, which uses the classical technique of positive acknowledgment and time-out periods. We first analyze the influence of the time-out on the packet transmission rate. Then we use a queuing analysis to obtain the ergodicity condition and to compute the buffer queue length probability distribution. Finally we compute buffer overflow when a finite number of packets are allowed to enter the node. This analysis allows us to obtain optimum values of the time-out in order to maximize throughput, or to minimize average transit delay through the node or buffer overflow probabilities.

We propose a new method for the control of a multiprogrammed virtual memory computer system. A mathematical model solved by decomposition permits us to justify that the method avoids thrashing. Simulation experiments are used to test the robustness of the predictions of the mathematical model when certain simplifying assumptions are relaxed and when a slightly simpler control technique based on the same principle is used. Comparisons are given with the case where an "optimal" control is used and with that with no control. We also provide a simulation evaluating the estimators used in an implementation of the control, as well as the responsiveness of the controlled system to transients in the workload.

The purpose of this paper is to analyze and optimize the behavior of the broadcast channel for a packet transmission operating in the slotted mode Mathematical methods of Markov chain theory are used to prove the inherent lnstablhty of the system If no control is apphed, the effective throughput of the system will tend to zero tf the population of user terminals ~s sufficiently large Two classes of control pohcles are examined, the first acts on admissions to the channel from active terminals, and the second modifies the retransmlss~on rate of packets In each case sufflc~ent conditions for channel stability are given.In the case of retransm~sslon controls it is shown that only pohcles which assure a rate of retransmlsslon from each blocked terminal of the form off = 1/n, where n is the total number of blocked terminals, will yield a stable channel It ts also proved that the optimal pohcy which maximizes the maximum achievable throughput wtth a stable channel IS of the formf = (1 -k)/n Simulations illustrating channel lnstabdlty and the effect of the opnmal control are prowded KEY WORDS AND PHRASES computer networks, packet swnchlng, broadcast channel, optimal control CR CATEGORIES 4 3, 4 6, 5 5, 6 2, 6 35 ~t has to repeat the transmission untd it succeeds Either a termmal which is not blocked

We develop a method based on diffusion approximations in order to compute, under some general conditions, the queue length distribution for a queue in a network. Applications to computer networks and to time-sharing systems are presented.

The allocation of the central processing unit (CPU) of a computer system in quanta of fixed length in round-robin fashion favors jobs with shorter total CPU processing time by reducing the time they spend waiting in queue below what it would be if all the lobs were served in first-come-first-served order This effect can be accentuated by the use of short quanta. The main disadvantage of this allocation policy is the resulting time the CPU spends in overhead activities when switching from one task to the other, this too will increase with smaller quanta. Thus, it appears useful to consider adaptive CPU allocation policies to reduce the overhead during high traffic conditions when saturation of this resource is more likely while keeping a small quantum during periods of low arrival traffic. In this paper we analyse such a policy, it is assumed that each time at least r (a threshold) arrivals occur during a quantum, the job currently using the CPU is allocated an additional quantum (if It is needed). Thus, the number of job arrivals during a quantum is used as a sensor of the intensity of arrival traffic. This policy, which can be easily implemented in hardware, is analysed using a mathematical model yielding the average response time for jobs as a function of mean total CPU time, the quantum size, r, and a fixed overhead for switching tasks, with a Poisson arrival process. Numerical results to illustrate the effect of this policy are presented.

A new treatment of the boundary conditions of diffusion approximations for interconnected queueing systems is presented. The results have applications to the study of the performance of multiple-resource computer systems. In this approximation method, additional equations to represent the behavior of the queues when they are empty are introduced. This reduces the dependence of the model on heavy traffic assumptions and yields certain results which would be expected from queueing or renewal theory. The accuracy of the approach is evaluated by comparison with certain known exact or numerical results.

Due to the practical complexity of addressing variable length records placed in arbitrary locations of a fixed-head disk (or drum), and because of difficulty of managing secondary memory space in such cases, variable length records are often stored with their first address at a fixed location of the magnetic support. We present a queuing model of such a scheme, assuming a Poisson arrival stream and arbitrary distributed record lengths. The stationary probability distribution of the number of transfer requests in queue and the expected response time are obtained. Numerical examples illustrating the results are presented.

Queueing network models are well suited for analyzing certain resource allocation problems associated with operating system design. An example of such a problem is the selection of the level of multiprogramming in virtual memory systems. If the number of programs actively competing for main memory is allowed to reach too high a value, trashing will occur and performance will be seriously degraded. On the other hand, performance may also suffer if the level of multiprogramming drops too low since system resources can become seriously under utilized in this case. Thus it is important for virtual memory systems to maintain optimal or near optimal levels of multiprogramming at all times.

The book is a general and fairly brief introduction to the structure of time-sharing computer systems and to the facilities that they provide. The authors indicate that their book is supposed to serve two purposes. It is supposed to assist in making comparisons between different systems in order to make a selection evaluation, and it should also serve as a textbook for an undergraduate course on time-sharing systems.

The replacement problem arises in computer system management whenever the executable memory space available is insufficient to contain all data and code that may be accessed during the execution of an ensemble of programs. An example of this is the page replacement problem in virtual memory computers. The problem is solved by using a replacement algorithm that selects code or data items that are to be removed from executable memory whenever new items must be brought in and no more free storage space remains. An automaton theoretic model of replacement algorithms is introduced for the class of ``random partially preloaded'' replacement algorithms, which contain certain algorithms of practical and theoretical interest. An analysis of this class is provided in order to evaluate their performance, using the assumption that the references to the items to be stored are identically distributed independent random variables. With this model, it is shown that the well-known page replacement algorithms FIFO and RAND yield the same long-run page-fault rates.

A new model for stochastic sequential machines is introduced. This model consists of a deterministic Mealy-type synchronous sequential machine some of whose inputs are random number generators while the outputs of another set of random number generators are used to perturb the output function of the deterministic Mealy machine. Thus this model is physically realizable in terms of random number generators, logic and memory elements. It is shown that this model and the Shannon model of a stochastic sequential machine are coextensive and a procedure is given, through a proof of this result, for obtaining one from the other. The model given here is then compared with the realizable model introduced by Nieh and Carlyle [1] and it is shown that their model and ours may be realized with identical random number generators for the case of input-state calculable stochastic sequential machines.

It is shown that a single-output Moore-type n-state linear machine may be realized with no more than 2n copies of the AND–OR-delay (AOD) module of Newborn, Weiner, and Hopcroft. This bound is significantly lower than that for arbitrary single-output Moore-type machines, which is 2n.