The classification of IoT traffic is important to improve efficiency and security of IoT-based networks, and state-of-the-art classification methods are based on Deep Learning. However, most of the current results require a big amount of data to be trained. This way, in real life situations, where there is a scarce amount of IoT traffic data, the models would not perform so well. Consequently, these models underperform outside their initial training conditions and fail to capture the complex characteristics of network traffic, rendering them inefficient and unreliable in real-world applications. In this paper, we propose a novel IoT Traffic Classification Transformer (ITCT) approach, utilizing the state-of-the-art transformer-based model named TabTransformer. The model, which is pre-trained on a large labeled MQTT-based IoT traffic dataset and may be fine-tuned with a small set of labeled data, showed promising results in various traffic classification tasks. Our experiments demonstrated that the ITCT model significantly outperforms existing models, achieving an overall accuracy of 82%. To support reproducibility and collaborative development, all associated code is made publicly available.

Emergency services response time is a critical factor for road traffic injury care, and the current scenario is that most accidents must be reported by a bystander or by those involved, which usually takes several minutes. This work proposes a crash response system using LoRa-based vehicular communications, aiming to reduce response times for road accidents, save lives, and take advantage of the license-free radio frequencies used by LoRa. The system uses broadcast beacons to build multi-hop routes from vehicle nodes to emergency services. The beacon message contains important information, such as the type of accident, latitude, longitude, the estimated number of people, and the last known speed. We have evaluated our proposal in the ns3 network simulator by simulating an urban environment in a 10km wide area and varying the node density. Results have shown that the network can transmit a beacon across several kilometers with minimal latency. The amount of nodes in the network is a critical factor, with greater node density network instances displaying a larger Packet Delivery Ratio (PDR) and Network Coverage but also a slightly longer latency. As a complementary contribution, all the code produced in this research is publicly available as Open Source Software.

The Internet of Things (IoT) has undergone rapid popularization, reaching a wide range of application domains, such as manufactures. Hence, more and more heterogeneous IoT devices have been deployed in a variety of industrial environments, progressively becoming common objects to the supply chain. The physical infrastructure of manufacturing systems has become complex and requires efficient and dynamic solutions for managing network performance and security. Network Function Virtualization (NFV) has attracted attention when the intention is to respond to security threats on Industrial IoT (IIoT). Few works use NFV to detect and mitigate security threats on IIoT networks, but even less consider performance indicators of the network context when placing the Virtual Network Functions (VNFs). Thus, this work introduces a Machine Learning (ML) approach to place security VNFs based on NFV performance to mitigate Distributed Denial of Service (DDoS) attacks on IIoT. Experiments considering a new composed data set and diverse ML techniques show ML classification as an alternative for IIoT scenarios, achieving, according to the best-performing technique, 99.40% of accuracy in relation to the ideal placement. To facilitate the reproduction of the work, all the code and data produced are publicly available.

MQTT's security has been a major concern because of its weak protocol implementations. Over the last few years, several fuzzing frameworks have been proposed to mitigate this issue. However, these frameworks lack sufficient knowledge of MQTT's specifications, requiring a considerable amount of network packets to cover all of its features and functionality. In this paper, we explain how to improve the efficiency of fuzzing frameworks for MQTT by using a grammar based on its specifications. Although defining a grammar is time-consuming and complex, these drawbacks are overshadowed by its benefits, such as deep state exploration and efficiency. Our improvements are implemented in MQTTGRAM, a new grammar-based fuzzer for MQTT. Due to these improvements, MQTTGRAM offers higher code coverage with significantly fewer packets than existing MQTT fuzzers. For instance, MQTTGRAM exchanges up to 9x fewer packets than its counterparts without reducing the line coverage.

Over the last few years, MQTT applications have been widely exposed to vulnerabilities because of their weak protocol implementations. For our preliminary research, we conducted background studies to: (1) determine the main cause of vulnerabilities in MQTT applications; and (2) analyze existing MQTT-based testing frameworks. Our preliminary results confirm that MQTT is most susceptible to malformed packets, and its existing testing frameworks are based on blackbox fuzzing, meaning vulnerabilities are difficult and time-consuming to find. Thus, the aim of my research is to study and develop effective fuzzing strategies for the MQTT protocol, thereby contributing to the development of more robust MQTT applications in IoT and Smart Cities.

Cloud-Fog Radio Access Networks (CF-RAN) were proposed as an alternative network architecture to alleviate the high fronthaul capacity requested in traditional Cloud RAN (CRAN) by moving some BaseBand Units (BBUs) from the cloud nodes to fog nodes closer to users. However, when BBU processing is moved into fog nodes, OPEX and CAPEX will increase, and the cost and energy savings introduced by CRAN will also reduce. Moreover, mobile traffic fluctuations may lead to an unbalanced resource utilization and energy- inefficient operation in fog nodes. To address this problem, processing functions in fog nodes could be activated and deactivated in function of network traffic and BBUs placed on fog nodes could be migrated to cloud nodes when network traffic is low. In this paper, we propose an Integer Linear Programming (ILP) formulation to address this dynamic resource allocation problem. By means of Network Functions Virtualization (NFV), virtualized BBUs (vBBUs) can be dynamically allocated and deallocated in fog nodes. Furthermore, considering the availability of cloud nodes and the optical fronthaul, vBBUs can be migrated from fog nodes to cloud nodes in order to balance processing loads and save energy. Compared to a baseline incremental algorithm without vBBU migration, our proposal reduces blocking probability in 89% and achieves power savings of 38%, while providing a very small rate of service interruption due to vBBUs migration.

Hybrid Cloud-Fog Radio Access Network (CF-RAN) is a recent network architecture proposed to increase network coverage from CRAN while leveraging power consumption in future 5G networks. In CF-RAN, the processing of baseband signals from Remote Radio-Heads (RRHs) can be performed in virtualized BaseBand Units (vBBUs) located in the cloud or in fog nodes that are instantiated in function of the network demand. Through a Time-and-Wavelength Division Multiplexing Passive Optical Network (TWDM-PON), virtualized PONs (VPONs) can be dynamically created to support transmissions from RRHs to vBBUs. However, due to traffic fluctuations, the amount of necessary vBBUs and VPONs may change along a day. In this paper, we propose a batch scheduling algorithm based on Integer Linear Programming (ILP) to perform reconfiguration of VPONs and migration of vBBUs among processing nodes in function of fluctuation on traffic demands. Our results show that, in comparison to an incremental algorithm without reconfiguration of VPONs and vBBUs migration capacities, our algorithm reduces power consumption and bandwidth wastage by up to 28% and 57%, respectively, and also eliminates blocking probability.

Optical networks are commonly adopted to implement the fronthaul of Cloud-RAN (CRAN) and Hybrid-CRAN (H-CRAN) architectures to support the 5G traffic. The Cloud-Fog RAN (CF-RAN) architecture promotes cloud and local fog processing of baseband signals with the novelty of activating on demand processing capabilities under the Fog Computing and Network Functions Virtualization (NFV) paradigms. Timeand-Wavelength-Division-Multiplexed Passive Optical Networks (TWDM-PON) are used to implement the fronthaul due to the capacity of Virtual-PON (VPON) formation, which can reduce power consumption by allocating as many Remote Radio Heads (RRH) as possible into a single optical channel. However, the TWDM-PON wavelengths must be assigned to the VPONs in an exclusive manner so different VPONs do not share a common wavelength and thus VPONs do not collide on the optical links. In this paper we propose an Energy-Efficient VPON Formation and Wavelength Dimensioning formulation through Integer Linear Programming (ILP) to efficiently form VPONs and dimension the wavelengths so collisions never occurs. Results from the experiments show that the ILP formulation was able to form VPONs and dimension the wavelengths to the network nodes on the most energy-efficient way, achieving gains in the order of 80.27% in comparison to a random VPON formation approach.

The adoption of microservice-based architecture has become increasingly popular. Microservice containerization is a technique used by developers to facilitate the deployment process of applications based on this architecture. There are several implementation models for microservices. In this paper we study and analyze the performance of these models in terms of the use of network, CPU, memory and disk. Pros and cons related to the development process are also discussed. Among the results obtained with measurements made in a public cloud, the significant reductions in network consumption (up to 99%) are noteworthy when using one container per microservice.

In the context of future Cloud Radio Access Networks (CRAN), optical networks will play an important role to provide the required transport capacity between cell-sites and processing pools, especially for future 5G scenarios. For instance, using CPRI fronthaul technologies a single antenna element can generate data up to 24.3Gbps even with current configurations of radio transmissions, and it is expected to generate up to Tbps with the advance of technology. So, the transport segment of a 5G network needs to be accurately planned to accommodate all the generated traffic. In this work, we propose the use of a Passive Optical Network (PON) jointly with the emergent paradigms of Fog Computing and Network Function Virtualization (NFV) to energy-efficiently support the high traffic transported in emergent mobile networks in an hybrid architecture called Cloud/Fog RAN (CF-RAN) that allows local and remote baseband processing. We introduce an Integer Linear Programming (ILP) model to schedule the processing of CPRI demands among the processing nodes of the network and turn on or off processing functions on demand. Our approach is able to accommodate demands on the nodes of the network in the most energy efficient way. We compare our results with CRAN and distributed architectures (DRAN) and show that an energy efficient planning can achieve considerable gains in power consumption.

Adaptivity is a capability that enables a system to choose amongst various alternatives to satisfy or maintain the satisfaction of certain requirements. The criteria of requirements satisfaction could be pragmatic and context-dependent. Contextual Goal Models (CGM) capture the power of context on banning or allowing certain alternatives to reach requirements (goals) and also deciding the quality of those alternatives with regards to certain quality measures (softgoals). It is used to depict facets of the decision making strategy and rationale of an adaptive system at the preliminary level of requirements. In this paper we argue the case for pragmatic requirements and extend the CGM with additional constructs to capture them and allow their analysis. We also develop an automated analysis which aids the planning and scheduling of tasks execution to meet pragmatic goals. Moreover, we evaluate our modelling and analysis regarding correctness and performance. Such an evaluation showed the applicability of the approach and its usefulness in aiding sensible decisions. It has also shown its capability to do so in a time short enough to suit run-time adaptation decision making.

Data centers power consumption corresponds to near 2% of the total world wide power consumption, with constantly increasing greenhouse effect and CO2footprints. Virtualization techniques improve the efficiency of data centers infrastructure sharing a same physical hardware among several Virtual Machines (VMs). An efficient VM placement can minimize even further the hardware and energy needs. In contrast to existing VM placement algorithms that usually focus on a single resource or assumes that resources demands are deterministic, this paper proposes and compares four energy-aware algorithms that consider multiple stochastic resources, including network bandwidth. We first formulate the problem as a multi objective optimization problem with stochastic resources and we present two algorithms based on this approach. We also formulate the problem as an evolutionary computation problem and we present two algorithms based on this approach. The objective is a joint strategy: minimize the required hardware to maximize the allocated VMs satisfying the resource requirements. Through simulations, we compare our algorithms using real VMs workloads from the PlanetLab project and showed the significant improvements on power consumption and network utilization. In average, the algorithms reduce power consumption by 87.90% and the network utilization by 9.94%.

This paper introduces a task scheduler for clouds which considers that estimations of the available network bandwidth are not precise. The efficacy of the scheduler was evaluated with different type of applications and degree of uncertainty of available bandwidth estimations. Results show that the proposed scheduler is robust to imprecise input information.

Network virtualization is a promising technology for the Internet of the Future. An open issue in virtualization is the management of network resources in a way that energy savings are achieved without compromising the Quality of Service (QoS) requirements of the virtual networks. The dynamic allocation and deallocation of virtual networks can lead the state of the substrate to a less than optimum energy consumption. This paper introduces two algorithms for the migration of virtual routers and/or links which aims to allocate resources so that energy consumption is minimized. The efficacy of the migration of virtual routers and/or links and its impact on energy consumption are analyzed based on results derived via simulations.

Network virtualization is a promising technique for building the Internet of the future since it enables the introduction of new features into network elements at low cost. An open issue in virtualization is how to search for an efficient mapping of virtual network elements onto those of the existing physical network. Mapping is an NP-hard problem and existing solutions take long time to find a solution. This paper presents four new approximated algorithms based on two integer linear programming formulations that runs fast and, also, consider various real network characteristics, which is neglected by other proposals in the literature.

Due to the fact that grid networks are non-proprietary systems, resource allocation mechanisms must cope with fluctuations in resource availability to guarantee the QoS requirements of the applications. Mechanisms are classified either as reactive mechanisms, which react to the fluctuations in run time, or as proactive mechanisms, which try to infer the occurrence of fluctuations in resource availability. This paper presents a study that evaluates the performance of these two different mechanisms in order to compare their advantages. Besides, a new mechanism, integrating the main characteristics of these mechanisms, is proposed. It produces makespan lower than those given by the other two mechanisms.

The Future Internet demands energy efficient communication to cope with the ever increasing power consumption. Virtualization techniques have proved to be effective in reducing power consumption of network devices. An open issue in virtualization for green networking is the search for an energy-efficient mapping of virtual networks onto physical networks. This paper introduces a new model for the mapping of virtual networks which aims at reducing the energy consumption. This model is based on an integer linear programming formulation and several parameters, corresponding to characteristic of real networks, are considered. Simulation results attest the efficacy of the proposal.

Network virtualization is a promising technique for building the Internet of the future since it enables the low cost introduction of new features into network elements. An open issue in virtualization is how to search for an efficient mapping of virtual network elements onto those of the existing physical network, also called the substrate network. Mapping is an NP-hard problem and existing solutions ignore various real network characteristics in order to solve problem in a reasonable time frame. This paper introduces two new algorithms for the solution of the mapping problem, both based on 0--1 integer programming, for the solution of the mapping problem which consider a whole new set of network parameters not taken into account by previous proposals. Simulation experiments confirm the efficiency of the proposed algorithms.

Both grid and cloud applications may require specific softwares for their execution. Introducing such requirements into the DAG describing the dependencies among the tasks of an application leads to a combinatorial problem with numerous possibilities of modified DAGs. This paper introduces an algorithm to obtain a single DAG with the minimum possible makespan by reducing the need for data transfer over the network.

Grid scheduling is essential to Quality of Service provisioning as well as to efficient management of grid resources. Grid scheduling usually considers the state of the grid resources as well application demands. However, such demands are generally unknown for highly demanding applications, since these often generate data which will be transferred during their execution. Without appropriate assessment of these demands, scheduling decisions can lead to poor performance. Thus, it is of paramount importance to consider uncertainties in the formulation of a grid scheduling problem. This paper introduces the IPDT-FUZZY scheduler, a scheduler which considers the demands of grid applications with such uncertainties. The scheduler uses fuzzy optimization, and both computational and communication demands are expressed as fuzzy numbers. Its performance was evaluated, and it was shown to be attractive when communication requirements are uncertain. Its efficacy is compared, via simulation, to that of a deterministic counterpart scheduler and the results reinforce its adequacy for dealing with the lack of accuracy in the estimation of communication demands.

Grid computing emerged as a computing paradigm to take advantage of the enormous availability of computing and communication resources in the Internet. Cloud computing succeeded grid computing as a new computer paradigm, offering resource transparency to the computation of user demands. In cloud computing, virtualization techniques are central to the provisioning of resources in clouds. This paper introduces a scheduler for grid applications in clouds which takes into account not only resource demands, but also software requirements of the applications. Moreover, it accounts for network link availability that has been widely neglected in the literature of grid scheduling.Experiments show the efficacy of the scheduler in different scenarios.

This paper introduces a procedure called traffic engineering for grids for enabling grid networks to self-adjust to resource availability. The proposal is based on monitoring the state of resources and on task migration. It involves several layers of the Internet architecture. Experiments executed in NS-2 are used to illustrate the efficacy of the procedure proposed.