The emerging heterogeneous Edge-IoT systems are supporting various essential and critical applications in public safety, transportation, education, and health. The reliability and resiliency of such systems are crucial given the varied demands of IoT applications and the potential for edge failure. This paper tackles two key Edge-IoT resiliency challenges: i) how to cushion the impact of sudden workload increase and allow graceful performance degradation for heterogeneous applications; ii) how to better prepare for and handle severe disruptions to avoid significant application performance loss. These are complicated challenges due to limited edge resources, heterogeneity of IoT applications, and the dynamicity of the Edge-IoT environment. Therefore, we develop a novel intelligent resource allocation framework namedREPAthat is application-centric to enable proactive resiliency and offer graceful degradation of IoT applications. The new framework: 1) proactively cushions and absorbs the disruptive impacts of edge workload increases, and provides graceful application degradation in moderate to severe edge system loads, across multiple Edge-IoT systems; 2) features a two-level optimization model that jointly optimizes the intra-zone and inter-zone edge resource allocation such that the application’s performance degradation and service disruption are minimized; 3) includes novel deep reinforcement learning (DRL) designs for resource allocation that are specialized for the complex Edge-IoT environment and provide faster and stable decision-making capabilities than the existing methods. Performance evaluation demonstrates the ability ofREPAto maintain the best performance of the IoT applications regardless of disruptive situations.

Edge computing and AI can potentially empower Unmanned Aerial Vehicle (UAV) systems with automated decision-making and resource support for monitoring in future science tasks such as emergency response, search and rescue, inspections, and wildfires. However, it is challenging to achieve autonomous and robust monitoring in such systems, given the dynamic environmental situations, the limited capabilities, and the unbalanced load of the UAVs. For instance, the monitoring activity levels at different locations might vary, which leads to an unbalanced monitoring load for the corresponding UAVs. Moreover, the UAVs require regular recharging/maintenance and can have malfunctions that will disrupt the monitoring task. In this article, we develop a novel proactive and robust Edge-UAV framework named PREUS to enable autonomous and efficient monitoring of dynamic environments when faced with dynamic environment situations and various UAV workload stresses that can jeopardize the monitoring performance. PREUS features a unique design to handle the varying UAV workload stress of the monitored area. It incorporates novel spatial, temporal, and proactive exploration vs. exploitation planning to balance the UAVs’ workloads in various locations with fluctuating activities. In addition, PREUS includes novel Deep Reinforcement Learning (DRL) design specialized to maximize coverage in the complex environments and provides faster and stabler decision-making capabilities than the existing methods. The positive impact brought by PREUS is demonstrated in terms of the achieved monitoring performance, including coverage and balanced UAV load.

Edge computing is being used to facilitate closer computing, storage, and networking resources to support various IoT applications including delay-sensitive ones. It is envisioned that the future Edge-IoT systems will incorporate heterogeneous IoT devices distributed over multiple geographical zones of certain institutions with edge resource demands that vary according to time and location. Edge servers (resource facilitators) are with limited resources and are susceptible to “outlandish” situations, such as service overloading, outage, and external attacks; they may also have to handle the roaming of IoT devices among different zones. These situations induce the need for alternative edge servers using an adaptive resource facilitation scheme to fulfill the demands of the IoT applications. In this article, we develop a novel intelligent and hierarchical resource facilitation framework named I-HARF that adapts to dynamic Edge-IoT situations, including outlandish situations, mobility, application’s sensitivity, and varying resource demand of IoT applications based on time and location. I-HARF achieves an adaptive facilitation and holistically addresses the facilitation technical barriers by: 1) adopting the hierarchical structure which efficiently migrates the resource facilitation from intrazone to interzone levels; 2) extending novel intrazone and interzone optimization models to boost the utilities of the edge servers and the IoT applications; and 3) developing a novel and unique actor dual-critic and collective actor–critic deep reinforcement learning (DRL) designs that intelligently facilitate the edge resources in both intrazone and interzone, respectively. The evaluation results demonstrate I-HARF’s capability enabling adaptive resource facilitation that adjusts according to the dynamic Edge-IoT situations.

Data-driven approaches are envisioned to build future Edge-IoT systems that satisfy IoT devices demands for edge resources. However, significant challenges and technical barriers exist which complicate resource management of such systems. IoT devices can demonstrate a wide range of behaviors in the devices resource demand that are extremely difficult to manage. In addition, the management of resources fairly and efficiently by the edge in such a setting is a challenging task. In this paper, we develop a novel data-driven resource management framework named BEHAVE that intelligently and fairly allocates edge resources to IoT devices with consideration of their behavior of resource demand (BRD). BEHAVE aims to holistically address the management technical barriers by 1) building an efficient scheme for modeling and assessment of the BRD of IoT devices based on their resource requests and resource usage; 2) expanding a new Rational, Fair, and Truthful Resource Allocation (RFTA) model that binds the devices BRD and resource allocation to achieve fair allocation and encourage truthfulness in resource demand; and 3) developing an enhanced deep reinforcement learning (EDRL) scheme to achieve the RFTA goals. The evaluation results demonstrate BEHAVE's capability to analyze the IoT devices BRD and adjust its resource management policy accordingly.

Network intrusion detection plays an important role in the Internet of Things systems for protecting devices from security breaches. Facing challenges of the rapidly increasing amount of diverse network traffic, recent research has employed end-to-end deep learning-based intrusion detectors for automatic feature extraction and high detection accuracy. However, deep learning has been proved vulnerable to adversarial attacks that may cause misclassification by imposing imperceptible perturbation on input samples. Though such vulnerability is widely discussed in the image processing domain, very few studies have investigated its perniciousness against network intrusion detection systems (NIDS) and proposed corresponding defense strategies. In this paper, we try to fill this gap by proposing Def-IDS, an ensemble defense mechanism specially designed for NIDS, against both known and unknown adversarial attacks. It is a two-module training framework that integrates multi-class generative adversarial networks and multi-source adversarial retraining to improve model robustness, while the detection accuracy on unperturbed samples is maintained. We evaluate the mechanism over CSE-CIC-IDS2018 dataset and compare its performance with the other three defense methods. The results demonstrate that Def-IDS is able to detect various adversarial attacks with better precision, recall, F1 score, and accuracy.

Edge computing is emerging to empower the future of Internet of Things (IoT) applications. However, due to heterogeneity of applications, it is a significant challenge for the edge cloud to effectively allocate multidimensional limited resources (CPU, memory, storage, bandwidth, etc.) with constraints of applications’ Quality of Service (QoS) requirements. In this paper, we address the resource allocation problem in Edge-IoT systems through developing a novel framework namedDeepEdgethat allocates resources to the heterogeneous IoT applications with the goal of maximizing users’ Quality of Experience (QoE). To achieve this goal, we develop a novel QoE model that considers aligning the heterogeneous requirements of IoT applications to the available edge resources. The alignment is achieved through selection of QoS requirement range that can be satisfied by the available resources. In addition, we propose a novel two-stage deep reinforcement learning (DRL) scheme that effectively allocates edge resources to serve the IoT applications and maximize the users’ QoE. Unlike the typical DRL, our scheme exploits deep neural networks (DNN) to improve actions’ exploration by using DNN to map the Edge-IoT state to joint resource allocation action that consists of resource allocation and QoS class. The joint action not only maximize users’ QoE and satisfies heterogeneous applications’ requirements but also align the QoS requirements to the available resources. In addition, we develop a Q-value approximation approach to tackle the large space problem of Edge-IoT. Further evaluation shows thatDeepEdgebrings considerable improvements in terms of QoE, latency and application tasks’ success ratio in comparison to the existing resource allocation schemes.

The considerable growth in demands for wireless services have led to spectrum scarcity challenge. Cognitive radio came into practice to deal with the scarcity problem by granting cognitive users access to the licensed spectrum. However, this solution requires efficient power allocation strategies to guarantee QoS for cognitive system, reduce power consumption, and protect primary users from the cognitive users' interference impact. In this paper, we investigate the energy efficient power allocation problem for cognitive radio networks in underlay mode. We propose a novel approximated online Q-learning scheme for power allocation in which cognitive users learn with conjecture feature to select the most appropriate power level. The power allocation problem is formulated as an optimization problem with the goal to maximize energy efficiency under QoS and interference constraints. The scheme is evaluated using software defined radio testbed and simulations. The evaluation results demonstrate the scheme capability to guarantee SINR for both primary and cognitive systems and mitigate interference with minimum power consumption in comparison with other schemes.

The emerging Internet of Things (IoT) is facing significant scalability and security challenges. On one hand, IoT devices are “weak” and need external assistance. Edge computing provides a promising direction addressing the deficiency of centralized cloud computing in scaling massive number of devices. On the other hand, IoT devices are also relatively “vulnerable” facing malicious hackers due to resource constraints. The emerging blockchain and smart contracts technologies bring a series of new security features for IoT and edge computing. In this paper, to address the challenges, we design and prototype an edge-IoT framework named “EdgeChain” based on blockchain and smart contracts. The core idea is to integrate a permissioned blockchain and the internal currency or “coin” system to link the edge cloud resource pool with each IoT device' account and resource usage, and hence behavior of the IoT devices. EdgeChain uses a credit-based resource management system to control how much resource IoT devices can obtain from edge servers, based on predefined rules on priority, application types, and past behaviors. Smart contracts are used to enforce the rules and policies to regulate the IoT device behavior in a nondeniable and automated manner. All the IoT activities and transactions are recorded into blockchain for secure data logging and auditing. We implement an EdgeChain prototype and conduct extensive experiments to evaluate the ideas. The results show that while gaining the security benefits of blockchain and smart contracts, the cost of integrating them into EdgeChain is within a reasonable and acceptable range.

Cognitive radio attracts researchers' attention recently in radio resource management due to its ability to exploit environment awareness in configuring radio system parameters. Cognitive engine (CE) is the structure known for deciding system parameters' adaptation using optimization and machine learning techniques. However, these techniques have strengths and weaknesses depending on the experienced network scenario that make one more appropriate than others. In this paper, we propose a novel design for the cognitive system called supervised cognitive system (SCS), which aims to perform radio parameters adaptation with the most appropriate CE learning technique for the encountered network scenario. To realize SCS, it is required to evaluate the performance of different CEs in different network scenarios and according to certain performance objectives. In addition, the ability to select the most appropriate CE learning technique for adaptation in the current network scenario is also a priority in our design. Therefore, SCS investigates the relationship between learning and performance improvement and it employs online learning to classify scenarios and select the most appropriate CE learning technique. The testbed implementation and evaluation results in terms of goodput, packet error rate, and spectral efficiency show that the proposed SCS achieves more than 50% in performance gain compared to the best standalone CE.

5G is the upcoming evolution for the current cellular networks that aims at satisfying the future demand for data services. Heterogeneous cloud radio access networks (H-CRANs) are envisioned as a new trend of 5G that exploits the advantages of heterogeneous and cloud radio access networks to enhance spectral and energy efficiency. Remote radio heads (RRHs) are small cells utilized to provide high data rates for users with high quality of service (QoS) requirements, while high power macro base station (BS) is deployed for coverage maintenance and low QoS users service. Inter-tier interference between macro BSs and RRHs and energy efficiency are critical challenges that accompany resource allocation in H-CRANs. Therefore, we propose an efficient resource allocation scheme using online learning, which mitigates interference and maximizes energy efficiency while maintaining QoS requirements for all users. The resource allocation includes resource blocks (RBs) and power. The proposed scheme is implemented using two approaches: centralized, where the resource allocation is processed at a controller integrated with the baseband processing unit and decentralized, where macro BSs cooperate to achieve optimal resource allocation strategy. To foster the performance of such sophisticated scheme with a model free learning, we consider users' priority in RB allocation and compact state representation learning methodology to improve the speed of convergence and account for the curse of dimensionality during the learning process. The proposed scheme including both approaches is implemented using software defined radios testbed. The obtained results and simulation results confirm that the proposed resource allocation solution in H-CRANs increases the energy efficiency significantly and maintains users' QoS.

Network efficiency and proper utilization of its resources are essential requirements to operate wireless networks in an optimal fashion. Cognitive radio aims to fulfill these requirements by exploiting artificial intelligence techniques to create an entity called cognitive engine. Cognitive engine exploits awareness about the surrounding radio environment to optimize the use of radio resources and adapt relevant transmission parameters. In this paper, we propose a hybrid cognitive engine that employs Case Based Reasoning (CBR) and Decision Trees (DTs) to perform radio adaptation in multi-carriers wireless networks. The engine complexity is reduced by employing DTs to improve the indexing methodology used in CBR cases retrieval. The performance of our hybrid engine is validated using software defined radios implementation and simulation in multi-carrier environment. The system throughput, signal to noise and interference ratio, and packet error rate are obtained and compared with other schemes in different scenarios.

Heterogeneous cloud radio access networks (H-CRAN) is a new trend of SC that aims to leverage the heterogeneous and cloud radio access networks advantages. Low power remote radio heads (RRHs) are exploited to provide high data rates for users with high quality of service requirements (QoS), while high power macro base stations (BSs) are deployed for coverage maintenance and low QoS users support. However, the inter-tier interference between the macro BS and RRHs and energy efficiency are critical challenges that accompany resource allocation in H-CRAN. Therefore, we propose a centralized resource allocation scheme using online learning, which guarantees interference mitigation and maximizes energy efficiency while maintaining QoS requirements for all users. To foster the performance of such scheme with a model-free learning, we consider users' priority in resource blocks (RBs) allocation and compact state representation based learning methodology to enhance the learning process. Simulation results confirm that the proposed resource allocation solution can mitigate interference, increase energy and spectral efficiencies significantly, and maintain users' QoS requirements.

The demand on mobile Internet related services has increased the need for higher bandwidth in cellular networks. The 5G technology is envisioned as a solution to satisfy this demand as it provides high data rates and scalable bandwidth. The multi-tier heterogeneous structure of 5G with dense base station deployment, relays, and device-to-device (D2D) communications intends to serve users with different QoS requirements. However, the multi-tier structure causes severe interference among the multi-tier users which further complicates the resource allocation problem. In this paper, we propose a cooperative scheme to tackle the interference problem, including both cross-tier interference that affects macro users from other tiers and co-tier interference, which is among users belong to the same tier. The scheme employs an online learning algorithm for efficient spectrum allocation with power and modulation adaptation capability. Our evaluation results show that our online scheme outperforms others and achieves significant improvements in throughput, spectral efficiency, fairness, and outage ratio. © 2016 IEEE.

Radio resource management becomes an important aspect of the current wireless networks because of spectrum scarcity and applications heterogeneity. Cognitive radio is a potential candidate for resource management because of its capability to satisfy the growing wireless demand and improve network efficiency. Decision-making is the main function of the radio resources management process as it determines the radio parameters that control the use of these resources. In this paper, we propose an adaptive decision-making scheme (ADMS) for radio resources management of different types of network applications including: power consuming, emergency, multimedia, and spectrum sharing. ADMS exploits genetic algorithm (GA) as an optimization tool for decision-making. It consists of the several objective functions for the decision-making process such as minimizing power consumption, packet error rate (PER), delay, and interference. On the other hand, maximizing throughput and spectral efficiency. Simulation results and test bed evaluation demonstrate ADMS functionality and efficiency.

Cognitive Radio is an intelligent Software Defined Radio that is capable to alter its transmission parameters according to predefined objectives and wireless environment conditions. Cognitive engine is the actuator that performs radio parameters configuration by exploiting optimization and machine learning techniques. In this paper, we propose an Adaptive Multi-objective Optimization Scheme (AMOS) for cognitive radio resource management to improve spectrum operation and network performance. The optimization relies on adapting radio transmission parameters to environment conditions using constrained optimization modeling called fitness functions in an iterative manner. These functions include minimizing power consumption, Bit Error Rate, delay and interference. On the other hand, maximizing throughput and spectral efficiency. Cross-layer optimization is exploited to access environmental parameters from all ТСРЯР stack layers. AMOS uses adaptive Genetic Algorithm in terms of its parameters and objective weights as the vehicle of optimization. The proposed scheme has demonstrated quick response and efficiency in three different scenarios compared to other schemes. In addition, it shows its capability to optimize the performance of ТСРЯР layers as whole not only the physical layer.

With the increasing adoption of wireless communication technologies, there is a need to improve management of existing radio resources. Cognitive radio is a promising technology to improve the utilization of wireless spectrum. Its operating principle is based on building an integrated hardware and software architecture that configures the radio to meet application requirements within the constraints of spectrum policy regulations. However, such an architecture must be able to cope with radio environment heterogeneity. In this paper, we propose a cognitive resource management architecture, called CogWnet, that allocates channels, re-configures radio transmission parameters to meet QoS requirements, ensures reliability, and mitigates interference. The architecture consists of three main layers: Communication Layer, which includes generic interfaces to facilitate the communication between the cognitive architecture and TCP/IP stack layers; Decision-Making Layer, which classifies the stack layers input parameters and runs decision-making optimization algorithms to output optimal transmission parameters; and Policy Layer to enforce policy regulations on the selected part of the spectrum. The efficiency of CogWnet is demonstrated through a testbed implementation and evaluation.

The explosive growth in mobile Internet and related services has increased the need for more bandwidth in cellular networks. The Long-Term Evolution (LTE) technology is an attractive solution for operators and subscribers to meet such need since it provides high data rates and scalable bandwidth. Radio Resource Management (RRM) is essential for LTE to provide better communication quality and meet the application QoS requirements. Cognitive resource management is a promising solution for LTE RRM as it improves network efficiency by exploiting radio environment information, intelligent optimization algorithms to configure transmission parameters, and mitigate interference. In this paper, we propose a cognitive resource management scheme to adapt LTE network parameters to the environment conditions. The scheme optimizes resource blocks assignment, modulation selection and bandwidth selection to maximize throughput and minimize interference. The scheme uses constrained optimization for throughput maximization and interference control. It is also enhanced by learning mechanism to reduce the optimization complexity and improve the decision-making quality. Our evaluation results show that our scheme achieved significant improvements in throughput and LTE system capacity. Results also show the improvement in the user satisfaction over other techniques in LTE RRM.