Data privacy and eXplainable Artificial Intelligence (XAI) are two important aspects for modern Machine Learning models. To enhance data privacy, recent machine learning models have adopted Federated Learning (FL). On top of that, additional privacy layers can be added, notably Differential Privacy (DP). On the other hand, to improve explainability, ML must consider more interpretable approaches with reduced number of features and less complex internal architecture. In this context, this paper aims to achieve a Machine Learning (ML) model that combines enhanced data privacy with explainability. So, we propose a FL solution, called Federated EXplainable Trees with Differential Privacy (FEXT-DP), that: (i) is based on Decision Trees, since they are lightweight and have superior explainability to neural networks-based FL models; (ii) provides additional layer of data privacy protection applying Differential Privacy (DP) to the Tree-Based model. However, there is a side effect adding DP: it harms the explainability of the system. So, this paper also presents the impact of DP protection on the explainability of the ML model. The carried out performance assessment shows the results of FEXT-DP in terms of numbers of rounds, Mean Squared Error and explainability.

Smartphones contain a significant amount of critical user information. Without proper security measures, this information can be at a risk of privacy leakage from unattended devices. To address this issue, we propose a deep learning-based continuous authentication mechanism using a combination of CNN, BiLSTM and attention. In the proposed continuous authentication model, the CNNs effectively learn spatial features from raw sensor data, the BiLSTMs effectively capture the temporal features of user behaviour patterns while the attention mechanism underscores important features. The proposed model achieves 98.8% accuracy and a 3.7% Equal Error Rate (EER) on the Extrasensory dataset, outperforming existing state-of-the-art continuous authentication models. To further enhance model transparency, we apply Local Interpretable Model-agnostic Explanations (LIME) within the Explainable AI (XAI) framework. LIME emphasizes the features that most influence each authentication decision. This interpretability not only builds trust in the proposed model but also supports better understanding of how specific user behaviors contribute to authentication outcomes.

The Internet of Things (IoT) increasingly relies on edge computing nodes to decentralize computation and enhance processing power near IoT devices. However, IoT edge computing nodes are generally not designed for highly intensive machine learning (ML) training. In current IoT architectures, multiple edge computing nodes are strategically positioned near IoT devices, each accessing only a portion of the data generated by the entire IoT network. In this paper, we bring the concept of Federated Learning (FL) to this scenario, by enabling each IoT edge computing node to run lightweight ML models on local datasets cooperatively. Our primary goal is to design a decision tree-based solution for cooperative IoT edge computing, termed Federated Decision Trees (FeDT). To achieve this, we propose four FL strategies based on decision trees, which aggregate the learning contributions of multiple FL clients while adhering to FL principles. Our results demonstrate that the proposed strategies can achieve approximately $80 \%$ of the performance of a centralized ML model in terms of Pearson correlation. Furthermore, compared to FedAVG, a classical FL solution, FeDT requires approximately four times fewer training rounds to converge.

Technological advancements in cellular networks have enabled to surpass many challenges in telecommunications, but some features remain restricted, such as throughput, packet loss, and latency. User equipment (UE), including mobile, smart devices, vehicles, IoT devices, and smart city infrastructure, requires seamless connectivity to share data and resources effectively. Multiple radio access technology (multi-RAT) scenarios offer a solution to the limitations of individual RATs by combining their strengths. Determining the optimal RAT for traffic steering (TS) in multi-RAT scenarios is challenging due to factors such as high mobility, ultra-dense networks, overall dynamic network conditions, and the unique needs of individual users. In this context, we propose a bi-level approach, called BIL-TS, which includes (i) centrally determining the optimality of RATs and (ii) locally making TS decisions. BIL-TS utilizes the Actor-Critic SARSA Reinforcement Learning (ACS-RL) in level (i) to evaluate the optimality of RATs by considering the entire network, and level (ii) leverages Linear Regression (LR) to make decisions of TS to the optimal RAT based on specific requirements of each UE. Simulation results show that our proposed BILTS approach significantly enhances efficiency in TS, resulting in higher throughput, reduced packet loss, and lower latency.

As Internet of Things (IoT) networks continue to evolve, they face increasing security threats, and methods to achieve the security properties and requirements of these networks from the perspective of data, communication and IoT device security, are still on demand. This paper focuses on Recurrent neural network (RNN)-based methods, which provide intrusion detection systems (IDSs) with the capability of analyzing unseen and complex patterns in exchanges between IoT devices. Two flow-based optimized standalone RNNbased IDSs (called Uni-Hybrid and Bi-Hybrid RNNbased IDSs) are proposed to enhance the security of IoT networks. Through experiments using the IoTID20 dataset, the proposed models yield some marked improvements over a chosen benchmark model in terms of precision, accuracy, recall, and F1-score, highlighting the potential of RRN-based models in addressing intrusion detection in IoT networks.

5G technology offers high bandwidth, stability, and reliability among connected vehicles, which is necessary for increasing data sharing in intelligent transportation. While providing these benefits with its small cellular range and densification, it also presents a challenge in frequent handovers (HOs). This issue can result in unnecessary HO, HO failures, and ping-pong effects, negatively impacting service delivery and compromising safety data sharing. A learning-oriented proactive HO decision-making strategy can ensure connection stability by making HO decisions based on real-time scenarios. This paper presents a high mobility and ultra-dense network-aware proactive HO decision-making (PAHD) approach, efficiently ensuring stable connectivity by predicting future HO. PAHD consists of two parts (i) Gaussian Process Regression for mobility prediction and (ii) Bidirectional Long Short-Term Memory for the prediction of network traffic density. Realistic simulated analyses have shown that PAHD significantly improves efficiency in HO decision-making.

Ensuring connection stability is crucial for both vehicular safety and user experience. With the increasing amount of data sharing among connected vehicles, there is a need for more bandwidth, stability, and reliability. While 5G technology can offer these benefits with its small cellular range and densification, it also presents a challenge in frequent handovers (HOs). This issue can result in unnecessary HO, HO failures, and ping-pong effects, negatively impacting service delivery and compromising safety data sharing. To this end, we present High- mobility and Ultra-density Aware Handover decision-making (HMUD-H) approach using the SARSA Reinforcement Learning algorithm for connection management, which efficiently makes HO decisions to ensure stable connectivity. The HMUD-H algorithm is adaptable and can handle dynamic, highly mobile, and ultra-dense vehicular networks. Realistic simulated analyses have demonstrated that our algorithm significantly reduces the number of HOs, average cumulative HO time, HO failures, and ping-pong effects, thus improving overall connection stability.

5G and vehicular networks have enabled Intelligent Transportation Systems (ITS) with better safety and infotainment services where connected vehicles are critical components for data sharing. However, a stable connection is mandatory to transmit data successfully across the network. The 5G technology enhances bandwidth, stability, and reliability but suffers from low communication ranges, which results in frequent and unnecessary handovers and connection drops. In this paper, we introduce a user-centric approach, Factor-distinct SARSA Reinforcement Learning (FD-SRL), which combines a time series data-oriented model LSTM and adaptive method SARSA Reinforcement Learning for Virtual Cell (VC) and handover (HO) management. Our proposed approach maintains stable connections by reducing the number of HOs, given the fast-paced changes due to mobility, network load, and communication conditions. Realistic simulations demonstrated that FD-SRL reduced the number of HOs and the average cumulative HO time, showing potential improvements in connection stability for 5G-based ITS.

Connected vehicles enable massive data sharing and support intelligent transportation services. Consequently, a stable connection is compulsory to transmit across the network successfully, where 5G technology introduces more bandwidth, stability, and reliability. However, 5G communication is susceptible to frequent handovers and connection drops. A user-centric perspective helps cope with the smaller communication range in ultra-dense 5G networks. We thus introduce a Connectivity-oriented SARSA Reinforcement Learning (CO-SRL) algorithm for user-centric to efficiently handle virtual cell (VC) management and reduce the number of handovers (HO). The adaptability of the algorithm copes with high vehicular mobility and dynamic traffic and communication, deciding on in-rage cellular towers and VC size. Realistic simulated analyses showed CO-SRL reduced the number of handovers and the cumulative handover time.

This paper proposes an optimal admission and placement stochastic controller that inserts security and depend-ability in the operational aspects of edge-cloud system under a 5G deployment. The proposed mechanism uses the frame-work of Semi-Markov Decision Making Process (SMDP) and seeks for an optimal policy that efficiently allocates the virtual resources to secure and run the services across the cloudified infrastructure. Driven by a new latency-oriented cost structure, the optimal controller achieves a dependable and secure operation by optimally balancing the service requests between the edge and the cloud system taking into account the service profile, the workload, and the traffic load. A structural analysis of the optimal policy reveals its implementation friendliness while a cloudnomics analysis shows that the optimal cost can be further optimized by fine tuning the parameters of the proposed cost structure.

Cloud firewalls stand as one of the major building blocks of the cloud security framework protecting the Virtual Private Infrastructure against attacks such as the Distributed Denial of Service (DDoS). In order to fully characterize the cloud firewall operation and gain actionable insights on the design of cloud security, performance models for the cloud firewall become imperative. In this article, we propose a multi-dimensional Continuous-Time Markov Chain model for the cloud firewall that takes into account the burstiness and correlation features of the legitimate and malicious data traffic. By adopting the Markov-Modulated Poisson process (MMPP) and the Interrupted Poisson Process (IPP), we identify the workload conditions under which the cloud firewall might be subject to a loss of availability. Furthermore, by comparing the IPP and Poisson attacks, we numerically verify that the cloud firewall is inherently vulnerable to a burstiness-aware attack which might seriously compromise its operation. Additionally, we characterize the joint harmful impact of burstiness and correlation on the cloud firewall that might lead to performance degradation. Finally, we design an elastic cloud firewall by proposing a MMPP-driven load balancing procedure that provisions virtual firewalls dynamically while fulfilling a Service Level Agreement (SLA) latency specification.

This work proposes an optimal security risk management mechanism to holistically minimize the risks of a Denial of Service (DoS) attack and Service Level Agreement (SLA) violations that might unfold at the 5G edge-cloud ecosystem. Using the Semi-Markov Decision Process framework, a cyber risk-aware controller is designed to optimally decide on the admission, placement, and migration of a service taking into consideration a user taxonomy and the service requirements. A new cost structure that balances the targeted security risks as well as the cost and the reward of a secure service provisioning is introduced to pave the way for a safe edge-cloud operation. To proactively restrict the population of untrusted users, we consider security controls in the form of a linear and an exponential cost functions and show that the former represents a more flexible and profitable pathway for a Mobile Network Operator to operate at the expense of an inflated security risk while the latter leads to the opposite outcome. Results show that the baseline mechanism might violate the SLA and expose the edge and the cloud to a DoS attack in levels that are 102, 1012, and 1014times higher than those of the proposed controller.

As the multitude and complexity of the cloud market increases, the evaluation and selection of cloud services becomes a burdensome task for the users. With the extraordinary rise of available services from various Cloud Service Providers (CSPs), the role of cloud brokers has become more and more important. This paper proposes an optimal cloud broker model to address the challenge of optimally allocating multiple cloud system resources to multiple mobile user's requests with different requirements. The cloud brokering mechanism is formulated as a Semi-Markov Decision Process (SMDP) model under the average system cost criteria. The overall system cost takes into consideration the cost of occupying computing resources, the communication costs, the request traffic, as well as various security risk degrees and resource requirements from the various mobile users. Through minimizing the overall system cost, the optimal resource allocation policy is calculated by means of the Value Iteration Algorithm. Some analysis are conducted and numerical results are presented, demonstrating the feasibility of the proposed cloud broker design.

In Opportunistic networks (OppNets), due to high mobility, short radio range, intermittent links, unstable topology, sparse connectivity, to name a few, routing is a very challenging task since it relies on cooperation between the nodes. This paper focuses on using the concept of centrality to alleviate this task. Unlike other nodes in the network, central nodes are those that are more likely to act as communication hubs to facilitate the message forwarding and thereby routing. In this paper, a recently proposed History-Based Prediction Routing protocol (HBPR) for OppNets is re-designed using this concept, yielding the so-called centrality-based HBPR (CHBPR) routing protocol. The proposed CHBPR scheme is evaluated by simulations using the Opportunistic NEtwork (ONE) simulator, showing superior performance compared to HBPR without centrality and Epidemic protocol with centrality, in terms of number of messages delivered at destination and overhead ratio, under varying number of nodes and Time-to-Live.

With the ever increasing data and computational demands from mobile users, heterogenous wireless networks (HetNets) and mobile cloud computing (MCC) have been advocated as a promising solution to meet these demands. Insufficient bandwidth is one of the most important challenges being faced by a successful implementation of the MCC technology due to heavy data traffic. The MCC implementation on HetNet increases the bandwidth available to each base station (BS) by frequency reuse. In this paper, a novel data storage method for big data files is proposed, along with a data correction technique to deal with the issue of failure of a data chunk retrieval. The proposed algorithm exploits the multiple paths that are available between a user and the cloud storage system in a MCC-HetNet environment. Since the bottleneck in the MCC is the wireless link between the user equipment (UE) and the BS, we have implemented the algorithm on the wireless links between the UE and the BS. The simulated results show that the proposed method outperforms the conventional data storage method between the mobile device and the cloud system.

In this paper, we propose an optimal Joint Call Admission Control (JCAC) in next generation wireless networks, where different radio access technologies (RAT) coexist in a co-located way. Moreover, we study the impact of the different RAT's radius coverage area in the system performance and optimal policy structure. The Semi-Markov Decision Process (SMDP) framework is used for modeling of the JCAC proposed and the value iteration algorithm is used to compute the optimal policy. Numerical results show that variation in the proportionality of the radius of RAT coverage area impact on system performance.