In federated learning (FL), achieving high model accuracy often comes at the cost of increased energy consumption, particularly when incorporating robust privacy-preserving techniques like homomorphic encryption (HE). While HE enhances security by enabling encrypted aggregation of local models—thus mitigating data leakage and inference attacks without requiring a trusted aggregator—it also introduces computational overhead that impacts energy efficiency.This paper presents a novel approach for Adaptive Encryption Federated Learning (AEFL). AEFL strategically tunes encryption complexity to achieve the best trade-off between model accuracy and energy efficiency without compromising security. When additional energy from harvested sources is available, AEFL enhances model accuracy by adjusting encryption parameters to support more intensive computations. During periods of limited energy, the system focuses on conserving resources while maintaining robust encryption to ensure data security. This adaptive approach maximizes energy efficiency without lowering the security threshold, allowing for improved model performance when energy conditions permit.Experimental results show that AEFL improves energy efficiency by over 35% and increases the pool of eligible clients for FL by approximately 18%, all while maintaining a security level above 128 bits. Importantly, these gains are achieved with minimal impact on model accuracy, demonstrating that significant energy savings are possible in FL environments without sacrificing data security or model performance.

In the rapidly advancing field of wireless communication, Unmanned Aerial Vehicles (UAVs) have become indispensable due to their extensive coverage capabilities and ability to access remote locations. Whether deployed as mobile base stations (BSs) or relays, UAVs significantly enhance network throughput and reliability. Alongside UAVs, Intelligent Reflecting Surfaces (IRS) have emerged as a cost-effective solution for improving communication quality through passive modulation arrays. Despite these advancements, the potential misuse of UAVs poses serious security risks, particularly in the form of communication eavesdropping. To address these challenges, this paper introduces a novel communication framework that integrates a UAV equipped with an adaptive IRS. The primary aim is to boost communication secrecy between BSs and multiple users, even in the presence of several UAV eavesdroppers. This objective is formulated as an optimization problem focused on maximizing the secrecy rate while considering UAV mobility constraints. To solve this non-convex problem, we propose a hybrid strategy that combines Genetic Algorithms and Gradient Descent techniques. This innovative approach efficiently determines suboptimal reflection angles and UAV trajectories for IRS-equipped UAVs, thereby enhancing the security of the communication network. This method not only addresses the complexity of the optimization but also provides a practical pathway to secure communications in environments with high eavesdropping risks.

Federated learning (FL) has gained attention for enabling efficient distributed learning while maintaining data privacy. However, the data privacy constraint reduces the transparency in the agents’ model update making the learning process vulnerable to Byzantine attacks. In this paper, a mathematical proof is provided to show that when the traditional model-combining scheme is used, the model will eventually diverge to non-useful solutions in the presence of Byzantine agents independently from their number or their contributions. A low complexity norm-control based aggregation approach is also proposed and shown to converge to the optimal and sub-optimal solutions in the absence or presence of Byzantine nodes, respectively. Monte-Carlo simulations are also conducted to verify and validate the mathematical derivations and the efficiency of the proposed approach in protecting the FL model.

The identification of malicious users within a large set of participants poses a significant challenge in the domains of cybersecurity, data integrity, user management, and particularly within federated learning (FL) environments. FL, a distributed machine learning approach, necessitates rigorous mechanisms for safeguarding data integrity, model accuracy by effectively managing and identifying malicious participants. Traditional methods require the sequential removal and evaluation of users to determine their impact on the system’s overall error rate or loss function, fall short in terms of efficiency and scalability, especially in FL contexts where data is distributed across multiple clients. To address these limitations, we propose the Refine and Identify Algorithm, a two-phased approach that efficiently narrows the search space for identifying malicious users by initially evaluating users in groups rather than individually and iteratively focusing on those groups with the highest potential for containing malicious users. A rigorous mathematical framework, including a proof of convergence and a detailed analysis of iteration necessities, underpins the algorithm’s efficacy. The convergence proof and analysis of iteration requirements provide a solid mathematical foundation for the proposed method’s effectiveness, paving the way for further optimization and application-specific tuning. Simulation results depict the efficiency of the proposed technique and show a significant reduction in computational resources and time required for identifying malicious users.

In the dynamic realm of wireless communication, Unmanned Aerial Vehicles (UAVs) have gained increasing prominence due to their exceptional capabilities, which include expensive coverage of large areas and access to challenging and hazardous locations. When employed as mobile base stations or relays, UAVs have shown remarkable enhancements in system throughput and reliability. In addition, Intelligent Reflecting Surfaces (IRS) present a very low cost solution that efficiently enhances wireless communication quality using passive modulation arrays. Nevertheless, the use of UAVs for malicious intents can also introduce heightened security challenges such as communication eavesdropping. In response to these challenges, we present in this paper, a communication framework that incorporates a UAV equipped with an adaptive IRS aiming to enhance the communication secrecy between the Base Station (BS) and Bob in the presence of several UAV eavesdroppers. We formulate the objective as an optimization problem that aims to maximize the secrecy rate while considering the mobility constraints. To address this complex non-convex problem, our innovative solution harnesses a hybrid approach that combines Genetic Algorithms and Gradient Descent techniques, resulting in an efficient computation of suboptimal reflection angles and UAV trajectories for IRS-equipped UAVs towards a more secure communication.

Despite its potential benefits, Federated learning (FL) is vulnerable to various types of attacks that can compromise the accuracy and security of the trained model. While several defense mechanisms have been proposed to protect FL against such attacks, attackers are continuously developing more advanced techniques to bypass these protection mechanisms.In this context, this paper proposes a novel attack mechanism that allows malicious users to optimize their crafted reports, maximizing potential damage while limiting the chances of being detected. Our proposed attack technique is a robust approach designed to bypass existing defense mechanisms in FL. Our contributions are mainly investigating the FL model attack from the attacker’s perspective, proposing a model relaxation approach to optimize a single poisoning ratio variable, and formulating a compromise between the chances of being detected and the amount of damage that the attack could cause. Additionally, we introduce three new attack designs, namely DTA, ATA, and NEA, which maximize the effect of the attack. The proposed Distance Target Attack (DTA) minimizes the distance from the target attack model, while the Accuracy Target Attack (ATA) deteriorates the accuracy of the global model. Furthermore, the Number Estimation Attack (NEA) aims to maximize the expected number of attackers that could bypass the aggregation detection mechanisms.The numerical results based on the KDD dataset confirm the ability of the proposed approach to deteriorate the global model accuracy. The experiments showed that the proposed DTA, ATA, and NEA attacks can significantly reduce the accuracy of the global model. These results demonstrate also the effectiveness and robustness of the proposed attack mechanism in compromising the accuracy and security of FL models.

This paper addresses the problem of coexistence between smart grid based wireless communication networks and other interfering signals. Particularly, the problem of interfering wireless local area network (WLAN) signals over smart utility networks (SUN) systems is analyzed. We develop a statistical model of the WLAN interferers, which is in turn used in predicting the capacity of SUN systems when affected by several WLAN interferers. Based on this analysis, a framework that promotes the application of an adaptive power control algorithm at the SUN transmitter is developed based on a Lagrangian optimization approach. Our proposed optimization approach involves solving a reduced-complexity algorithm which yields an optimal power allocation for each SUN packet. The results show that an adaptive power control approach results in a significant improvement of the SUN’s network capacity when compared with fixed power transmissions.

Federated learning (FL) enables an effective and private distributed learning process. However, it is vulnerable against several types of attacks, such as Byzantine behaviors. The first purpose of this work is to demonstrate mathematically that traditional arithmetic-averaging model-combining approach will ultimately diverge to an unstable solution in the presence of Byzantine agents. This article also proposes a low-complexity, decentralized Byzantine resilient training mechanism. The proposed technique identifies and isolates hostile nodes rather than just mitigating their impact on the global model. In addition, the suggested approach may be used alone or in conjunction with other protection techniques to provide an additional layer of security in the event of misdetection. The suggested solution is decentralized, allowing all participating nodes to jointly identify harmful individuals using a novel cross-check mechanism. To prevent biased assessments, the identification procedure is done blindly and is incorporated into the regular training process. A smart activation mechanism based on flag activation is also proposed to reduce the network overhead. Finally, general mathematical proofs combined with extensive experimental results applied in a healthcare electrocardiogram (ECG) monitoring scenario show that the proposed techniques are very efficient at accurately predicting heart problems.

Federated Learning (FL) has attracted a lot of attention in numerous applications due to recent data privacy regulations and increased awareness about data handling issues, combined with the ever-increasing big-data sizes. This paper proposes a server-less, robust FL training mechanism that allows any set of participating data-owners to train a neural network (NN) model collaboratively without the assistance of any central node and while being resilient to Byzantine attacks. The proposed approach makes use of a dual-way update mechanism to allow each node to take a model forwarding decision towards a global collaborative decision of isolating any malicious updates. The efficiency of the proposed approach in detecting cardiac irregularities is verified using simulation results conducted based on the Physikalisch-Technische Bundesanstalt Database electro-cardiogram (PTBDB ECG) dataset.

Federated Learning (FL) is a machine learning approach that enables private and decentralized model training. Although FL has been shown to be very useful in several applications, its privacy constraints cause a lack of model update transparency which makes it vulnerable to several types of attacks. In particular, based on detailed convergence analyses, we show in this paper that when the traditional model-combining scheme is used, even a single Byzantine node that keeps sending random reports will cause the whole FL model to diverge to non-useful solutions. A low complexity model combining approach is also proposed to stabilize the FL system and make it converge to a suboptimal solution just by controlling the model norm. The Physikalisch-Technische Bundesanstalt extra-large electrocardiogram (PTB-XL ECG) dataset is used to validate the findings of this paper and show the efficiency of the proposed approach in identifying heart anomalies.

The Internet of Things (IoT) is used to interconnect a massive number of heterogeneous resource-constrained smart devices. This makes such networks exposed to various types of malicious attacks. In particular, jamming attacks are among the most common harmful attacks to IoT networks. Therefore, an anti-jamming power allocation (PA) strategy is first proposed in this article for health monitoring IoT networks by exploiting the game theory to minimize the worst case jamming effect under multichannel fading. This strategy uses an iterative algorithm based on gradient descent to identify the Nash Equilibrium (NE) of the game. An artificial neural network (ANN) model is also proposed to accelerate the convergence of the algorithm making it more suitable for IoT networks. Furthermore, novel data population (DP), extension, and balancing techniques are proposed to enhance the efficiency of the proposed strategy in combating jamming attacks even for network configurations that were never used in the training phase. In addition, time and spatial diversities are exploited using a heterogeneous iterative algorithm to enhance the security of the network.

A secure energy efficient approach is proposed to connect Internet of Things (IoT) sensors that operate with limited power resources. This is done by optimizing simultaneously the energy efficiency, the communication rate and the network security while limiting the potential data leakage and tracking the finite battery status evolution. The proposed model uses spatial diversity in addition to artificial jamming introduced by an intermediate device to forward the data from the sensors to the destination and to secure the communication links without draining the rechargeable batteries. The energy harvested by the source is also maximized without affecting the security level of the network. The outage secrecy capacity is derived to evaluate the security level. Furthermore, the system power stability is analyzed using Markov chains and statistical approaches to validate the efficiency of the proposed technique in maintaining the system in a self-sufficient mode and making it operate without the assistance of external power resources.

A jamming combating power allocation strategy is proposed to secure the data communication in IoT networks. The proposed strategy aims to minimize the worst case jamming effect on the intended transmission under multi channel fading and total power constraints by modelling the problem as a Colonel Blotto game Nash Equibrium (NE). Both Logistic Regression as well as a specifically designed algorithm are used to iteratively and rapidly obtain the equilibrium strategy. The conducted theoretical derivations and Monte Carlo simulations confirm that the proposed approach can secure the IoT network with a limited amount of power and with a number of iterations that is much reduced compared to state-of-the-art techniques.

In spite of the wide spread of practical crypto- systems and ciphers nowadays, they still lack a unique metric to measure the secrecy level they provide. Their strength is measured in an ad-hoc way by exposing them to different kinds of attacks. In addition, their ability to hold secure against these attacks is evaluated in time and computations. In this paper, we introduce an approach for calculating the equivocation of the secret key used in these ciphers. In addition, we prove that it can be used as an indicator for the work required to break the cipher. This will help in unifying the metrics used in evaluating the strength of the ciphers and in comparing them with the classical information theoretic secreacy measures.

The open nature of the wireless communication medium renders it vulnerable to jamming attacks by malicious users. To detect their presence and to avoid such attacks, several techniques are present in the literature. Most of these techniques aim to reduce the effect of the jamming signals by increasing the transmission power or by using complex coordination schemes. However, the implementation of such power consuming techniques might be challenging or not feasible in limited resources Internet-of-Things (IoT) devices. Therefore, a defending strategy against jamming attacks in health monitoring IoT networks is proposed in this article. This strategy operates in orthogonal frequency-division multiplexing channels and takes into consideration the effect of slow fading channels in the strategy design. Specifically, the jamming combating problem is formulated as a Colonel Blotto game where the equilibrium defines the minimization of the worst case jamming effect on the IoT sensors communications. Then, the optimal power allocation strategy for all the potential jammer power ranges is derived by investigating the Nash equilibrium of the game. This proposed strategy is shown to be efficient in combating jamming attacks while minimizing the IoT sensors power consumption.

Network security is an critical subject in any distributed network. Recently, machine learning has proven their efficiency for intrusion detection. By using a comprehensive dataset with multiple attack types, a well-trained model can be created to improve the anomaly detection performance. However, high dimensional data sets are a significant challenge for machine learning. In fact, learning algorithms considering all features in the input data, may cause over-fitting to irrelevant aspects of the data and increase the computational time caused by the process of similar features that provide redundant information, which is a critical problem especially for users with constrained resources. In this paper, we propose a new and efficient feature selection technique for intrusion detection in modern networks called Iterative Per Group Feature Selection (IPGFS). IPGFS reduces the number of features in the input data and selects the best features using the performance accuracy of the classifier. The features are sorted and selected according to their accuracy score. Both the UNSW and NSLKDD datasets are used in this paper to validate the proposed model and verify its efficiency in detecting intrusions. The simulation results show that the proposed model can reduce the number of features for the two dataset while successfully detecting intrusions with better accuracy compared to state-of-the-art techniques. Index Cloud security, feature selection, accuracy, machine learning techniques.

The Internet of things (IoT) is one of the most exposed networks to attackers due to its widespread and its heterogeneity. In such networks, jamming attacks are widely used by malicious users to compromise the private and secure communications. Many techniques are proposed in the literature to secure the network from malicious jamming attacks. However, most of these techniques require either the implementation of complex coordination schemes or the use of high transmission power and are therefore challenging to implement in limited resources IoT networks. In this paper, a low complexity anti-jamming defending strategy using smart power allocation under limited power constraints is proposed for health monitoring IoT networks. This strategy is designed by formulating the worst case jamming effect minimization problem as a Colonel Blotto game while considering the slow channel fading effect. By analyzing the Nash Equilibrium (NE) of the game, making use of efficient and fast equilibrium approximation techniques, designing a fast numerical solving approach, training an artificial neural network (ANN) to enhance the accuracy of the estimation, an anti-jamming power allocating strategy is proposed and is shown to be effective in reducing the power consumption and in combating jamming attacks with less resources. A data population scheme is also proposed to make the proposed ANN exploit as much possible the available data to provide accurate NE estimation.

Recently, machine learning techniques are gaining a lot of interest in security applications as they exhibit fast processing with real-time predictions. One of the significant challenges in the implementation of these techniques is the collection of a large amount of training data for each new potential attack category, which is most of the time, unfeasible. However, learning from datasets that contain a small training data of the minority class usually produces a biased classifiers that have a higher predictive accuracy for majority class(es), but poorer predictive accuracy over the minority class. In this paper, we propose a new designed attacks weighting model to alleviate the problem of imbalanced data and enhance the accuracy of minority classes detection. In the proposed system, we combine a supervised machine learning algorithm with the node1past information. The machine learning algorithm is used to generate a classifier that differentiates between the investigated attacks. Then, the system stores these decisions in a database and exploits them for the weighted attacks classification model. Thus, for each attack class, the weight that maximizes the detection of the minority classes will be computed and the final combined decision is generated. In this work, we use the UNSW dataset to train the supervised machine learning model. The simulation results show that the proposed model can effectively detect intrusion attacks and provide better accuracy, detection rates and lower false alarm rates compared to state-of-the art techniques.1In this document we will use the words “node” to represent computing, storage, physical, and virtual machines.

Advances in the design of Full-Duplex (FD) transceivers with low residual Self-Interference (SI) levels has led to their deployment in various wireless communication applications. One promising field that FD transceivers could play a major role in reshaping its dynamics is physical layer security. This paper investigates the performance of FD transceivers in the context of Military FD Radios (MFDR). Particularly, the adopted system model builds on recent investigations regarding the feasibility of deploying an MFDR in a scenario where it simultaneously jams a receiver node while intercepting the signal of a transmitter node, thus simultaneously disrupting and intercepting a tactical communication link of an opponent team. The secrecy performance of the MFDR is theoretically quantified by deriving its outage in interception probability expression. The results reveal that a well-performing SI cancellation scheme can increase the secrecy performance of an MFDR.

With the significant increase of wireless network nodes and traffic load in recent years, especially in the emerging internet-of-things (IoT) and vehicular networks, the design of a fast adaptive relay selection algorithm that is able to cope with a quickly changing environment became a necessity. In particular, the problem of multiple relay selection and beamforming under individual power constraints is investigated in this paper when the amplify-and-forward protocol is used to forward the data to the destination. The proposed algorithm first performs relay selection and beamforming using iterative convex optimization. The selection decisions are stored and processed before being used by a proposed multi-agent machine-learning (ML) model to imitate with high accuracy the optimal selection decision in real time with much less computational complexity. Simulation results confirm that the performance of the proposed technique is very close to the exhaustive search (ES) and to well known algorithms but with an execution time that is thousands of times shorter than traditional techniques.

Many proposed data center architectures are constructed with a huge number of network devices in order to support the increasing cloud based services. These devices are used to achieve the highest performance in case of full utilization of the network. However, the peak capacity of the network is rarely reached. As a result, many devices are set into idle state which increases the network energy consumption and lead to a non-proportionality between the consumed energy and the network load. In this paper, we present a new approach that reduces the data center energy consumption with a reduced trade off on network performance. By exploiting the correlation in time of internode communication and some topological features, the proposed approach uses the outdated traffic matrix to control the set of active communication links and ports in the network (switches ports and nodes ports). The ports activation management process is done using a proposed algorithm that guarantees network connectivity. Extensive simulations have been conducted to validate the performance of the proposed scheme in terms of average path length and energy consumption.

In this paper, the problems of relay selection and distributed beamforming are investigated for bi-directional dual-hop amplify-and-forward frequency-division duplex cooperative wireless networks. When using individual per-relay maximum transmission power constraint, it has been proven that the relay selection and beamforming optimization problem becomes NP hard and requires exhaustive search to find the optimal solution. Therefore, we propose a computationally affordable suboptimal multiple relay selection and beamforming optimization scheme based on the ℓ1norm squared relaxation. The proposed scheme performs the selection for the two transmission directions, simultaneously, while aiming at maximizing the aggregated SNR of the two communicating nodes. Furthermore, by exploiting the previous solutions to accelerate the algorithm's convergence, our proposed algorithm converges to a suboptimal solution compared to the exhaustive search technique with much less complexity.

Underlay cognitive networks should follow strict interference thresholds to operate in parallel with primary networks. This constraint limits their transmission power and eventually the coverage area. Therefore, in this paper, we first design a new approach for asynchronous narrow-band interference (NBI) estimation and mitigation in orthogonal frequency-division multiplexing cognitive radio networks that does not require prior knowledge of the NBI characteristics. Our proposed approach allows the primary user to exploit the sparsity of the secondary users' interference signal to recover it and cancel it based on sparse signal recovery theory. We also propose two subcarrier selection schemes that allow the primary user to further reduce the effect of the secondary users' interference based on sparse signal recovery algorithms. We show that although the primary and secondary transmissions are performed at the same time, the performance of our proposed techniques approach the interference- free limit over practical ranges of NBI power levels.

In this paper, the performance of an orthogonal frequency division multiplexing overlay cognitive radio network with subcarrier selection schemes is investigated. We propose three subcarrier selection techniques that reduce the level of interference at the primary base station based on collected channel state information from the different network nodes. Approximated outage probability expressions are also derived and verified by simulations for the different studied techniques. In addition, we propose and investigate a new approach for asynchronous narrowband interference (NBI) estimation and mitigation in cognitive radio networks. The proposed approach does not require prior knowledge of the NBI characteristics and allows the primary user to exploit the sparsity of the secondary users interference to recover it based on sparse signal recovery theory and approach the interference-free limit over practical ranges of NBI power levels.

In this paper, three novel techniques are proposed and investigated for multiple relay selection in dual hop OFDM networks. These techniques are based on the exploitation of sparse signal recovery theory and on carefully-designed groupings of the subcarriers depending on the channel quality. In particular, the proposed techniques use the Orthogonal Matching Pursuit algorithm which enables them to outperform existing techniques in terms of both outage probability and computation complexity. Furthermore, a detailed performance-complexity tradeoff investigation is presented for the different studied techniques and verified by Monte Carlo simulations.

The scarcity of radio resources coupled with the high data rate demands in the last few years, induced incremental relaying techniques one of the most promising technologies. This is simply because it allows reaping spatial diversity and saving the channel resources at the same time. In this contribution, we make use of the Decode and Forward incremental relaying technique in OFDM cooperative systems by proposing and analyzing two subcarrier mapping algorithms. Based on outage probability, diversity order and power gain analysis, we derive the optimal parameters for these algorithms to enhance the outage performance of DF cooperative system by reducing the effect of error propagation. Also, an incremental relay sharing scheme is proposed allowing multiple sources to share the same relay and results in an improved spectral efficiency, which is critical given the high cost and scarcity of RF spectrum driven by the growing big data.

In this contribution, we investigate the outage performance of OFDM-based Decode and Forward (DF), Amplify and Forward (AF) and Controlled DF (CDF) cooperative systems under IQ Imbalance (IQI). In particular, tractable and compact approximate outage probability expressions are derived and the effect of the different IQI parameters is analyzed for different SNR ranges. Furthermore, by localizing the error floor in terms of IQI and SNR for each technique, we demonstrate when it is more beneficial to invest in increasing the transmission power or in compensating the imbalance. Moreover, we prove that the IQI compensation should be concentrated in the relay for some techniques and in the destination for some others. A comparative study between AF, DF, CDF and direct link transmission techniques is also conducted for different IQI parameters, SNR ranges, transmission rates and relay's position. Hence, this work may create a paradigm for future studies of more effective adaptive IQI compensation techniques that concentrate the compensation on the right IQI, SNR ranges, transceivers depending on the used transmission technique.

Selective OFDMA is a powerfull relaying technique with subcarrier based selection. This scheeme has a high potential to considerably improve the performance of cooperative OFDM systems. However, in practice, the implementation of this technique may suffer from multiple problems caused by the separation of the subcarriers from each other's. In this Contribution, we propose a new selection scheme denoted “Optimized Selective OFDMA” that approaches selective OFDMA performance in terms of outage probability while reducing its synchronization problems that may be caused by the subcarriers separation. The outage probability and diversity order of the proposed scheme are derived. The number of the used paths is also analyzed. We proved that for high SNR, the introduced scheme gives exactly the same performance of selective OFDMA with only one path for all the subcarriers.

In this contribution, we investigate Decode and Forward (DF) OFDM Ad-hoc Routing strategy with bottleneck maximization. We derive the exact outage probability and diversity order for the different scenarios: with and without subcarrier grouping and with and without joint selection, where joint selection refers to the selection of the last two hops together. When joint selection is performed, numerical results show that a power gain can be obtained when using subcarrier grouping technique.