Detecting and mitigating poisoning attacks in Federated Learning (FL) poses a significant challenge, as malicious clients can severely impair model performance through adversarial updates. In this work, we present Dynamic Ensemble Clustering for Poison-Aware Federated Learning (DECPA-FL), a novel defense framework that operates at both the client and sample levels. Our approach leverages three dynamic model clusters—normal, poison, and hybrid—each designed to address varying data properties. Unlike conventional FL methods that treat all client input identically, DECPA-FL utilizes an adaptive Isolation Forest mechanism that progresses via federated rounds to identify poisoned samples with enhanced accuracy. The system’s Poisoning-Aware Loss Function provides reduced weights to potentially contaminated data, while its adaptive learning rate mechanism adjusts training parameters based on identified poison ratios. We evaluate DECPA-FL on the CICIDS 2017 network intrusion dataset and achieve an F1-score of 96.06%, outperforming centralized baselines by 8.25% and traditional FL by 18.83%. Our method maintains robust performance even under 15% poisoning rates, where existing approaches suffer substantial degradation. Through DECPA-FL, we offer an effective and resilient defense for secure federated learning in adversarial and privacy-critical domains.

Federated learning (FL) is a decentralized paradigm that enables collaborative model training across multiple institutions—such as hospitals—without sharing sensitive medical data (e.g., MRI or CT scans). The most common algorithm, Federated Averaging (FedAvg), aggregates locally trained model weights to form a global model. However, FedAvg struggles under statistical heterogeneity (non-independent and identically distributed data from diverse scanners, imaging protocols, and patient demographics) and system heterogeneity (varying client computation and memory capacities), often leading to unstable optimization and slow convergence.To address these challenges, we propose CLS-Transfer Federated Learning (CLS-TFL), a framework that improves robustness and efficiency in heterogeneous medical FL. Our approach leverages a pretrained Vision Transformer (ViT) while adapting only the [CLS] (classification) token head—and optionally the final transformer block—keeping the main encoder frozen. This design reduces trainable parameters and stabilizes non-IID training by focusing adaptation on the [CLS] representation that summarizes global image semantics.We further introduce a resource-aware orchestrator that estimates client training speeds during a warmup phase and dynamically allocates local epochs to satisfy a time budget while ensuring adequate sample coverage for rare data types. Experiments on a non-IID MRI classification dataset with client-specific perturbations (blur, noise, brightness, resolution) demonstrate that FedAvg achieves 73.56 % test accuracy, whereas CLS-TFL reaches 89.49 %. Incorporating resource-aware allocation further improves accuracy to 92.29 % while reducing average round time from 66.10 to 62.40 seconds.Unlike existing optimizer- or client-selection-based approaches, CLS-TFL unifies ViT [CLS]-based adaptation with time-budgeted orchestration, improving both accuracy and time-to-accuracy under real-world heterogeneity. The framework is modular, easily deployable, and compatible with personalization and fairness extensions in medical federated learning.

Privacy-preserving and secure data sharing are critical for medical image analysis while maintaining accuracy and minimizing computational overhead are also crucial. Applying existing deep neural networks (DNNs) to encrypted medical data is not always easy and often compromises performance and security. To address these limitations, this research introduces a secure framework consisting of a learnable encryption method based on block-pixel operation to encrypt the data and subsequently integrate it with the Vision Transformer (ViT). The proposed framework ensures data privacy and security by creating unique scrambling patterns per key, providing robust performance against leading bit attacks and minimum difference attacks.

As blockchain technology gains traction for enhancing data security and operational efficiency, traditional centralized authentication systems remain a significant bottleneck. This paper addresses the challenge of integrating decentralized authentication and access control within distributed networks. We propose a novel solution named ChainGuard, a fully decentralized authentication and access control mechanism based on smart contracts. ChainGuard eliminates the need for a central server by leveraging blockchain technology to manage user roles and permissions dynamically. Our scheme supports user interactions across multiple organizations simultaneously, enhancing security, efficiency, and transparency. By addressing key challenges such as scalability, security, and transparency, ChainGuard not only bridges the gap between traditional centralized systems and blockchain's decentralized ethos but also enhances data protection and operational efficiency.

With UAVs on the rise, accurate detection and identification are crucial. Traditional unmanned aerial vehicle (UAV) identification systems involve opaque decision-making, restricting their usability. This research introduces an RF-based Deep Learning (DL) framework for drone recognition and identification. We use cutting-edge eXplainable Artificial Intelligence (XAI) tools, SHapley Additive Explanations (SHAP), and Local Interpretable Model-agnostic Explanations(LIME). Our deep learning model uses these methods for accurate, transparent, and interpretable airspace security. With 84.59% accuracy, our deep-learning algorithms detect drone signals from RF noise. Most crucially, SHAP and LIME improve UAV detection. Detailed explanations show the model's identification decision-making process. This transparency and interpretability set our system apart. The accurate, transparent, and user-trustworthy model improves airspace security.

Healthcare industries face challenges when experiencing rare diseases due to limited samples. Artificial Intelligence (AI) communities overcome this situation to create synthetic data which is an ethical and privacy issue in the medical domain. This research introduces the CAT-U-Net framework as a new approach to overcome these limitations, which enhances feature extraction from medical images without the need for large datasets. The proposed framework adds an extra concatenation layer with downsampling parts, thereby improving its ability to learn from limited data while maintaining patient privacy. To validate, the proposed framework’s robustness, different medical conditioning datasets were utilized including COVID-19, brain tumors, and wrist fractures. The framework achieved nearly 98% reconstruction accuracy, with a Dice coefficient close to 0.946. The proposed CAT-U-Net has the potential to make a big difference in medical image diagnostics in settings with limited data.

Conflicting transactions within blockchain networks not only pose performance challenges but also introduce security vulnerabilities, potentially facilitating malicious attacks. In this paper, we explore the impact of conflicting transactions on blockchain attack vectors. Through modeling and simulation, we delve into the dynamics of four pivotal attacks - block withholding, double spending, balance, and distributed denial of service (DDoS), all orchestrated using conflicting transactions. Our analysis not only focuses on the mechanisms through which these attacks exploit transaction conflicts but also underscores their potential impact on the integrity and reliability of blockchain networks. Additionally, we propose a set of countermeasures for mitigating these attacks. Through implementation and evaluation, we show their effectiveness in lowering attack rates and enhancing overall network performance seamlessly, without introducing additional overhead. Our findings emphasize the critical importance of actively managing conflicting transactions to reinforce blockchain security and performance.

Proof of Stake (PoS) blockchains offer promising alternatives to traditional Proof of Work (PoW) systems, providing scalability and energy efficiency. However, blockchains operate in a decentralized manner and the network is composed of diverse users. This openness creates the potential for malicious nodes to disrupt the network in various ways. Therefore, it is crucial to embed a mechanism within the blockchain network to constantly monitor, identify, and eliminate these malicious nodes without involving any central authority. In this paper, we propose MRL-PoS+, a novel consensus algorithm to enhance the security of PoS blockchains by leveraging Multi-agent Reinforcement Learning (MRL) techniques. Our proposed consensus algorithm introduces a penalty-reward scheme for detecting and eliminating malicious nodes. This approach involves the detection of behaviors that can lead to potential attacks in a blockchain network and hence penalizes the malicious nodes, restricting them from performing certain actions. Our developed Proof of Concept demonstrates effectiveness in eliminating malicious nodes for six types of major attacks. Experimental results demonstrate that MRL-PoS+ significantly improves the attack resilience of PoS blockchains compared to the traditional schemes without incurring additional computation overhead.

An efficient, scalable, and provably secure dynamic auditing scheme is highly desirable in the cloud storage environment for verifying the integrity of the outsourced data. Most of the existing work on remote integrity checking focuses on static archival data and therefore cannot be applied to cases where dynamic data updates are more common. Additionally, existing auditing schemes suffer from performance bottlenecks and scalability issues. To address these issues, in this paper, we present a novel dynamic auditing scheme for centralized cloud environments leveraging an enhanced version of the B-tree. Our proposed scheme achieves the immutable characteristic of a decentralized system (i.e., blockchain technology) while effectively addressing the synchronization and performance challenges of such systems. Unlike other static auditing schemes, our scheme supports dynamic insert, update, and delete operations. Also, by leveraging an enhanced B-tree, our scheme maintains a balanced tree after any alteration to a certain file, improving performance significantly. Experimental results show that our scheme outperforms both traditional Merkle Hash Tree-based centralized auditing and decentralized blockchain-based auditing schemes in terms of block modifications (e.g., insert, delete, update), block retrieval, and data verification time.

In this paper, we consider a cell-free massive multiple input multiple output (CF-MMIMO) communication system, where users are supported by access points (AP) in conjunction with intelligent reflecting surfaces (IRS) mounted on unmanned aerial vehicles (UAVs). Although aerial IRS (aIRS) offers agile support for expanding network coverage in CF communication systems, the effective operation of such a complex network necessitates a channel state information (CSI) acquisition scheme that exhibits low run-time computational complexity. We propose an artificial intelligence (AI)-based approach to design and develop an efficient channel prediction scheme for CSI acquisition in the CF-MMIMO network supported by aIRS considered. Simulation results demonstrate the effectiveness of the proposed scheme in predicting channel gains across a wide range of signal-to-noise ratios (SNR) while maintaining low computational complexity during real-time operations.

Although wide-scale integration of cloud services with myriad applications increases quality of services (QoS) for enterprise users, verifying the existence and manipulation of stored cloud information remains an open research problem. Decentralized blockchain-based solutions are becoming more appealing for cloud auditing environments because of the immutable nature of blockchain. However, the decentralized structure of blockchain results in considerable synchronization and communication overhead, which increases maintenance costs for cloud service providers (CSP). This paper proposes a Merkle Hash Tree based architecture named Entangled Merkle Forest to support version control and dynamic auditing of information in centralized cloud environments. We utilized a semi-trusted third-party auditor to conduct the auditing tasks with minimal privacy-preserving file-metadata. To the best of our knowledge, we are the first to design a node sharing Merkle Forest to offer a cost-effective auditing framework for centralized cloud infrastructures while achieving the immutable feature of blockchain, mitigating the synchronization and performance challenges of the decentralized architectures. Our proposed scheme outperforms it's equivalent Blockchain-based schemes by ensuring time and storage efficiency with minimum overhead as evidenced by performance analysis.

Pilot contamination poses a critical challenge for channel estimation in dense cell-free (CF) distributed multiple-input multiple-output (CF-DMIMO) wireless networks. State-of-the-art channel estimation schemes require inversion of a high-dimensional channel covariance matrix, which is practically infeasible for dense CF-DMIMO networks owing to the requirement of large storage and high dimensional computational complexity. In this work, we investigate channel estimation problem for a CF-DMIMO network, where both terrestrial and aerial users are jointly supported by distributed access points. We formulate the problem of estimating channel coefficients from the received in-phase/quadrature (I/Q) samples as a non-linear regression problem and propose two deep-learning aided channel estimation schemes for the considered network, namely, deep model-agnostic neural network (DMANN) and deep successive contamination cancellation (DSCC) schemes. Compared to the state-of-the-art channel estimation schemes for CF-DMIMO networks, the proposed schemes (i) tackle the unavoidable pilot contamination issue in dense CF-DMIMO networks while estimating the channel gains for both terrestrial and aerial users; (2) does not require prior knowledge of signal-to-noise ratios; and (3) works well in the presence of non-Gaussian correlated noise. Simulation results demonstrate the effectiveness of the proposed schemes over state-of-the-art channel estimation schemes in various use cases of the CF-DMIMO networks.

Advanced Persistent Threat (APT) has dramatically changed the landscape of cybersecurity. APT is carried out by stealthy, continuous, sophisticated, and well-funded attack processes for long-term malicious gain thwarting most current defense mechanisms. There is a need for a defense strategy that continuously combats APT over a long time-span in imper-fect/incomplete information on attacker's actions. We propose the stochastic evolutionary game model to simulate the dynamic adversary to address this need in this work. We add the player's rationality parameter c to the Logit Quantal Response Dynamics (LQ RD) model to quantify the cognitive differences of real-world players. We propose an optimal decision-making plan by calculating the stable evolutionary equilibrium that balances a trade-off between defense cost and benefit. Cases studies conducted on Energy Delivery Systems (EDS) indicate that the proposed method can help the defender predict possible attack action, select the related optimal cyber defense remediation over time, and gain the maximum defense payoff.

Prioritized cyber defense remediation plan is critical for effective risk management in Energy Delivery System (EDS). Due to the complexity of EDS in terms of heterogeneous nature blending Information Technology (IT) and Operation Technology (OT) and Industrial Control System (ICS), scale and critical processes tasks, prioritized remediations should be applied gradually to protect critical assets. In this work, we propose a methodology for prioritized cyber risk remediation plan by detecting and evaluating paths to critical nodes in EDS. We propose critical nodes characteristics evaluation based on nodes' architectural positions, measure of centrality based on nodes' connectivity and frequency of network traffic, as well as the controlled amount of electrical power. The paper also examines the relationship between cost models of budget allocation for removal of vulnerabilities on critical nodes and its impact on gradual readiness. The proposed cost models were empirically validated in an actual network ICS test-bed computing nodes criticality. Two cost models were examined, and although varied, we concluded the lack of correlation between types of cost models to most damageable attack path and critical nodes readiness.

The ubiquitous adoption of cellular technologies, such as, Long Term Evolution (LTE) and Wide-band Code Division Multiple Access (WCDMA) has not diminished the impact of Global System for Mobile communication (GSM) technologies. Despite 20 years of deployment, GSM still poses significant security threats to its users due to adversaries exploiting protocol vulnerabilities. In this paper, we present an attack which leverages cross-layer information from network or data link layers to craft an attack vector targeting the physical layer. The cross-layer attack provides attacker sufficient knowledge to specifically target cells, control channels, and mobile stations (MS) with minimal investment of communication and energy resources. We have designed and implemented an experimental testbed, which comprises of, software defined radios (SDR) (USRP and gnuradio), open source GSM channel sniffer (gr-gsm), and distributed processing engine (Apache Spark). The experimental testbed will also facilitate cloning a base transceiver station (BTS) which will benefit from availability of cross-layer information to create customized attack vectors. The cross-layer attack capability on our experimental testbed provides a cost effective scheme to achieve desired benefits with optimal usage of communication and energy resources. Experimental results shows that the testbed can be used to successfully attack multiple mobile stations with minimal usage of power resources.