Mohammed K. Alzaylaee

The growing number of Internet of Things (IoT) devices requires decentralized Edge Intelligence solutions. As the current FL-based IDS systems are decentralized solutions for privacy protection, face two major problems: (1) network traffic manipulation through adversarial evasion attacks (2) privacy threats from gradient-based inference attacks and (3) server-side Robustness issue. The current methods which use Differential Privacy (DP) or adversarial training result in 5-15% accuracy reduction which makes them unsuitable for deployment. The key novelty of our work is the integration of a novel dual-defense framework that uniquely reconciles the conflict between differential privacy noise and adversarial gradient requirements, effectively eliminating the conventional ”accuracy tax” along with server-side Robust Aggregation. Our research develops an enhanced two-stage federated system which is robust and protects privacy while delivering secure IoT edge intelligence solutions. The core system FedEdge-ID provides 99.73% detection performance across different edge devices. Spectre-Fed enhances the FedEdge-ID framework via three key defenses: (1) Hybrid Loss Adversarial Training (α=0.5) to fortify decision boundaries against evasion, (2) Gradient-Guided Adaptive Privacy with decreasing noise injection (σ0=0.0005, γ =0.95) for secure gradient updates, and (3) Robust Trimmed Mean Aggregation to counter Byzantine poisoning. Experiments demonstrate that Spectre-Fed’s client-side (Layer 1) defense achieves 99.72% clean accuracy with only a 0.01% utility loss versus the non-private baseline. It shows strong adversarial resilience, retaining 99.34% accuracy against FGSM attacks (ϵ =0.01), a mere 0.38% degradation from the clean state. When integrated with server-side Robust Aggregation (Layer 2), the system sustains 99.59% accuracy even under active label-flipping attacks from 20% of clients, while preserving high utility compared to the baseline. The system achieves optimal privacy-utility balance through its formal privacy protection and its ability to resist adversarial attacks which makes it suitable for zero-trust IoT systems.

The cyber-physical network (CPS) combines sensing, communication, and control in physical processes, making them very susceptible to sophisticated cyber-attacks that may cause safety-critical effects. There are two core shortcomings to existing intrusion detection systems (IDS): generative-only models have little transparency of decision-making, while explainable-only models have low robustness in the presence of imbalanced and zero-day attacks. This paper presents a sequentially integrated trustworthy intrusion detection (ID) framework that combines generative learning and explainable AI (XAI) to boost robustness and transparency. The generative module enhances training data diversity, while the explainability module provides post-hoc interpretations during inference. The framework exploits a Generative Adversarial Network (GAN) to tackle the issue of class imbalance and boost generalization to untrained attacks, and SHAP, LIME, and attention-based approaches give instance-level and feature-level explanations. The proposed framework is proven to significantly improve the performance of the state-of-the-art methods through extensive evaluations on industrial CPS, industrial IoT-based CPS, and smart grid-based datasets. In terms of quantitative metrics, it achieves 97.4% accuracy, 96.1% F1-score, 3.1% false alarm rate, 0.983 AUC, and 91.2% zero-day attack detection (compared to 84.7% for GAN-only and 78.6% for XAI-only). The proposed framework provides a qualitative framework that enables the derivation of transparent, human-understandable decisions that are not compromised in terms of detection performance while maintaining an inference latency of 2.6 ms per sample, suitable for real-time CPS monitoring. The results validate the effectiveness of the integration of generative learning and explainable AI as a secure, transparent, and trusted solution to securing modern cyber–physical networks, where current generative-only and explainable-only IDS methods are unable to cover the essential needs.

The proliferation of Internet of Things (IoT) devices has introduced unprecedented security challenges, necessitating efficient intrusion detection systems (IDS) capable of operating under severe resource constraints. This research presents a hardware-informed empirical study of quantized neural-network-based intrusion detection for resource-constrained IoT platforms, using an ARM Cortex-M4 deployment target as a reference. We evaluate FP32, FP16, and INT8 TensorFlow Lite model variants derived from a lightweight 1D-CNN and assess their trade-offs in clean-data accuracy, model size, estimated inference latency, estimated energy consumption, and adversarial robustness. INT8-quantized model achieves 99.10% accuracy on clean data while maintaining 97.50% adversarial accuracy under Projected Gradient Descent (PGD) attacks with perturbation budget = 0.3. The quantized model achieves 12.0× latency reduction (0.083 vs. 0.995 ms) and 92.7% energy reduction (0.0083 vs. 0.1135 mJ) when compared to FP32. The memory footprint of the model is reduced by 55.7% from 58.02 to 25.72 KB. Our comprehensive analysis includes confusion matrices, ROC curves (AUC = 0.9964 for INT8), adversarial robustness heatmaps, and statistical significance testing via McNemar’s test. The results establish INT8 quantization as a viable solution for deploying robust IDS on resource-constrained IoT devices, achieving practical deployment feasibility without reducing detection performance.

Numerous Security experts agree that Intrusion Detection Systems (IDS) are inevitable in securing computer networks, especially against malicious attacks such as Distributed Denial of Service (DDoS). Another limitation of traditional IDSs, which involve both signature-based and anomaly-based models, is that they cannot detect new attacks and often produce many false positives. Signature-based systems are effective only with known attack patterns and are ineffective against unknown ones. In contrast, false positives greatly affect anomaly-based systems due to their high sensitivity levels. In this research, we aim to enhance the accuracy and efficiency of DDoS attack detection by combining Random Forest (RF) and Convolutional Neural Network (CNN)-Long Short-Term Memory (LSTM) models with attention mechanisms, thereby optimizing Intrusion Detection Systems (IDS). The CICIDS2017 dataset is used to train the model, comprising 23,659 benign data points and 17,258 DDoS records. As Random Forest and CNN-LSTM demonstrate their interpretation and temporal feature extraction capabilities, respectively, the hybrid approach combines them based on attention mechanisms over essential features. Moreover, our proposed model outperforms traditional models. The hybrid model’s accuracy, precision, recall, and F1 score are 99.25%, 99.15%, 99.34%, and 99.24%, respectively. Compared to this, our standalone Random Forest model achieves an accuracy of 99.93%, and the CNN-LSTM with attention mechanism scores 99.25%. It can also detect with excellent capabilities, with an AUC of 0.9993. The integration of the attention mechanism significantly enhances the model's effectiveness in real-time intrusion detection. Finally, the future scalability of the model for more complex attack scenarios and real-world deployments is left for further work.

Smart homes that integrate Internet of Things (IoT) devices face increasing cybersecurity risks, posing significant challenges to these environments. The study explores security threats in smart homes ecosystems, categorizing them into vulnerabilities at the network layer, device level, and those from cloud-based and AI-driven systems. Research findings indicate that post-quantum encryption, coupled with AI-driven anomaly detection, is highly effective in enhancing security; however, computational resource demands present significant challenges. Blockchain authentication together with zero-trust structures builds security resilience, although they need changes to existing infrastructure. The specific security strategies show their effectiveness through ANOVA, Chi-square tests, and Monte Carlo simulations yet lack sufficient scalability according to the results. The research demonstrates the requirement for improvement in cryptographic techniques, alongside AI-enhanced threat detection and adaptive security models which must achieve a balance between performance and efficiency and real-time applicability within smart home ecosystems.

Healthcare networks prove to be an urgent issue in terms of intrusion detection due to the critical consequences of cyber threats and the extreme sensitivity of medical information. The proposed Auto-Stack ID in the study is a stacked ensemble of encoder-enhanced auctions that can be used to improve intrusion detection in healthcare networks. The WUSTL-EHMS 2020 dataset trains and evaluates the model, constituting an imbalanced class distribution (87.46% normal traffic and 12.53% intrusion attacks). To address this imbalance, the study balances the effect of training Bias through Stratified K-fold cross-validation (K = 5), so that each class is represented similarly on training and validation splits. Second, the Auto-Stack ID method combines many base classifiers such as TabNet, LightGBM, Gaussian Naive Bayes, Histogram-Based Gradient Boosting (HGB), and Logistic Regression. We apply a two-stage training process based on the first stage, where we have base classifiers that predict out-of-fold (OOF) predictions, which we use as inputs for the second-stage meta-learner XGBoost. The meta-learner learns to refine predictions to capture complicated interactions between base models, thus improving detection accuracy without introducing bias, overfitting, or requiring domain knowledge of the meta-data. In addition, the auto-stack ID model got 98.41% accuracy and 93.45% F1 score, better than individual classifiers. It can identify intrusions due to its 90.55% recall and 96.53% precision with minimal false positives. These findings identify its suitability in ensuring healthcare networks’ security through ensemble learning. Ongoing efforts will be deployed in real time to improve response to evolving threats.

The growing incidence of cyberattacks necessitates a robust and effective Intrusion Detection Systems (IDS) for enhanced network security. While conventional IDSs can be unsuitable for detecting different and emerging attacks... | Find, read and cite all the research you need on Tech Science Press