Energy-Efficient Distributed Intrusion Detection Leveraging Graph-Smoothness and Sparsity in Consumer IoT WSNs

An effective security approach is essential for wireless sensor networks (WSNs) against various security threats, which helps to estimate the lifetime of the network. This makes it highly valuable to design an Intrusion Detection System (IDS) that consumes less energy and identifies malicious activity precisely. Traditional IDS techniques are limited to detecting only certain types of attacks and often suffer from reduced network longevity. Intrusion detection plays a critical role in consumer electronics by safeguarding user privacy, ensuring reliable device operation, and protecting interconnected smart environments from increasingly sophisticated cyber threats. This paper proposed a Distributed ADMM-based IDS (D-ADMM-IDS) using graph-smoothness priors and sparsity-driven anomaly decomposition for real-time threat detection. The proposed framework optimizes locally at each sensor, exchanging only compressed dual variables, thereby reducing communication by 11.1% compared to existing schemes. The system adopted a convex composite objective, ensuring convergence within 25 ADMM iterations, with residuals below 2.3×10−3and 2.0×10−3. The simulations show that D-ADMM-IDS achieves a detection accuracy of 96.4% while maintaining a false positive rate of only 1.9%. Additionally, the proposed method reduces average detection latency 3.0 s and energy consumption 0.875 J per node.