An Energy-Adaptive Event-Driven Hybrid Trust Framework for Detection of Man-In-The-Middle Attacks in Wireless Sensor Networks

Abstract

In wireless sensor networks (WSNs) and edge computing architectures for cyber-physical systems, Man-in-the-Middle (MIM) routing attacks present a severe security vulnerability. While continuous per-packet cryptographic encryption guarantees payload integrity, it exhausts node battery resources due to heavy computational overhead. Conversely, passive intrusion detection systems conserve energy but struggle to isolate behaviorally adaptive adversaries. To address this fundamental security-efficiency tradeoff, this paper introduces an autonomous dual-engine active-passive framework with adversarial machine learning hardening. The proposed protocol combines stochastic $1 / K$ sampling for truncated HMAC-SHA256 authentication with a lightweight Random Forest classifier trained on side-channel latency and inter-arrival jitter telemetry. A localized edge self-healing state machine at Cluster Head nodes tracks dynamic link trust scores to isolate compromised forwarding links and establish authenticated bypass routes within seconds. Discrete-event simulations across topologies scaling up to 500 nodes demonstrate that the framework achieves an intrusion detection accuracy of 98.4% under standard attacks and 96.2% under evasive jittermatching attacks, providing a 24.9 percentage point advantage over baseline intrusion detection systems. Furthermore, the framework reduces total verification energy consumption by up to 77.8% compared to continuous cryptographic routing, achieving rapid detour convergence in 3.20 s.