Intelligent trust-based routing and attack detection mechanism using hybrid computing techniques

With the expansion of IoT and WSN, the arsenal of attacks that can be directed against these networks blackhole, sinkhole, Sybil, selective forwarding all serve to lower the reliability one might reasonably expect from them. Traditional trust-based approaches and machine-learning (ML) alone cannot provide a high level of security for our IoT/WSNs, and they typically have little predictive power. Besides, they don’t take into account situational behavioural information. In this paper, we offer an intelligent hybrid security architecture that combines multi-parameter trust Calculation with edge-based ML attack detection to increase the reliability of secure packet delivery in distributed IoT/WSN environments. We use node behaviour metrics packet forwarding ratio, residual energy, link quality, and routing anomalies to calculate trust scores, while an edge-deployed lightweight ML classifier offers real-time anomaly identification with sub-2 ms inference latency. A hybrid decision engine combines trust and output from ML algorithms to dynamically separate out malicious nodes and optimize routing paths. Experimental evaluation using simulation-generated datasets and attack scenarios demonstrates that the proposed model can enhance the packet delivery rate by 30%, lower malicious packet propagation by 45% on average, and uphold a detection accuracy of 97-98% along with low false positive percentages. These results confirm the framework’s validity, versatility, and practicality for securing IoT/WSN installations that often have restricted resources in facing continuously altering threats.