Continuous Lightweight Authentication and Behavioural Risk Verification for Compromised Sensor Node Detection

Abstract

Wireless Sensor Networks are prone to attacks as nodes can easily get compromised. Traditional authentication protocols only authenticate the node at the time of entry which allows the node to continue acting however it pleases afterward. To address this issue, we present ZeroTrust-WSN: a novel continuous authentication & behavioural risk verification model for Wireless Sensor Networks that allows us to identify malicious sensor nodes with the same low overhead of authentication. Our model consists of lightweight authentication, behavioural checks at intervals, trust calculation, risk scoring, and reactions such as limiting access, re-authentication, rekeying and quarantine. Behaviours analysed include packet forwarding rates, legitimacy of neighbouring nodes during communication, idle listening, energy left, past trust, and proof of deviations. We implemented our model using Python and NS-3 and compared it to three baseline models: Blockchain_ SecAuth, Lightweight Decision Tree IDS and THN-NBA-MND. Results show that ZeroTrust-WSN had an authentication accuracy of 96.8% with 96.2% success rate at detecting compromised nodes with a false positive rate of 3.1% and an average verification delay of 14.6 ms while retaining 95.4% of the network's energy- efficiency when under attack. On all security categories, our model stayed within ~94%-97% and outperformed all baseline models by continuously verifying nodes rather than assigning static trust values. This shows our model's ability to be used as an effective security solution for WSNs as it provides continuous, adaptive, low-overhead security that doesn't tax the limited computations of sensor nodes.