Detection and Prediction of FDI Attacks in IoT Systems via Hidden Markov Model

摘要

False data injection (FDI) attacks aim to threaten the security of Internet of Things (IoT) systems by falsifying a device's measurements without being detected. In this paper, we propose a process for detecting and predicting FDI attacks, which aims to predict future attacks before they occur and induce IoT devices to behave reliably. First, we propose a novel artificial intelligence (AI)-based detection and prediction module that uses a hidden Markov model (HMM) to observe the behavior of IoT devices and predict their future actions. Next, we design a distributed trust management module that establishes trust between devices using a set of weighted votes. To defend against FDI attacks in communication channels, we formulate a bandwidth optimization problem to meticulously allocate bandwidth to trusted devices. In addition, we propose an efficient incentive mechanism that uses reputation rewards to encourage trustworthy behavior and uses a punishment mechanism to neutralize malicious behavior. Simulations show that the proposed process outperforms recent benchmark FDI attack detection algorithms in the literature in terms of significantly improving attack detection accuracy and reducing attack detection latency.