Fog Computing-Based Intrusion Detection Architecture to Protect IoT Networks

摘要

With the deployment of billions of Internet of Things (IoT) devices, more and more cyber attacks involving or even targeting such devices are rife. Cyberattack vectors are in constant evolvement in terms of diversity and complexity. Thus, to detect novel cyberattacks, we use anomaly-based techniques which model the expected behavior of the IoT device to identify occurrences of attacks. In this paper, we propose a new distributed and lightweight intrusion detection system (IDS). To provide efficient and accurate intrusion detection, the proposed IDS combines variational AutoEncoder and multilayer perceptron. The IDS operates within a two-layered fog architecture, an anomaly detector within fog node, and attack identification module within the cloud. The proposed approach is evaluated on two recent cyber attack datasets. The experimental results showed that the proposed system is able to characterize accurately the normal behavior within fog nodes, and detect different attack types such as DDoS attacks with high detection rate (99.98%) and low false alarms rate (less than 0.01%). The proposed system outperforms other existing techniques in terms of detection and false positive rates.