Vireshwar Kumar (Assistant Professor, IITD) (PI), Rajeev Shorey (Adjunct Faculty, IITD) (co-PI), Dan Kim (Professor, University of Queensland) (co-PI)
Sponsored by: Ministry of Electronics and Information Technology (MeitY)
To facilitate smart applications such as smart vehicles and smart manufacturing, energy-constrained devices are inter-connected through bandwidth-constrainedcommunication protocols to form the internet-of-things (IoT). Due to such constraints, an IoT network fails to employ conventional security protocols, which makes them vulnerable to security threats. One resource-efficient way to secure the IoT network is to deploy a comprehensively trained machine learning (ML)-based intrusion detectionsystem (IDS). However, the existing IDSs are vulnerable to adversarial machine learning (AML)-assisted attacks that occur during the inference phase of the ML model, i.e. when the IDS employs the ML model for analysing specific network traffic to detect the presence of a malicious attack. In an AML-assisted attack, the attacker makes carefully crafted perturbations in the network packets with the goal to evade detection by the IDS. This project aims to resolve this issue by developing a novel algorithm for an AML-resilient IDS for detecting AML-assisted attacks in IoT networks. It involves demonstration of the proof of concept and evaluation of the effectiveness of the proposed IDS by conducting extensive experiments on real-world IoT devices.
