Active Learning-Based Backtracking Attack Against Source Location Privacy of Cyber-Physical System

A Cyber-Physical System (CPS) is a networked system with communication protocols and technologies that are susceptible to potential vulnerabilities and security threats. For example, an adversary may try to uncover the Source Location Privacy (SLP) of a target to perform an attack by analyzing network traffic in a random manner. However, this approach is usually inefficient. Considering the inefficiency of traditional random walk methods, we incorporate an active learning mechanism to equip with the backtracking framework to effectively distinguish between the fake and real network traffic with a limited number of learning samples. In this paper, we present an Active Learning-based Backtracking Attack (ALBA) which leverages muti-layer sampling technique to discover SLP. ALBA offers a more intelligent and faster approach in discovering the SLP. A new muti-layer sampling strategy is also proposed to enhance the performance of the backtracking attack. The effectiveness of our proposed ALBA is evaluated using the NSL-KDD dataset, which simulates the communication environment of CPS. The experimental results show that ALBA outperforms existing state-of-the-art schemes.