Model-Free False Data Injection Attack via Eavesdropping: An Online Approach
Nachuan Yang, Xiaoxu Lyu, Ling Shi · 2024
This paper considers the design of false data injection attacks towards cyber-physical systems without knowing system parameters. The existing research usually assumes that the attacker knows the plant's parameters. However, this is often not the real case since the plant's parameters are sensitive information that is not transmitted and thereby cannot be eavesdropped by external attackers. We propose an inference-based CPS attack mechanism where an attacker simultaneously updates an inference system and designs the injection attacks. More specifically, we introduce the notion of system imitator to cyber-security for the first time and propose a bilevel imitation-based cyber-attack mechanism. We show that the attack mechanism can be implemented in an online manner and the convergence is theoretically guaranteed. Besides, we present a new cyber-threat model in communication networks, called network congestion attack, to illustrate the applicability of our approach.