Integrity attacks on state estimation with varying data access: From visibility to full restriction

Abstract

This paper explores optimal integrity attacks with varying data access to innovations transmitted from smart sensors to remote state estimators. The data access refers to attackers’ capability or authorization to eavesdrop on and tamper with raw data packets. Unlike prior studies where the innovations were entirely intercepted and modified by attackers, different restrictions on data access are considered in this work. Within this context, entries of each innovation are separated into two parts: safe data and suspicious data. The first level of data access enables the interception of complete data, while modifications are limited to the suspicious data only. The second level restricts eavesdropping or tampering with the safe data, permitting only the interception and modification of suspicious data. To maintain stealthiness, the integrity attacks must adhere to stringent constraints, ensuring the statistical characteristics of the entire innovations remain unaltered. In this varying setting, optimal attack policies that make full utilization of available information are proposed and the corresponding coefficients are derived explicitly. Moreover, the optimality of proposed strategies is demonstrated rigorously, showing that no existing attack outperforms the proposed ones. Finally, the effectiveness of theoretical results is demonstrated through numerical examples and comparative studies.