Detection of False Data Injection Attacks in Industrial Wireless Sensor Networks Exploiting Network Numerical Sparsity

Existing studies on false data injection attacks, a type of stealth attacks against sensor networks aimed at compromising the system in the cyber-physical security domain, have primarily been conducted on wired systems for applications such as advanced metering infrastructure in smart grid. However, the emerging trend of the widespread deployment of industrial wireless sensor networks for various new functionalities as well as for replacement of legacy systems, on the other hand, calls for both data aggregation methods that are cost-effective, scalable and easily implementable, as well as feasible approaches to detect injected false data in coordination with such data aggregation models. In this paper, we propose a numerical sparsity-based detection scheme operating upon a network coding-based data aggregation model paired with compressed sensing-based decoding, against attacks that alter the overall network sparsity by compromising and injecting falsified data into multiple sensor nodes in the network. Both the applicative scope and performance of the proposed scheme are analyzed and compared to a more straightforward but realistically challenging approach of directly examining network compressibility, i.e. the number of sufficiently large readings of active nodes extracted from the decoded network signal. Numerical studies illustrate the proposed method is applicable for the usually sparsely active industrial wireless sensor networks, and offers faster, reliable decisions when the aforementioned false data injection attacks are launched.