Data Inference from Publicly Available Data: Threats and Defense Methods in Power Systems

In recent years, data disclosure has become a global trend. While making public data of power systems available facilitated research and development efforts, it also brought more risks to the security of power systems. Although the public data may not directly reveal sensitive information, attackers can use the public data to reversely infer sensitive data. Recent studies have proven the existence of data inference threats. However, it is unclear what the actual maximum inference threat is, so targeted defense can not be easily developed. In this paper, we first establish a model to evaluate the availability of different electricity prices on data inference threats. Then, we derive the maximum inference threat and show how to reach it under different network topologies. To defend against sensitive data inference threats, we propose a data disclosure strategy based on differential privacy technologies. Experimental results show that the inference method can reach the theoretical maximum value and the defense method can balance the availability of public data and the privacy of sensitive data in all test systems.