Smart home cyberattack detection framework for sponsor incentive attacks

Abstract

Sponsor incentive pricing is an emerging pricing scheme in smart home energy systems. It allows a supplier of smart appliances to co-pay the electricity bills of customers due to using their products, which boosts the sales of smart appliances in the sponsor program. More importantly, it facilitates to shift energy usage of incentive customers from peak to non-peak hours by designing the time-varying rate, which is inversely proportional to the total demand. Despite its effectiveness in bill reduction and load balancing, the sponsor incentive pricing scheme is vulnerable to various forms of cyberattacks. In this paper, we develop the first detection technique for cyberattacks that exploits the sponsor incentive pricing scheme. Our short-term detection algorithm uses binary logistic regression for learning the energy usage patterns and identifying anomaly ones. Leveraging the partially observable Markov decision process framework, our long-term detection algorithm optimizes the utility's decision of on-site inspections while minimizing the labor cost and the financial loss due to cyberattacks. The simulation results demonstrate that, for detection of rebate rate cyberattacks, the proposed algorithm reduces the peak-to-average ratio of the aggregate energy usage profile by 26.57% and 9.58%, compared with the no-detection scenario and a natural heuristic detection technique, respectively. For detection of sponsor incentive abuse, the maliciously elevated electricity bill is reduced by 46.03% and 15.30%, respectively.