Security and Privacy in Shared HitLCPS Using a GA-Based Multiple-Threshold Sanitization Model

In Cyber-Physical Systems (CPS), especially in human-in-the-loop situations (also known as HitLCPS), the security and privacy for keeping sensitive information private is considered an emerging topic in recent decades. Many techniques in privacy-preserving data mining (PPDM) can be applied directly to HitLCPS. However, most of them to date have focused on handling singular threshold problems for data sanitization. If a sensitive itemset includes more items, it has a higher probability of being identified due to its specificity. In this work, we propose a new concept of multiple support thresholds to assist in resolving this issue. The proposed method assigns a stricter threshold for an itemset. Furthermore, a genetic-algorithm (GA)-based model is involved in the designed algorithm to minimize side effects. In our experimental results, the GA-based PPDM approach is compared with traditional Greedy PPDM approaches. The strong experimental results clearly show that our proposed method can give similar performance to conventional algorithms while still maintaining higher-levels of security and privacy protection than previous methods.