Cloud-based privacy- and integrity-protecting density peaks clustering

Artificial intelligence (AI) has been extensively used in today’s society. The full utilization of AI requires massive amounts of data and strong computing power. Cloud provides such a platform for AI with large storage and powerful computation capabilities. However, the cloud also poses significant security issues since it is not fully trustworthy. Especially, third-party clouds are generally profit-driven, and thus they have strong financial incentives to execute less computation to save costs, thereby not ensuring the integrity of their returned results. In this work, we design a privacy- and integrity-protecting density peaks clustering algorithm in the ciphertext domain (PIDPC). Particularly, we develop efficient strategies guaranteeing the integrity of outsourced DPC results by leveraging game theory. It uses economic approaches to create incentives for the single cloud to compute honestly. Also, we design three optimization techniques to improve the efficiency of PIDPC. Besides, extensive experiments demonstrate that PIDPC guarantees clustering accuracy while ensuring the privacy of training data and the integrity of returned results.