Toward Fault-Tolerant and Secure Frequent Itemset Mining Outsourcing in Hybrid Cloud Environment

Abstract Due to the rising costs of maintaining IT infrastructures for large-scale data mining, it is becoming a trend for data owners to outsource data mining tasks together with storage to cloud service providers, however, which also arouses security concerns on unauthorized breaches of data confidentiality and result integrity. Existing solutions yet seldom protect data privacy whilst guaranteeing result integrity. To address these issues, this paper proposes a series of privacy-preserving building blocks by employing Shamir’s secret sharing scheme. Based on those subprotocols, an efficient frequent itemset mining protocol is designed under hybrid cloud setting, in which the public unreliable cloud and semi-trusted cloud cooperate to mine frequent patterns over the encrypted database. Our scheme not only protects the privacy of datasets from frequency analysis attack, but also verifies the integrity of mining results. Theoretical analysis demonstrates that the scheme ensures security as well as fault tolerance under our threat model. Experimental evaluations show that our proposed protocol outperforms the similar solution regarding efficiency while it can detect and correct cloud servers’ errors effectively.