Abstract

Outlier detection is the technique of finding outliers or anomalies in a dataset. Outlier detection is an important data mining task to detect outliers because of its several application fields such as intrusion detection, credit card fraud, fraudulent financial transactions, environmental science, and many other machine learning applications. In this study, we consider a problem that an organization with less computational capability wants to detect the outliers from its categorical dataset using efficient data mining algorithms through outsourced computation. To address the above problems, we introduce two post quantum secure protocols for detecting outliers from a dataset using attribute value frequency (AVF) and weighted attribute value frequency (WAVF) algorithms through outsourced computation. To ensure the security of the protocols, we use the modified technique of the CKK scheme (IEEE Trans. Inf. Forensics Security, 2016) with the BFV fully homomorphic encryption scheme since BFV’s implementation works well for small plaintext moduli. In the end,

Keywords