Least Square Privacy Preserving Technique for Intrusion Detection System

Abstract

Network intrusion is a foremost growing concern threat in the cyberspace, which can be damage the network architecture in a multiple ways by modifying the system configuration/parameters. Hackers/Intruders are familiar with signature based intrusion detection models and they are making successful attempts to crash the networks. Hence, it is necessary to preserve user privacy on intrusion data. PPDM techniques forms a necessary but existing techniques such as Encryption, Perturbation, Data Transformation, Normalization, L-Diversity, K-Anonymity methods forms excessive generalization and suppression problems. In this paper, LSPPM distortion technique using Least Square Method with ensemble classification model have been proposed for providing efficient privacy preservation on intrusion data. The proposed methodology is validated on benchmark NSL_KDD intrusion dataset. A comparative analysis of NSL_KDD class attributes is performed for better classification in terms of accuracy, FAR, F-Score and time taken to build LSPPM-NIDS. The experimental results of state-of-art PPDM methods are also analyzed before and after distortion, and privacy measures to ascertain the degree of privacy offered.