Privacy Preserving Ant Colony Optimization based Neural Learning Classifier

摘要

Due to the advancement of technologies and usage of smart devices generate huge volumes of personal data which are more sensitive and sharing of these data is important for decision making in the competitive business world. But the knowledge discovery process creates privacy and legal issues. Therefore, privacy preserving is necessary to hide sensitive data in the knowledge discovery process. In this paper, a privacy preserving ant colony optimization based neural learning classifier algorithm is proposed, which allows to learn knowledge from the neural network without revealing the sensitive data to other parties. The objective of privacy preserving data classification is to build classifiers that non-reveal the sensitive information in the data being classified. In this paper, Backpropagation algorithm is used for classification and optimizing the weights of neural network is done by ant colony optimization. The activation function in the neural network is securely computed using ElGamal scheme and the data is vertically partitioned. Experiments were conducted to observe the effectiveness of the classifier on real world datasets downloaded from the University of California, Irvine (UCI) machine learning repository and it is evident that the proposed privacy preserving ant colony optimization based neural learning classifier is promising from the Receiver Operating Characteristics analysis.