PPCS-MMDML: Integrated Privacy-Based Approach for Big Data Heterogeneous Image Set Classification

Abstract

In the current digital world, the Big Data and Deep learning are the two fast maturing technologies. The classification algorithms from Deep learning provide key and prominent advances in major applications. Deep learning spontaneously learns hierarchical illustrations in deep architectures using supervised and unsupervised methods for classification. The image classification is a bustling research area and applying it in big data will be a great contest. With analysis on big data, it is noticeable that the veracity characteristic unnerving the privacy requirement of data shared. While the data is shared for feature selection process, the privacy is in need for user and databank holders. Also since the feature selection process influences the performance of a classifier, a privacy-based feature selection process is mandatory. In this paper, we propose an integrated technique using PPCS (privacy-preserving cosine similarity) and MMDML (multi-manifold deep metric learning) algorithms for a secure feature selection and efficient classification process on Cancer Image datasets.