Security Model against private data sharing by Streaming (OTT) platforms using Generative Adversarial Networks

Abstract

The expansion of television streaming services has transformed media consumption, providing unparalleled convenience and access to content. Streamers frequently gather comprehensive user data, encompassing viewing patterns, individual preferences, and financial details. This data can be commercialised via collaborations with external advertisers and data brokers, thereby engendering considerable privacy violations, identity theft, and user confidence deterioration. Generative Adversarial Networks (GANs) present a promising method for improving detection techniques of data transmitted to third parties. GANs can be trained to replicate standard data flow patterns and detect anomalies that suggest unauthorised data sharing. Additionally, GANs can produce synthetic data that simulate authentic user behaviour, thereby aiding in developing resilient real-time detection models. Moreover, GANs can create sophisticated data anonymisation techniques to monitor whether user data has been shared. This paper introduces an innovative, multifaceted security and privacy model utilising GANs and deep learning methodologies to identify and alleviate these threats. It compares the GAN model against traditional Support Vector Machine and Random Forests Classifier models. Our methodology integrates anomaly detection and graphs convolutional networks with generative adversarial networks to detect dubious data-sharing activities. The proposed model illustrates the efficacy of deep learning models, specifically GANs, in identifying unauthorised data-sharing platforms.