Towards a Privacy-Aware and Outsourced Fake News Detection Framework

With the ubiquitous nature of Online Social Network (OSN) platforms, such as Facebook and X, the spread of fake news on these platforms has been increasing at an alarming rate. Users are able to re-post conspiracy theories and also manipulate mainstream media news articles in order to perpetuate their narratives. Therefore, it is important for OSN providers to implement effective methods to detect user posts that have an elevated likelihood of being fake news. In this paper, we propose a privacy-aware and outsourced framework for fake news detection. Our framework provides a collaborative outsourced environment where multiple OSN providers can train the fake news detection model based on their aggregated datasets in a privacy-preserving manner. We utilize a cryptographic hash and investigate three popular text vectorization techniques under the Random Forest classifier. Our experimental results show the accuracy and efficiency of the proposed framework using a real-world dataset.