Machine Learning Based Detection System for Identifying Deepfake Images and Social Bots

This paper presents a machine learning-based approach to detect two major forms of misinformation on social media platforms: deepfake images and social bots. For deepfake image detection, we propose a novel hybrid model combining DenseNet121 with custom Convolutional Neural Networks, achieving 97% accuracy on a dataset of 140,000 images. The social bot detection component utilizes a dual-branch architecture, processing tweet content through LSTM networks and user metadata through dense layers, attaining 91 % accuracy on the CIC truth seeker dataset. Both models demonstrate balanced performance across classes, addressing the challenges of evolving misinformation tactics. By integrating these detection systems, we aim to enhance the integrity and reliability of information ecosystems in digital spaces. Our approach contributes to the ongoing efforts to combat the spread of misinformation on social media platforms.