A Comprehensive Survey on Detection of Fake Multimedia Content
Ayush S Acharya, Ashish A Shisal, Saurabh K Butale, Omkar B Latpate, Shalaka Prasad Deore · Cureus Journal of Computer Science. · 2025
The rapid rise of fake news and deepfake media poses serious challenges to digital content integrity and public trust. Despite numerous detection models being developed, there remains a critical gap in creating solutions that are both highly accurate and robust across diverse modalities (text, image, and audio) and in real-time applications. This paper presents a comprehensive survey of state-of-the-art detection techniques for identifying fake multimedia content. We critically evaluate traditional and modern machine learning models, such as convolutional neural network (CNN), long short-term memory (LSTM), and Transformer-based models like bidirectional encoder representations from transformers (BERT) and robustly optimized BERT approach, alongside hybrid approaches that integrate explainable AI and digital watermarking for improved interpretability and detection performance. Our comparative analysis highlights that the Inception-ResNet-v2 model achieves the highest accuracy (99.81%) for deepfake image detection, while CNN + recurrent architectures perform best (96%) for audio deepfakes. In text-based fake news detection, hybrid LSTM-CNN models incorporating explainable AI report up to 99% accuracy on benchmark datasets. These findings illustrate the effectiveness of multimodal approaches but also expose limitations related to adversarial robustness, scalability, and cross-media generalization. The paper concludes by identifying future research directions, including the development of lightweight, real-time systems, cross-modal generalization frameworks, and policy-level interventions to mitigate the societal impact of synthetic media.