Real-Time Deepfake Detection in Online Communication: A Review of AI-Based Audio-Visual Detection Techniques and a Proposed Yii2–MySQL Framework

The rapid advancement of artificial intelligence (AI) and generative models has increased the sophistication of manipulated audio and video content, creating significant challenges for the authenticity of online communication. Real-time deepfake detection has therefore become an important research area in multimedia security and digital forensics. This paper presents a literature-based review of AI techniques for detecting video deepfakes, synthetic speech and multimodal audio-visual manipulation. It examines convolutional neural networks, transformer-based architectures, acoustic feature analysis, temporal inconsistency detection and cross-modal feature fusion. Particular attention is given to detection latency, computational requirements, media compression, robustness and generalization to previously unseen manipulation techniques. Based on the reviewed literature, the paper proposes a conceptual web-based integration framework using the Yii2 PHP framework and MySQL database, supported by an independent machine-learning inference service. The proposed architecture separates multimedia analysis from application management to support detection requests, result storage and monitoring. The study identifies research gaps and discusses future directions for reliable, efficient and privacy-aware deepfake detection in online communication environments.