Bengaluru, September 06, 2026 – As manipulated videos and synthetic media become increasingly difficult to distinguish from authentic content, students from the Department of Computer Science & Engineering at East Point College of Engineering & Technology (EPCET), Bengaluru, have developed a web-based AI-powered deepfake video detection system designed to help identify whether a video has been manipulated.
Developed by Adarsh R, Chandu Kumar G, K N Siddarth, and Likith Roshan H B under the guidance of Asst. Prof. Nithyananda C R, the project uses a hybrid AI approach combining a ResNeXt50 convolutional neural network (CNN) with an LSTM model to analyse facial patterns across multiple frames of a video.
The prototype is designed as a web application where users can upload a video and receive a detection result. According to the project brief, the system was developed with an interactive algorithm visualiser and was designed to extract meaningful features from video frames to distinguish manipulated content from authentic footage.
The project team reports that the system successfully identifies deepfake videos using the hybrid CNN-LSTM approach and provides a visualisation of the detection process. The work also highlights the growing role of AI in strengthening digital literacy and helping users assess the authenticity of increasingly sophisticated video content.
Speaking about the project, Rajiv Gowda, CEO, East Point Group of Institutions, said: “Deepfakes are no longer only a technology story; they are a trust story. What makes this project relevant is that our students are applying AI to a problem that ordinary users, businesses and institutions increasingly face - knowing whether what they see is real. Student innovation becomes meaningful when it addresses challenges that are already shaping society.”
The project brief identifies future enhancements including faster processing, support for additional video formats, improved accuracy, and support for videos containing multiple faces. The current work is presented as a functional prototype and should not be treated as a commercial or forensic-grade verification system.
Click Here for More Latest News