VeriSight Sentinel: Deepfake Detection and Reporting System

Abstract

This project focuses on Deepfake Detection and Reporting, addressing the growing concern of synthetic media manipulation using artificial intelligence. Deepfakes, which use deep learning techniques to alter or generate realistic audio, video, or images, pose serious threats to privacy, misinformation, and digital trust. The proposed system employs Convolutional Neural Networks (CNN) and Machine Learning algorithms to detect inconsistencies in facial expressions, blinking patterns, and pixel-level artifacts that are often present in deepfake content. A trained model analyzes input media to identify forged elements with high accuracy. Once detected, the system can automatically generate a detailed report highlighting the probability of manipulation, affected regions, and potential sources. This tool can be integrated into social media platforms, news verification systems, and digital forensics. By combining detection and real-time reporting, the project aims to strengthen digital media integrity, promote responsible AI usage, and empower users to identify and report deepfakes effectively and efficiently.

Introduction

In the digital era, deepfake technology—powered by advanced AI models like Generative Adversarial Networks—enables the creation of highly realistic but manipulated media (images, videos, audio, and text). While it has useful applications, it also poses serious risks such as misinformation, fraud, and reputational damage, making detection increasingly difficult.

To address this issue, the project VeriSight Sentinel proposes an AI-based system for detecting and reporting deepfakes across multiple media types. The system integrates machine learning, deep learning, and generative AI to identify manipulations and generate detailed authenticity reports, improving digital trust.

The framework uses:

• CNN models (e.g., ResNet, EfficientNet) for image analysis
• CNN–LSTM hybrid models for video (capturing spatial and temporal patterns)
• Spectrogram-based CNNs for audio detection
• Transformer-based models like BERT for text analysis

It is implemented as a web-based platform (Flask backend, React frontend) with secure storage and real-time analysis capabilities.

The system improves upon existing solutions by providing a comprehensive, multimodal detection approach, unlike earlier models that focused mainly on facial videos or single media types. It also includes explainable outputs for better user understanding.

However, challenges remain, such as detecting subtle manipulations, handling diverse media formats, and ensuring high accuracy across all modalities.

Conclusion

The project VeriSight Sentinel – Deepfake Detection and Reporting System aims to develop a secure, authenticated, and AI-driven platform capable of detecting deepfakes across multiple media types. Currently, the Image Deepfake Detection Module has been successfully implemented using CNN, ResNet, and EfficientNet models, achieving accurate results in identifying manipulated images. The system also ensures secure storage of results using SQLite and allows authenticated access for users. The project work is still ongoing, and further development will focus on integrating video, audio, and text detection modules to make VeriSight Sentinel a complete, multimodal deepfake detection and reporting system.

References

[1] Deressa Wodajo and Solomon Atnafu. Deepfake video detection using convolutional vision transformer. arXiv preprint arXiv:2102.11126, 2021. [2] Irene Amerini, Leonardo Galteri, Roberto Caldelli, and Alberto Del Bimbo. Deepfake video detection through optical flow based cnn. In Proceedings of the IEEE/CVF International Conference on Computer Vision Workshops, pages 0–0, 2019. [3] David Guera and Edward J Delp. Deepfake video detection using recurrent ¨ neural networks. In 2018 15th IEEE international conference on advanced video and signal-based surveillance (AVSS), pages 1–6. IEEE, 2018. [4] Nicolo Bonettini, Edoardo Daniele Cannas, Sara Mandelli, Luca Bondi, Paolo ` Bestagini, and Stefano Tubaro. Video face manipulation detection through ensemble of cnns. In 2020 25th International Conference on Pattern Recognition (ICPR), pages 5012–5019. IEEE, 2021. [5] Hany Farid. Photo forensics. MIT press, 2016. [6] Lorant. Lincoln-picture-story-his-life. www.amazon.com, 1969. [7] Phillip Isola, Jun-Yan Zhu, Tinghui Zhou, and Alexei A Efros. Image-to-image translation with conditional adversarial networks. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 1125–1134, 2017. [8] Mart´?n Abadi, Paul Barham, Jianmin Chen, Zhifeng Chen, Andy Davis, Jeffrey Dean, Matthieu Devin, Sanjay Ghemawat, Geoffrey Irving, Michael Isard, et al. Tensorflow: A system for large-scale machine learning. In 12th {USENIX} symposium on operating systems design and implementation ({OSDI} 16), pages 265–283, 2016. [9] F. Chollet et al. Keras. https://keras.io, 2025. [10] Ayush Tewari, Michael Zollhofer, Hyeongwoo Kim, Pablo Garrido, Florian Bernard, Patrick Perez, and Christian Theobalt. Mofa: Model-based deep convolutional face autoencoder for unsupervised monocular reconstruction. In Proceedings of the IEEE International Conference on Computer Vision Workshops, pages 1274–1283, 2017. [11] Grigory Antipov, Moez Baccouche, and Jean-Luc Dugelay. Face aging with conditional generative adversarial networks. In 2017 IEEE international conference on image processing (ICIP), pages 2089–2093. IEEE, 2017.

Copyright

Copyright © 2026 Sakshi Yadav, Vaishnavi Misal, Princy Singh, Jitesh Bhoir, Prof. Vijayalaxmi Tadkal. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.