Copyright © 2026 Authors retain the copyright of this article. This article 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.

@article{179618,
        author = {JEEVANANDHAM N and JEESMON S J},
        title = {Enhancing Deepfake Detection with Diversified Self-Blending Images and  Residuals},
        journal = {International Journal of Innovative Research in Technology},
        year = {2025},
        volume = {11},
        number = {12},
        pages = {8692-8702},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=179618},
        abstract = {Deep technology based on advanced artificial intelligence and deep learning has transformed manipulation of media by rendering the possibility of generating extremely realistic but inauthentic images and videos that are quite indistinguishable from the actual content. On the positive side, it is going to help entertain the and educating sectors, but on the negative side, it is giving big challenges in digital security and personal privacy and spreading of the wrong information. Malicious applications in this sense include identity theft fraud and false news that makes people lack the trust to use digital media for their convenience. To discuss each one of these apprehensions, this paper discusses a novel approach to the detection of deepfakes that incorporates diversified self-blending images with residual learning. Real combined with manipulated content into the same image - A self-blending technique tends to bring forward minor inconsistency and artifacts related to that image. In contrast, residual learning identifies faint anomalies that often go unnoticed by other traditional detection techniques. The suggested system is designed based on a hybrid framework, CNN for spatial feature extraction and ResNet for anomaly detection. That design allows for the full examination and robust detection of deepfakes under any datasets and manipulation types. Training with augmentation data improves its robustness and generalization capacity, allowing it to outperform the most existing methods in the task of detecting high-quality deepfakes created by recent GAN-based advanced generators. This work proves that this system is the best in terms of accuracy, recall, and precision. The given system can add to the level of digital security and ensure a protected privacy for everyone. This paper is one more step towards having a secure, trustworthy digital atmosphere and a springboard for developing further innovations of detection technologies of deepfakes.},
        keywords = {Deepfake Detection System Artificial Intelligence Convolutional Neural Network (CNN) Residual Networks (ResNets), Self-Blending Images, Residual Learning, GANs: Generative Adversarial Networks, Anomaly Detection, Augmentation of Images and Media Manipulation Artifacts for Digital Security. Early Deepfake Detection Hybrid Framework. High Quality Deepfakes Feature Extraction Information Prevention Privacy Protect Trustworthy digital ecosystem Digital Media Authenticity.},
        month = {May},
        }