A Hybrid Deep Learning Approach for Image Forgery Detection Using Transfer Learning and Error Level Analysis

Abstract

The rapid expansion of digital media has made image manipulation a widespread concern, contributing to the circulation of misleading and fabricated content across online platforms. Manual identification of such altered images is both challenging and unreliable. This work presents a hybrid approach for detecting image forgery by combining transfer learning with forensic analysis techniques. The proposed system enables users to upload digital images, which are processed through a structured pipeline that includes validation, preprocessing, and deep learning-based classification. A pre-trained MobileNet model is employed to extract relevant features and determine whether an image is authentic or tampered. To improve interpretability, a confidence score is generated for each prediction, indicating the reliability of the result. Furthermore, Error Level Analysis (ELA) is incorporated to visually highlight regions that may have been manipulated. The system is implemented using a backend framework for processing, OpenCV for image operations, and TensorFlow for model integration. Experimental results demonstrate that the approach achieves high detection accuracy while also providing visual evidence, making it a practical solution for real-world applications.

Introduction

The document focuses on the growing problem of digital image forgery, which has become widespread due to advanced editing tools and social media. Such manipulated images are often used to spread misinformation, mislead audiences, and create false evidence, making reliable detection important for fields like journalism and digital forensics. Traditional detection methods based on manual inspection and basic image processing are slow, subjective, and ineffective against complex manipulations.

To overcome these limitations, recent research uses Artificial Intelligence and deep learning techniques, especially Convolutional Neural Networks (CNNs), which automatically learn image features and achieve higher accuracy. Transfer learning with pre-trained models like MobileNet improves efficiency and performance, especially with limited data. In addition, forensic techniques such as Error Level Analysis (ELA) help identify tampered regions by detecting compression inconsistencies. However, many existing systems lack explainability and visual evidence of manipulation.

The proposed system introduces a hybrid approach that combines transfer learning (MobileNet) with ELA. The system processes input images through preprocessing, feature extraction, classification, and decision-making stages. It classifies images as authentic or tampered, provides confidence scores, and uses a threshold-based decision rule.

Conclusion

In this work, a deep learning-based system for detecting image forgery has been successfully developed and implemented. The proposed system utilizes a combination of Error Level Analysis (ELA) and transfer learning techniques to identify manipulated regions and classify images as authentic or tampered. The use of the MobileNet architecture enables efficient feature extraction and accurate classification of input images while maintaining computational efficiency. The integration of the trained model into a web-based interface provides a user-friendly platform that allows users to upload images, perform real-time forgery detection, and visualize the results with highlighted manipulated regions. The system ensures smooth interaction and accessibility, making it practical for applications in digital forensics and media verification. Experimental results demonstrate that the system achieves high accuracy, along with balanced precision, recall, and F1-score values, indicating that the model is reliable and effective in detecting image forgeries. The incorporation of ELA further enhances the system’s capability by providing visual interpretation of suspicious areas, improving transparency and trust in the detection process. Overall, the proposed system offers a robust, scalable, and efficient solution for image forgery detection. It can assist in preventing the spread of fake images, support forensic investigations, and improve the authenticity verification process in various real-world scenarios.

References

[1] K. D. Kadam, S. Ahirrao, and K. Kotecha, “Multiple Image Splicing Dataset (MISD): A Dataset for Multiple Splicing,” Data, vol. 6, no. 10, p. 102, Sep. 2021. [2] R. Agarwal, O. P. Verma, A. Saini, A. Shaw, and A. R. Patel, “The Advent of Deep Learning-Based Techniques,” in Innovative Data Communication Technologies and Application. Singapore: Springer, 2021. [3] M. A. Elaskily, M. H. Alkinani, A. Sedik, and M. M. Dessouky, “Deep Learning-Based Algorithm (ConvLSTM) for Copy-Move Forgery Detection,” Journal of Intelligent & Fuzzy Systems, vol. 40, no. 3, pp. 4385–4405, Mar. 2021. [4] A. Mohassin and K. Farida, “Digital Image Forgery Detection Approaches: A Review,” in Applications of Artificial Intelligence in Engineering. Singapore: Springer, 2021. [5] K. B. Meena and V. Tyagi, Image Splicing Forgery Detection Techniques: A Review. Cham, Switzerland: Springer, 2021. [6] S. Gupta, N. Mohan, and P. Kaushal, “Passive Image Forensics Using Universal Techniques: A Review,” Artificial Intelligence Review, vol. 55, no. 3, pp. 1629–1679, Jul. 2021. [7] M. M. Qureshi and M. G. Qureshi, Image Forgery Detection & Localization Using Regularized U-Net. Singapore: Springer, 2021. [8] Y. Rao, J. Ni, and H. Zhao, “Deep Learning Local Descriptor for Image Splicing Detection and Localization,” IEEE Access, vol. 8, pp. 25611–25625, 2020. [9] K. M. Hosny, A. M. Mortda, N. A. Lashin, and M. M. Fouda, “A New Method to Detect Splicing Image Forgery Using Convolutional Neural Network,” Applied Sciences, vol. 13, no. 3, p. 1272, Jan. 2023. [10] S. S. Ali, I. I. Ganapathi, N.-S. Vu, S. D. Ali, N. Saxena, and N. Werghi, “Image Forgery Detection Using Deep Learning by Recompressing Images,” Electronics, vol. 11, no. 3, p. 403, Jan. 2022. [11] K. Kadam, S. Ahirrao, K. Kotecha, and S. Sahu, “Detection and Localization of Multiple Image Splicing Using MobileNet v1,” IEEE Access, vol. 9, pp. 162499–162519, 2021. [12] R. Sari and M. Fahmi, “Effect of Error Level Analysis on CNN-Based Digital Image Forgery Detection,” Procedia Computer Science, vol. 179, pp. 335–343, 2021. [13] S. Qazi, I. Hussain, and M. Saleem, “ResNet50v2-Based Deep Learning Model for Image Splicing Detection Using Transfer Learning,” IEEE Access, vol. 8, pp. 182633–182646, 2020. [14] L. Castillo Camacho and Z. Wang, “A Comprehensive Review of Deep Learning-Based Image Forensics,” Forensic Science International: Digital Investigation, vol. 36, p. 301090, 2021. [15] R. Katiyar and V. Bhavsar, “Explainable CNN-Based Image Forgery Detection Using Grad-CAM,” Journal of Electronic Imaging, vol. 31, no. 5, p. 053029, 2022. [16] R. Khalil, A. Al-Ali, T. Zia, and S. Al-Maadeed, “Enhancing Digital Image Forgery Detection Using Transfer Learning,” IEEE Access, vol. 11, pp. 102345–102359, 2023. [17] R. Hebbar and R. Kunte, “Transfer Learning Approach for Splicing and Copy-Move Image Tampering Detection,” Multimedia Tools and Applications, vol. 80, no. 19, pp. 29607–29623, 2021. [18] S. Ali, B. Ahmed, and U. Khan, “Recompression-Based Preprocessing for Image Forgery Detection Using CNNs,” Journal of Visual Communication and Image Representation, vol. 89, p. 103633, 2022. [19] M. Abdalla, H. Mirza, and S. Hussain, “Deep Transfer Learning for Image Forgery Detection,” Pattern Recognition Letters, vol. 125, pp. 701–708, 2019. [20] X. Yu, Y. Liu, and X. Zhao, “MCNet: Multi-Domain Feature Learning for Image Manipulation Classification in JPEG Images,” IEEE Transactions on Information Forensics and Security, vol. 15, pp. 2035–2049, 2020. [21] S. Qazi, M. Saleem, and I. Hussain, “Deep Transfer Learning for Image Forgery Detection Using Steganalysis Features,” Expert Systems with Applications, vol. 133, pp. 1–12, 2019.

Copyright

Copyright © 2026 Ajna Ashraf, Arsha K B, Navami P M, T S Rithuparna, Ms. Mahshiya Mishab. 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.