A Hybrid Deep Learning Framework for Digital Image Forgery Detection using Error Level Analysis and Grey Wolf Optimized Resnet50 Features

Dr. P. Harish M. Tech, ph. D
Department of Electronics and Communication Engineering Annamacharya Institute of
Technology & Sciences, Tirupathi, India harishpasupulati.aits@gmail.com
Shaik Saad Ahamed
Department of Electronics and Communication Engineering
Annamacharya Institute of Technology & Sciences,
Tirupathi saadbaby960@gmail.com
Kaveti Sai Ram
Department of Electronics and Communication Engineering
Annamacharya Institute of Technology & Sciences, Tirupathi, India
Sram82445@gmail.com
M Muneeswar
Department of Electronics and Communication Engineering Annamacharya Institute of
Technology & Sciences,Tirupathi,balu1muppala@gmail.com

M Radhika
Department of Electronics and Communication Engineering Annamacharya Institute of
Technology & Sciences, Tirupathi, India
radhikamudumuku1@gmail.com

Abstract— Digital image manipulation has become increasingly common due to the availability of advanced editing tools, creating a need for reliable forgery detection techniques. This work proposes a hybrid framework for detecting tampered images by integrating image preprocessing, forensic analysis, deep feature extraction, and optimization techniques. Initially, the input image undergoes preprocessing using a median filter and anisotropic diffusion filter to reduce noise while preserving edge information. Error Level Analysis (ELA) is then applied to highlight compression inconsistencies that may indicate image manipulation. Deep features are extracted from the ELA-enhanced images using the Resnet50 model. To improve feature relevance and reduce redundancy, Grey Wolf Optimization (GWO) is employed to selectthe most informative features from the extracted deep  feature set.