HYBRID LIGHTWEIGHT AI MODELS FOR ACCURATE IMAGE FORGERY DETECTION

Abstract

The rapid growth of digital image editing tools and social media platforms has significantly increased the spread of manipulated and forged images, creating serious concerns regarding information authenticity and security. Image forgery detection has therefore become an important research area in digital forensics. This paper presents an efficient approach for image forgery detection based on the fusion of lightweight deep learning models. The proposed system combines multiple lightweight convolutional neural network architectures to improve detection accuracy while maintaining low computational complexity and faster processing speed. The fusion strategy enables the extraction of complementary spatial and texture-based features from suspicious images, allowing the system to effectively identify tampered regions and manipulated content. The framework is designed to detect common image forgery techniques such as copy-move forgery, splicing, and image retouching. Data preprocessing and augmentation techniques are applied to improve model robustness and generalization performance. Experimental evaluation demonstrates that the fusionbased lightweight model achieves high detection accuracy with reduced memory consumption compared to traditional deep learning approaches. The proposed method is suitable for real-time and resourceconstrained applications, including mobile devices and cloud-based forensic systems. Overall, the system provides a reliable, scalable, and computationally efficient solution for modern digital image forgery detection challenges.