This research proposes that image forgery detection is the main significant obstacle encountered in systems that function in actual time, is social networking sites and internet-based data portals. The scope of traditional methods of identification depending on the remnants of picture alterations is limited by preconceived assumptions, such as size, contrast, and hand-crafted features. In this paper, we offer a fusion-based judgment technique for identifying picture fraud. Three lightweight deep learning models—Squeeze Net, Mobile NetV2, and Shuffle Net—serve as the framework for the decision fusion. The fusion decision system's implementation consists of two steps. The retained weights of the lightweight deep learning models are used to determine whether the photos are fabricated in the outset. The outcomes of the forged photos are then contrasted with those of the previously trained models using the fine-tuned weights. The experimental findings imply that, in comparison to cutting-edge methods, the fusion-based decision technique achieves higher accuracy.