EFFICIENT IMAGE TAMPERING CLASSIFICATION VIA ADAPTIVE HYPERPLANE EVOLUTION

Image manipulation has acquired a fatal significance as digital falsifications have been 
introduced, and this threat has endangered credibility and trust in visual media. The 
conventional watermarking and hashing algorithms lack efficiency as far as the detection of 
more complex manipulations and transformations is concerned and the more diverse and 
intelligent models of detection should be introduced. The proposed study will enhance accuracy 
and robustness of image tampering detector by incorporating two algorithms, that is Adaptive 
Multi-Layer Feature Disentanglement (AMFD) that detects feature and Adaptive Hyperplane 
Evolution Classification (AHEC) that will classify the features. The data employed in the 
assessment is CASIA 2.0 and the suggested AMFD algorithm estimations the independent and 
discriminative characteristics by disentangling the multi-layer representations in a hierarchical 
way and reducing redundancy to improve the quality and uniqueness of the discovered features. 
To perform the optimal separation of the classes, as well as to classify tampered and authentic 
images correctly, the proposed AHEC algorithm classifies pieces separately using adaptively 
evolving hyperplanes on the feature space. The experimental outcomes suggest that both 
algorithms are more effective than the ones that are already known with AMFD implying 
increased compactness of features, ability to separate them, and structural similarity, and AHEC 
implying increased improvements in accuracy, sensitivity, precision, and robustness.