ENHANCING MEDICAL IMAGE SECURITY: A PCA-DEEP LEARNING APPROACH FOR ATTACK DETECTION

Abstract

The integrity and authenticity of medical images are critical for accurate diagnosis and treatment planning. However, they are increasingly vulnerable to malicious attacks like tampering and manipulation. This work proposes a novel deep learning-based approach for medical image attack detection. The proposed model leverages Principal Component Analysis (PCA) as a feature extractor to capture essential information from the images while reducing dimensionality. Subsequently, two deep learning classifiers, Convolutional Neural Network (CNN) and Inception, are employed for classification. The model is trained and tested on a deepfake dataset, simulating potential attack scenarios in the medical domain. This approach aims to identify tampered medical images and enhance the security of medical image analysis. This research work investigates the effectiveness of the proposed model in distinguishing between genuine and manipulated medical images, paving the way for safeguarding the integrity of medical data and ensuring reliable decision-making in healthcare.