GAN-Based One-Class Classification SVM for Real time Medical Image Intrusion Detection

Abstract

Medical data attack and detection technology has been a hot topic in the past few decades precisely as numerous attacks on hospitals and clinics led to the loss of data. Although many methods have been developed for detection and discrimination of fake images, the problem has not yet been properly solved. Classifying the data method tends to produce higher error when compare to other methods due to the large variance directions. One-class classification is a fairly competitive method for detecting fake medical images due to the data's unbalanced nature. However, it can also produce higher error when compare to other methods. One of the most effective ways to improve the accuracy of one-class classification is by implementing covariance-guided support vector machine (iCOSVM) especially with a real time system. Therefore, in this paper, we present a case study that uses incremental covariance-guided support vector machine to build suitable detection system. The results of the study showed that the proposed detection system is very accurate and efficient. It utilizes the training data to improve its accuracy and minimize its error. The iCOSVM supports incremental projections, improves significantly the performance of the one-class support vector machine. Additionally, our proposed detection system is very accurate and efficient comparing to other incremental one class classifications algorithms, outperforming the batch learning system as well.