AI Driven Exploit Mitigation for Zero Day Vulnerability Using SVM and Autoencoder
Padmalakshmi. R. S, Shawn Basil. C. K · 2025
This study is about the action taken to reduce the risk or impact of a "zero-day vulnerability" (flaw or weakness in a software that is unknown to the makers or the public but that can be exploited by attackers). This study helps to analyze the exploits in a software to prevent cyberattacks. This study evaluates the performance of SVM and Autoencoder to classify the "zero-day vulnerability" dataset. SVM classifier got an accuracy of 94.9%, precision and recall values of 20.83% and 30%, f1-score at 23% and specificity of 94%. Whereas, autoencoder got an accuracy of 81.6%, precision and recall values of 21% and 28%, f1-score at 21.05% and specificity of 81%. According to the achieved values, SVM provides better accuracy than autoencoder in analyzing the dataset.