Zero day Attacks Defense Technique for Protecting System against Unknown Vulnerabilities

Abstract

Every organization connected to the internet has one common threat of zero-day attacks. Zero-day exploits are unnoticed until a specific vulnerability is actually identified and reported. Zero-day attacks are difficult to defend against because it is mostly detected only after it has completed its course of action. Protecting networks, applications and systems from zero-day attacks is the daunting task for organization’s security persons. This paper analyzed the research efforts in relation to detection of zero-day attacks. The fundamental limitations of existing approaches are the signature generation of unknown activities and the false alarming rate of anomalous behavior. To overcome these issues, this paper proposes a new approach for zero-day attacks analysis and detection, which senses the organization’s network and monitors the behavioral activity of zero-day exploit at each and every stage of their life cycle. The proposed approach in this paper provides a machine learning based framework to sense network traffic that detects anomalous behavior of network in order to identify the presence of zero-day exploit. The proposed framework uses supervised classification schemes for assessment of known classes with the adaptability of unsupervised classification in order to detect the new dimension of classification.