A Novel Light Weighted Feature Extraction Model for Web Defacement Intrusion Detection

Wide spread use of websites in the cyber-space along with their availability in the public domain has increased the cyber attacks against these platforms. Defacement attack which results in the variation of the website appearance is a common attack launched against organizational websites for some motivations. In this paper, to monitor a website against defacement attack, a machine learning- based approach is proposed. From the continent of the website, a number of features related to text, tags and links are extracted and investigated. In the case of any indication of defacement attack, to evaluate the appearance of the website, its screenshot is taken and fed to a convolutional neural network for final decision. This network is trained using well-known available data sets. Results demonstrate the effectiveness of the proposed defacement detection method with an accuracy rate of 99.8%.