Malicious Webpage Classification

Abstract

Attacks through the web pages containing malicious content have become an increasingly threat to the web security in the recent years. Thus, detection of the malicious URL is an important task to reduce the security threats. To detect malicious URL or web pages, there are several ways but the most traditional technique is through the Black List detection. The Black list contains the list of malicious web pages that are maintained so that user can be aware about the web pages before accessing any webpage. But, the problem with the black list is that it is not an effective method as malicious web pages change frequently, and also growing numbers of web pages that pose scalability issues. A part from blacklist technique, various research techniques have been proposed that use machine learning technique and some use CNN (Convolution Neural Network) to classify web pages into category: malicious or benign. In the paper, a literature survey on classification of malicious web pages is presented that compares various machine techniques with parameter: precision, recall, and F1 score. This survey shows that the Machine learning techniques are better if the features used are textual but when there are images in web page, CNN performs better for the malware image classification.