Detecting Malicious URLs using Lexical Analysis and Network Activities

The majority of unlawful actions have utilised web URLs. This behaviour has a significant financial impact on businesses, as well as a loss of intellectual or copyright property and a negative reputation. This effect is a major loss for any organization, which may be small or large depending on the loss. This research work has introduced a lightweight approach for the detection and categorization of malicious URLs according to their type, and finally it has been shown that the lexical analysis is effective and efficient for the proactive detection of these URLs. The existing research works have used three multi-class classification algorithms: K-nearest neighbours (KNN), Random Forest, and C4.5 to classify the network intrusion into four categories like benign, defacement, malware, and phishing; 78 different lexical features are used and one more obfuscating technique across 70,000 URLs. The proposed research work has used four different algorithms: Random Forest, K-Nearest Neighbour (KNN), Decision Tree, and Extra Tree Classifier, and utilized the top 15 lexical features, which are sufficient and necessary to classify all the four categories over 700,000 URLs.