Machine Learning Approach for Detection of Spam Url: Performance Evaluation of Selected Algorithms

Abstract

The Internet, web consumers and computing systems have become more vulnerable to cyber-attacks. Spam which exist in different form has recently becomes one of the techniques attackers use to get confidential information from their victims. Whatever is the form of spam, Uniform Resource Locator (URL) serves as a key driver for spam. Hence, detection of spam URL has attracted attention of many researchers. Machine learning approach is one of the approaches researchers have used in this area of study. Meanwhile, no researcher has reported 100% accuracy with any machine leaning algorithm and not all machine learning algorithms has been explored in this area of research. Consequently, this study presents performance evaluation of some selected algorithms with the aim of identifying best algorithm in terms of accuracy, precision, sensitivity, specificity, mean Squared Error. WEKA data mining tool was used carry out experiment on the selected algorithms. The results of our experiment revealed that K-NN out performed other algorithms with highest values in accuracy, precision, sensitivity and with lowest values in specificity and mean squared error.