Phishing Website Detection using Hyper-parameter Optimization and Comparison of Cross-validation in Machine Learning Based Solution

The development of the internet and web technologies has led to their use in a variety of services. It has led to an increase in cyber security problems over time, the most well-known of which is the phishing assault, in which malicious websites appear as legitimate websites to steal the details of unsuspecting users needed for unauthorised access. Anti-phishing software and machine learning (ML) techniques are now used as mitigation measures, and they have been successful in detecting phishing activities. On the other hand, hackers are developing new methods to get through existing defences. Despite this, there is a continuing need for innovative and effective website phishing detection solutions due to the vitality of phishing operations. In this study, phishing website detection is proposed using machine learning classifiers and incorporates several cross-validation techniques to obtain the highest level of accuracy (97.3%). The random forest's hyper parameter tuning has finally been used to get successful results. The Phishtank repository, a collection of authentic and phishing websites, is used to assess the efficiency of the proposed system. The results of the experiment showed that hyper parameters outperformed a few chosen baseline classifiers. High detection accuracy was reported by Hyper parameter (97.6%). Therefore, the proposed approach is suggested for addressing complex phishing assaults.