Published Paper Details:

The protection of personal identification numbers and passwords is a significant barrier for cybersecurity. Deceptive login pages soliciting personal information deceive billions of users daily. A variety of nefarious approaches are employed to deceive individuals into accessing harmful websites, such as phishing emails, clickjacking, malware, SQL injection, session hijacking, man-in-the-middle attacks, denial of service, and cross-site scripting. The offender creates a fraudulent yet convincingly comparable website to deceive victims into divulging their credentials. Researchers have offered many security solutions to mitigate these vulnerabilities; however, these methods are both ineffective and susceptible to error. We introduce and implement a client-side defence system that employs machine learning to detect phishing attempts and recognise fraudulent web sites. Our machine learning algorithm serves as a proof of concept for the Google Chrome plugin PhishCatcher, which classifies URLs as either trustworthy or suspicious. The random forest classifier evaluates a login page for authenticity after acquiring four web properties. The precision and validity of the extension were evaluated on multiple real-world web applications. The findings exhibited a precision and accuracy rate of 98.5% when evaluated on 400 authentic URLs and 400 identified phishing URLs. We assessed the latency of our technique using forty phishing URLs. We improved Random Forest by integrating XGBOOST, a technique that evaluates datasets through forest trees or ensembles of estimators to optimise features more efficiently and achieve superior accuracy.