Blockage of Phishing Attacks Through Machine Learning Classification Techniques and Fine Tuning its Accuracy

Blocking phishing attacks using machine learning classification techniques is a promising approach to enhancing email security. Phishing attacks can involve the use of fictitious identities to trick victims into giving personal information or installing risky software. Machine learning may help distinguish these fake emails by looking at various features and patterns. Phishing exploits inherent human tendencies in order to elicit the disclosure of sensitive information via online means. Phishing assaults often start via the use of electronic mail and include a range of potential aims. The cybercriminal will use social engineering tactics to deceive the victim into clicking on the link included inside the phishing email, so redirecting them to the compromised website. The complexity of these attacks increases as hackers include a human element into their fraudulent activities, therefore creating communications that possess a real appearance. Phishing faces billions of dollars in annual expenditures for Internet users. Identity thieves employ seductive methods to sift through a pool of trusting Internet users in search of personal information. Phishing software and phony emails are used by scammers to get personal information, including passwords and other bank account details. There are three types of tuning elements used: feature selection, hyperparameter optimization, and data balancing. The experiment utilizes two separate datasets acquired from online sources, namely the UCI repository and the Mendeley repository, with the inclusion of the eight most often used machine learning algorithms. The proposed research aims to block phishing attacks and improve accuracy using feature extraction and the CatBoost algorithm. The results will be shown, and the tuning parameters will be shown to increase the effectiveness of machine learning methods.