Phishing Email Detection Based on Binary Search Feature Selection.

Area of Science:

• Cybersecurity
• Machine Learning
• Data Science

Background:

• Phishing emails pose a significant threat in cybersecurity.
• Email remains a primary vector for phishing attacks, targeting unsuspecting users.
• Effective detection of phishing emails is crucial for digital security.

Purpose of the Study:

• To propose and evaluate a novel method for detecting phishing emails.
• To utilize binary search feature selection (BSFS) with a Pearson correlation coefficient for feature ranking.
• To assess the performance of BSFS against other feature selection methods.

Main Methods:

• Employed binary search feature selection (BSFS) for identifying relevant features.
• Utilized Pearson correlation coefficient as a ranking metric for feature selection.
• Extracted 41 features across four dimensions: email subject, body, hyperlinks, and readability.
• Compared BSFS with Sequential Floating Forward Selection (SFFS) and Wrapper Feature Selection (WFS).

Main Results:

• The proposed BSFS method achieved a high accuracy of 97.41% in phishing email detection.
• BSFS outperformed SFFS (95.63% accuracy) and WFS (95.56% accuracy).
• BSFS demonstrated efficiency, requiring less time to identify the optimal feature set compared to SFFS.

Conclusions:

• Binary search feature selection (BSFS) is an effective and efficient technique for phishing email detection.
• BSFS offers superior accuracy and comparable or better time efficiency than other tested feature selection algorithms.
• The method successfully identifies critical features for enhanced email security.