One of the significant challenges in the realm of cyber attack detection is the scarcity of training data, which remains a formidable obstacle. Despite the utilization of established network monitoring tools like Wireshark, a vast number of individuals are still at risk due to the absence of information regarding website behaviors and features that can signify an impending attack. Interestingly, most attacks don’t hinge on intricate coding or evasion techniques employed by threat actors but rather stem from victims lacking fundamental tools to detect and thwart these attacks.
In this context, machine learning is emerging as a transformative force in comprehending the nature of cyber-attacks. This study leveraged machine learning techniques with a focus on Phishing Website data, aiming to compare five algorithms and provide valuable insights that can empower the general public to avoid falling into phishing traps. The study’s findings underscore the effectiveness of the Neural Network algorithm as the top performer. The model identifies several key features indicative of phishing websites, including the inclusion of an IP address in the domain name, longer URLs, the use of URL shortening services, the presence of “@” and “-” symbols in the URL, non-trusted SSL certificates with a validity period of less than 6 months, domains registered for less than one year, and favicon redirection from other URLs.
The Neural Network, founded on multi-layer perceptron principles, serves as the cornerstone of artificial intelligence. This suggests that, in the future, the task of phishing detection may become automated and driven by artificial intelligence, enhancing our cyber security measures.