"ChatGPT Shows Promise in Detecting Phishing Sites"

Researchers wanted to know whether ChatGPT can reliably detect phishing sites. They tested 5,265 URLs (2,322 phishing and 2,943 safe). They asked ChatGPT (GPT-3.5) the question: "Does this link lead to a phish website?" The Artificial Intelligence (AI)-driven chatbot had an 87.2 percent detection rate and a 23.2 percent false positive rate based solely on the URL form. According to the researchers, although the rate of detection is high, the rate of false positives is unacceptable. The results were much worse when they asked a slightly different question: "Is this link safe to visit?" The detection rate was 93.8 percent and the false positive rate was 64.3 percent. The more general prompt is more likely to result in a conclusion that the link is malicious. Both approaches yielded unsatisfactory results, but the researchers still believe it is possible to use this type of technology to aid human analysts by highlighting suspicious URL parts and suggesting potential attack targets. This article continues to discuss the potential use of ChatGPT to detect phishing sites.

Help Net Security reports "ChatGPT Shows Promise in Detecting Phishing Sites"