Reproduce “PhishGen: Understanding LLM Capabilities in Automated Phishing Website Generation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Fujiao Ji, Yerim Kim, Yena Cho, Seonhye Park, Hyoungshick Kim, Doowon Kim
PhishGen tests whether LLM-generated login pages collect credentials end to end, not merely render. It injects synthetic credentials, submits them, and checks local persistence. While 96.2% of safeguard-bypassing outputs were valid websites, 10,889 of 28,121 pages passed functionality testing (38.7%), and mean user-perceived similarity was 2.02/5. GPT 4.1 reached 89.7%; the controlled test does not measure real-world campaign success.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.