AI-Generated Fishing Attacks and Detection Methods

This paper presented a comprehensive survey of AI-generated phishing attacks and the state-of-the-art detection methodologies designed to counter them. Our analysis reveals that the integration of Large Language Models into the cybercriminal toolkit represents a fundamental shift in the phishing threat landscape—one that renders traditional rule-based and simple ML-based detection approaches largely ineffective. The most promising detection approaches leverage the same transformer architectures that enable sophisticated phishing generation, creating a technological arms race between attackers and defenders. Fine-tuned BERT and RoBERTa models demonstrate strong performance but require continuous updates against evolving generative models. Hybrid multi-modal systems combining NLP analysis with behavioral signals, metadata analysis, and threat intelligence represent the current gold standard. We proposed the SHIELD framework as a practical multi-layered defense architecture. Future work should focus on developing benchmark datasets specifically for AI-generated phishing, creating explainable detection systems, and establishing ethical frameworks for responsible disclosure in this dual-use research domain.

[1] A. Vaswani, N. Shazeer, N. Parmar et al., "Attention is all you need," Advances in Neural Information Processing Systems, vol. 30, pp. 5998–6008, 2017. [2] R. Karanjai, "Targeted phishing campaigns using large language models," arXiv preprint arXiv:2301.00676, 2022. [3] J. Hazell, "Large language models can be used to effectively scale spear phishing campaigns," arXiv preprint arXiv:2305.06972, 2023. [4] M. Bethany, K. Hayawi, and S. Senouci, "PhishBench: A comprehensive benchmark for LLM-generated phishing content detection," in Proc. IEEE S&P 2024. [5] S. Roy, X. He, and X. Yue, "Generating targeted phishing attacks using large language models," in Proc. ACM CCS 2023, pp. 2215–2228. [6] OpenAI, "GPT-4 technical report," arXiv preprint arXiv:2303.08774, 2023. [7] Anti-Phishing Working Group (APWG), "Phishing Activity Trends Report Q1 2024," APWG Publications, 2024. [8] J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, "BERT: Pre-training of deep bidirectional transformers for language understanding," in Proc. NAACL-HLT 2019. [9] Y. Liu, M. Ott, N. Goyal et al., "RoBERTa: A robustly optimized BERT pretraining approach," arXiv preprint arXiv:1907.11692, 2019. [10] S. Liao, R. Engle, C. Weidman et al., "RAIDER: Real-time AI-powered integrated detection for email phishing," IEEE Trans. Inf. Forensics Security, vol. 19, no. 3, pp. 1892–1907, 2024. [11] T. Gangavarapu, C. D. Jaidhar, and B. Chanduka, "Applicability of machine learning in spam and phishing email filtering: Review and approaches," Artif. Intell. Rev., vol. 53, pp. 5019–5081, 2023. [12] P. Bountakas and C. Xenakis, "HELP: A holistic approach to enterprise-level phishing detection systems," Computers & Security, vol. 127, p. 103106, 2023.