ChatGPT: A New Generation of Dynamic Machine Based Attacks?
summary
Claims that large language models change attack economics get illustrated with specifics: GPT‑3 has 175 billion parameters and ~400 billion tokens; attackers use the API to generate highly personalized phishing and forged Slack-style conversations, produce ransomware in Go and simple obfuscation, and assemble Metasploit exploits; jailbreaks like DAN jailbreak and prompt engineering bypass some safeguards while outputs are often partial but usable; GPT‑4 adds image-to-code and Duolingo integration so GPT‑4 converts images to HTML, widening attack surfaces.
AI-written summary · about
chapters (8)
- 0:00Introduction to ChatGPT
- 5:11Social engineering risks
- 10:36Forging communication
- 17:44Malware and code generation
- 26:06Detection and defense
- 29:58Bypassing guardrails
- 34:28Evolution of GPT-4
- 38:54Conclusion and Q&A
related talks
30mBattling the Machine: Cutting-Edge Attacks and Defenses in Generative AI
36mHow NOT to Train Your Hack Bot: Dos and Don'ts of Building Offensive GPTs
51mBeyond the Hype: Research on How Cybercriminals Are Really Using GenAI
48mGPT3+ Models in Cybersecurity Tasks
01h 24mOffensive Cybersecurity by AI Promises and Pitfalls
46mConversations with ChatGPT: illusion or reality?
01h 04mRobust Generalization in the Era of LLMs: Jailbreaking Attacks and Defenses
17mWhat You Need to Know and How to Respond to the New Wave of GenAI Identity Attacks