CatPhish Automation - The Emerging Use of Artificial Intelligence in Social Engineering
summary
Demonstrates a reproducible pipeline that turns open LLM APIs into autonomous phishing bots: code an API loop, prepopulate the conversational context to give the model a persona (GPT-3.5-turbo via API, RLHF-trained), scrape target profiles to feed the victim model, and inject mission text so the bot solicits secrets. Builds and tests personas (e.g. Sarah Ellis) that infer details and execute scams including a live Social Security Administration scam, credential harvesting and wire‑fraud scripts, and shows the approach reliably succeeds in demos.
AI-written summary · about
chapters (17)
- 0:00Intro
- 1:44Roadmap
- 9:13The Imitation Game
- 11:20Eliza
- 13:09Replika
- 13:59Transformer Architecture
- 14:43GPT-3 Released
- 21:48Transform it to code
- 23:09Establish a sense of self-identity
- 25:22Train on target identity
- 26:55Step 4 - Execute the attack (pretext)
- 28:38Step 4 - Execute the attack (accomplice)
- 35:41Social Security Scam
- 38:45Credential Harvesting Scam
- 39:01Wire Fraud Scam
- 39:50Key takeaways?
- 46:02What you can do?
related talks
51mDeepPhish: Simulating Malicious AI
51mDeepPhish: Simulating Malicious AI
11mHow to Navigate Emerging AI-Driven Security
42mWeaponizing Data Science for Social Engineering: Automated E2E Spear Phishing on Twitter
31mAI-Powered Solutions for Cybersecurity
20mAI Red Teaming and Automation: Exploring Societal Risks in...
23mCyberattacks easily implemented with AI?
33mAgentic ProbLLMs: Exploiting AI Computer-Use and Coding Agents