How to Build Synthetic Persons in Cyberspace
summary
Built a family of Cyberspace Cognitive agents — CyCog-A, CyCog-D — inside a single VM that runs standard tools (nmap, sqlmap, snort, iptables) with an IRC C2 and a Java RAT. Agents use Soar/ACT-R-style goal graphs, an asset DB and persona files to plan, learn and act; instance‑based learning models user click behavior from email and proxy logs. Demo successfully phishes an inbox, infects three Windows hosts and exfiltrates about four to five “Project Titan” files, and the attacker model proves stronger at adversary emulation than pen testing.
AI-written summary · about
chapters (20)
- 0:00Start
- 1:46(Partial) Al Taxonomy
- 3:23Non-Symbolic AI Example
- 5:34black hat The Tactical Battlefield of 2050
- 6:25Al and Cyber Attacks
- 7:30Imperfect AI
- 9:29Autonomous Agents
- 9:52CyCog Agent Genealogy
- 13:43CyCog-A User Interface
- 15:06Inherently Explainable Al
- 17:51Modeling TTPS
- 19:36Agent Code
- 21:25CyCog Agents
- 22:39ACT-R Agent Learning
- 23:53CyCog Abstraction Layer
- 26:27Putting It All Together
- 35:31Keeping Track of Cyberstuff
- 36:15Formal TTP Models
- 36:56TTP Model + ATT&CK
- 38:12black hat Using ML to Build TTP Models
related talks
35mThe long road to autonomous cybersecurity
50mCatPhish Automation - The Emerging Use of Artificial Intelligence in Social Engineering
33mAgentic ProbLLMs: Exploiting AI Computer-Use and Coding Agents
01h 24mBuilding your own software factory
24mAI Agent Swarms Are Amazing
40mHow To Tame Your Agents?
49mAI Agents: Collaboration in Real-Life Applications
32mFriend or Replicant: How Attackers Automate and Disguise Themselves