ISTA FEST EDITION - Workshop: Hack the Agent: Hands-on AI Agent Security
Стъпка 1 от 3
Избери билети
Избери билети
За събитието
Securing AI Agents & OWASP Top 10
Exploitation, Runtime Guardrails & Hardening
AI agents are rapidly becoming part of modern applications, capable of executing code, calling external tools, accessing sensitive data, and interacting with cloud-native platforms. While these capabilities unlock new possibilities, they also introduce an entirely new attack surface where traditional application security falls short.
This hands-on workshop builds on the conference talk and focuses on ASI02: Tool Misuse and Exploitation from the OWASP Top 10 for Agentic Applications.
Participants will exploit a vulnerable AI agent by manipulating its interaction with external tools, then analyze why the attack succeeds and progressively harden the application using practical security controls such as least agency, scoped identities, runtime guardrails, and sandboxing.
By the end of the workshop, you'll have hands-on experience exploiting and defending AI agents, along with practical techniques you can immediately apply to build more secure AI-powered applications.
🎯 What attendees will learn
- ✔Exploit a vulnerable AI agent using real-world attack techniques
- ✔Understand why tool misuse is one of the most critical risks for AI agents
- ✔Analyze how agent autonomy, tools, and trust boundaries contribute to security failures
- ✔Implement practical defenses including least agency, runtime guardrails, scoped identities, and sandboxing
- ✔Apply guidance from the OWASP GenAI Security Project and the OWASP Top 10 for Agentic Applications
💻 Requirements
- Bring your laptop.
- Every participant will receive a dedicated, isolated lab environment. No cloud account or local setup is required.
- An SSH client is required. If you don't have one installed, we'll help you set it up before the workshop.
- We'll be available throughout the workshop to answer questions and help if you get stuck.