AIHEM Deep Dive: Tool Use & Function Calling Attacks
Master attacks against AI agent tool use. Learn to exploit function calling, manipulate tool selection, and bypass tool safety mechanisms.
- Date and time
- Thursday, May 21, 202610:00 AM - 1:00 PM PST
- Location
- virtual
- Price
- Free
- Capacity
- 30 seats
- Topics
- Function Calling, Tool Manipulation, Safety Bypasses, Parameter Injection
Speakers
- Guard0 Security TeamSecurity Researchers
Training Overview
Tool use is what makes agents powerful—and dangerous. This training focuses specifically on attacking the tool/function calling capabilities of AI agents.
Topics Covered
Tool Use Architecture
- How agents select tools
- Function calling mechanisms
- Tool definition vulnerabilities
- Parameter handling
Attack Techniques
- Forcing unintended tool calls
- Parameter injection
- Tool confusion attacks
- Chaining tool calls maliciously
Safety Bypass Methods
- Circumventing tool restrictions
- Escalating tool permissions
- Accessing hidden tools
- Disabling safety checks
Hands-on Labs
- Lab 1: Tool enumeration and mapping
- Lab 2: Parameter injection attacks
- Lab 3: Tool confusion exploitation
- Lab 4: Safety mechanism bypass
Prerequisites
- AIHEM Fundamentals completed
- Understanding of API design
- JSON/schema knowledge helpful
Materials
- Tool attack playbook
- Lab environment
- Tool security assessment guide
This event has ended.
View upcoming events