Simulation attack on LLM using prompt injection technique
Simulation attack on LLM using prompt injection technique
Speaker: Mrigakshi Goel
Abstract: Attackers can manipulate LLMs through crafted inputs, causing them to execute the attacker's intentions. This can be done directly by adversaries prompting the system or indirectly through manipulated external inputs, potentially leading to data exfiltration, social engineering, and other issues. This kind of attack is one of the OWASP top 10 attacks on LLMs and is called prompt injection. We will be doing a deep dive into this attack with a short demo on how an attacker can manipulate LLMs.
Track: Security Operations