Hi All
There's no denying we're seeing explosive adoption of GenAI and LLM use cases and applications
That said, it's not too late for organizations to utilize one of the best secure software development methodologies out there, Threat Modeling to ensure secure adoption of LLM's.
This is an excellent paper discussing "Threat Modeling and Risk Analysis for LLM-Powered Applications"
It covers:
- Potential attacks against LLM-powered applications and their potential impacts
- LLM specific attack vectors and risks
- Widely used Threat Modeling methodologies and adapting them for LLM applications (cc: 😷 Adam Shostack)
- An example LLM-application threat model
Definitely a key activity organizations should be having their security, development and engineering teams implement as they move forward adopting LLM and GenAI use cases for organizational outcomes, and doing so securely.
Regards
Caute_Cautim
