From Text to Flaws: Vulnerabilities in Apps Using Generative AI and LLMs
summary
Overview of how large language models (LLMs) and generative AI are integrated into production apps, explaining LLM architectures and typical application patterns. Describes common vulnerabilities that arise in apps using generative AI (with concrete examples) and gives practical countermeasures developers can implement to mitigate those risks.
AI-written summary · about
chapters (16)
- 0:00LLM introduction
- 3:51How LLM apps work
- 11:11Architecture and vulnerabilities
- 13:58Basic prompt injections
- 15:45Exfiltration via external tools
- 17:51Code injections
- 19:59Multimodal attacks
- 20:20GPT Custom and data leaks
- 24:19Classic protection techniques
- 25:25LLM-specific solutions
- 30:55Advanced defense techniques
- 34:12Dual LLM and quarantine
- 35:40Human-in-the-loop
- 36:03Canaries and monitoring
- 38:18Best practices and design
- 40:19Conclusion
related talks
40mFrom Text to Flaws: Application Vulnerabilities with Generative AI and LLMs
40mFrom Text to Exploits: Vulnerabilities of Apps Using Generative AI and LLMs
42mFrom Text to Flaws: Application Vulnerabilities with Generative AI and LLMs
51mFrom Text to Flaws: vulnerabilities in applications with Generative AI and LLMs
45mOverview of Risks, Vulnerabilities and Paths to Safer Use of Generative AI
49mSecuring AI Apps with the OWASP Top Ten for Large Language Models
31mHacking with Words...
47mCurrent Security Threats in Large Language Model Applications