OneShield Privacy Guard: Deployable Privacy Solutions for LLMs
summary
OneShield Privacy Guard is a deployable framework integrated with IBM Krenite that enforces real-time, context-aware privacy across input, inference and output, claiming improved PII detection and better privacy-utility tradeoffs. Deployments include an enterprise multilingual data-and-model factory and an open-source PR triage that cuts manual intervention by 30%. Detection uses regex + ML with lightweight models (CNNs and Naive Bayes), semantic multilingual classifiers, and outperforms or matches tools like AWS Comprehend on many PII types.
AI-written summary · about
chapters (8)
- 0:00OneShield Privacy Guard intro
- 1:35Privacy in enterprise LLMs
- 3:02Protecting LLM inputs and outputs
- 5:30Lightweight PII detection models
- 8:47Deployment results and compliance
- 9:31Open-source challenges and ranking
- 12:44Lessons learned and future work
- 13:48Q&A
related talks
28mPrivacy Engineering Networking Break AMA
28mHow Privacy Engineers Deliver ROI
16mPrivacy in the Public Sector: Lessons Learned and Strategies for Success
08mLLMs Without Leaks: Keep Your Code, Data & IP
06mData Privacy Concerns and Solutions When using LLMs in Production
29mBeyond the Hype: Building Trustworthy and Reliable LLM Applications with Guardrails
43mPanel: Privacy Design Patterns for AI Systems: Threats and Protections
44mPrivacy by Design Machine Learning