Framework anonymizes sensitive UI data with type-preserving placeholders for cloud-based GUI agents. Detects PII across screenshots, XML, and instructions via layered architecture. Achieves top privacy-utility trade-off on benchmarks.
Key Points
- 1.Available-but-invisible access principle
- 2.PII detector and UI transformer
- 3.Minimal utility degradation
Impact Analysis
Protects personal data without hindering task automation. Improves trust in MLLM-powered agents.
Technical Details
Uses secure proxy and gatekeeper for consistent anonymization. Supports local computations for raw values.