Surprise Masking: Private Data Filtering with Model Likelihoods
Surprise Masking uses perplexity-based filtering against a public reference model to strip private data from training corpora—a practical pre-processing step that could reduce memorisation risk for enterprise fine-tuning.
Summary written by editorial AI · Source link below
arXiv:2610.22988v1 Announce Type: new Abstract: Data useful for language model (LM) training can contain private or unwanted information that is at risk of being learned or memorized. We propose Surprise Masking (SM), a simple form of perplexity filtering that uses a publicly available reference model to identify unwanted spans and suppress learning from them while preserving the rest of each example. On synthetic code benchmarks, we show SM improves the privacy-utility tradeoff over standard m
Editorial Analysis
Enterprises fine-tuning LLMs on internal data face memorisation-driven data leakage; a lightweight filtering step could reduce GDPR exposure before training begins.
Test perplexity-based data filtering on a representative sample of your training data to quantify how much sensitive content it catches before model fine-tuning.
Forward-looking interpretation drafted by editorial AI under human review — not a reproduction of the source. See methodology.
External link — opens at arXiv Crypto & Security in a new tab.
More from the AI Security Desk
- OpenAI AI Agents Demonstrate VM Escapes and Secret Cross-Sandbox Communication1d
- Exposed Infrastructure Reveals How Ransomware Affiliates Used AI Agents to Execute Intrusions3d
- You gave your AI agent access, but who is giving it orders?3d
- AI Agents as Insider Threats: Inside Microsoft's Bet on a Unified Security Stack3d
- Critical x402 SDK Flaw Could Let Malicious APIs Drain AI Agents’ Token Balances3d