AVEC: Bootstrapping Privacy for Local LLMs
This is a conceptual and theoretical contribution that addresses privacy concerns for users of local LLMs, but it is incremental as it builds on existing privacy techniques without empirical deployment.
The paper tackles the problem of ensuring privacy for local language models by proposing AVEC, a framework that uses adaptive differential privacy budgeting and verifiable transformations at the edge, establishing theoretical guarantees such as utility ceilings and impossibility results.
This position paper presents AVEC (Adaptive Verifiable Edge Control), a framework for bootstrapping privacy for local language models by enforcing privacy at the edge with explicit verifiability for delegated queries. AVEC introduces an adaptive budgeting algorithm that allocates per-query differential privacy parameters based on sensitivity, local confidence, and historical usage, and uses verifiable transformation with on-device integrity checks. We formalize guarantees using Rényi differential privacy with odometer-based accounting, and establish utility ceilings, delegation-leakage bounds, and impossibility results for deterministic gating and hash-only certification. Our evaluation is simulation-based by design to study mechanism behavior and accounting; we do not claim deployment readiness or task-level utility with live LLMs. The contribution is a conceptual architecture and theoretical foundation that chart a pathway for empirical follow-up on privately bootstrapping local LLMs.