The paper investigates how different types of personally identifiable information (PII) are learned and leaked by large language models for code (LLM4Code). By creating a diverse PII dataset, fine‑tuning models of various scales, and analyzing training dynamics through a structural causal model, the authors find that leakage risk varies by PII type—easy‑to‑learn data such as IP addresses leak more often, while harder types like keys and passwords leak less. The study provides causal evidence that leakage risks are type‑dependent and suggests that defenses should be tailored to specific PII categories and their learnability.
arXiv:2606. 16952v1 Announce Type: cross Abstract: The rapid adoption of generative AI and Large Language Models (LLMs) has spurred interest in synthetic data as a privacy-preserving alternative to sensitive real-world datasets.
arXiv:2606. 16952v2 Announce Type: replace-cross Abstract: The rapid adoption of generative AI and Large Language Models (LLMs) has spurred interest in synthetic data as a privacy-preserving alternative to sensitive real-world datasets.
arXiv:2606. 10481v1 Announce Type: cross Abstract: Parameter-efficient fine-tuning of large language models (LLMs) can exhibit problematic memorization of individual training examples.
arXiv:2606. 09401v1 Announce Type: new Abstract: Recent work has applied differential privacy (DP) to adapt large language models (LLMs) for sensitive applications, offering theoretical guarantees.
arXiv:2607. 15970v1 Announce Type: cross Abstract: The flourishing code hosting platforms and coding agents enable even beginners with private data to build tailored Machine Learning (ML) models using available code quickly.
The paper examines how supervised fine-tuning (SFT) of large language models can leak personally identifiable information (PII) when the fine-tuning data contains user-provided sensitive details. It introduces COVA, a coverage-aware decoding algorithm that improves targeted PII reconstruction from SFT models, especially when an adversary has limited contextual knowledge about a target. Experiments on medical and legal Q&A datasets show that even small proprietary SFT datasets can lead to significant privacy leakage via PII reconstruction.
arXiv:2607. 22695v1 Announce Type: new Abstract: Large Language Models (LLMs) are capable of generalizing human language for the completion of never-before-seen tasks, leading to widespread deployment.
Conformal Privacy Auditing (CPA) is a distribution‑free framework that calibrates re‑identification risk for each released document against large language model (LLM)‑empowered adversaries. It outputs a conformal ambiguity set of candidate identities that is guaranteed to contain the true identity with a user‑chosen confidence level under exchangeability, along with an interpretable leakage proxy derived from the set size. CPA supports both logit‑access and sampling‑only attackers, enabling audits of both open‑source and proprietary models, and demonstrates calibrated coverage across various benchmarks and attacker configurations.
arXiv:2506. 14003v5 Announce Type: replace Abstract: Machine unlearning (MU) for large language models (LLMs), commonly referred to as LLM unlearning, seeks to remove specific undesirable data or knowledge from a trained model, while maintaining its performance on standard tasks.
InferPriv is a new benchmark that evaluates inferential privacy leakage in large language model (LLM) assistants. It measures whether an honest-but-curious recipient can deduce a user’s sensitive attributes from an LLM’s reply during routine tasks, a type of leakage not captured by existing direct‑leakage benchmarks. Testing five frontier models, the study finds that inferential leakage rates (16.2–25.2%) far exceed direct leakage, even when direct leakage is near zero under privacy‑aware prompts.
arXiv:2606. 17110v1 Announce Type: cross Abstract: Large Language Models are increasingly trained on proprietary or sensitive data, from private healthcare and financial records to user conversations containing secrets.
Privacy risks in text-only Large Language Models (LLMs) are well studied, particularly their tendency to memorize and leak sensitive information. However, Multi-modal Large Language Models (MLLMs), which process both text and images, introduce unique privacy challenges that remain underexplored.