aiqbench analysis
This research introduces Robust Privacy (RP), a framework that leverages certified robustness to mask fine-grained signals at the inference stage, effectively countering model inversion attacks that are often mischaracterized as training-stage threats. By proving prediction invariance within a specific neighborhood, RP significantly reduces adversary advantage and lowers attack success rates while maintaining high model accuracy, outperforming standard techniques like DP-SGD in the privacy-utility tradeoff. The approach offers a direct method to mitigate side-channel leakage through the inference interface without sacrificing utility.
Robust Privacy: Inference-Stage Privacy through Certified Robustness
arXiv:2601.17360v2 Announce Type: replace Abstract: An adversary observing a model's released prediction can infer sensitive attributes of the queried input, or even reconstruct representatives of the model's training data. The inference interface thus acts as a side channel for privacy leakage. We introduce Robust Privacy (RP), an inference-stage privacy notion inspired by certified robustness: if a model's prediction is provably invariant within a radius-R neighborhood around an input x with confidence at least $1-\alpha$, then x enjoys $(R,\alpha)$-Robust Privacy, under which we prove that any adversary observing the released prediction has at most $\alpha/2$ advantage in distinguishing x from any input within distance R of x. Building on RP, we formalize Robust Attribute Privacy (RAP), an attribute-level privacy notion that characterizes the set of sensitive-attribute values that remain compatible with a released prediction. On a classification task, RP increases the median length of the RAP-compatible inference interval from 23.50 to 29.96, reducing attribute-inference precision. Model inversion attacks, often treated as a training-stage threat, in fact rely on fine-grained signals leaked through the inference interface; RP masks these signals at the inference stage, reducing attack success rate (ASR) from 73% to 4% on a black-box inversion attack. This direct targeting of the leakage channel enables RP to dominate DP-SGD and randomized response in the privacy-utility tradeoff space: RP retains 98.4% accuracy at 21% ASR, whereas DP-SGD must drop accuracy to 61.7% to reach a comparable ASR. Across both experiments, increasing the smoothing sample size N strengthens privacy and improves utility together. Finally, we examine model distillation as a scope boundary and show that RP mitigates attribute-level and instance-level inference-stage privacy leakage, but not function-level extraction through model distillation.
Abstract:An adversary observing a model's released prediction can infer sensitive attributes of the queried input, or even reconstruct representatives of the model's training data. The inference interface thus acts as a side channel for privacy leakage. We introduce Robust Privacy (RP), an inference-stage privacy notion inspired by certified robustness: if a model's prediction is provably invariant within a radius-R neighborhood around an input x with confidence at least $1-\alpha$, then x enjoys $(R,\alpha)$-Robust Privacy, under which we prove that any adversary observing the released prediction has at most $\alpha/2$ advantage in distinguishing x from any input within distance R of x. Building on RP, we formalize Robust Attribute Privacy (RAP), an attribute-level privacy notion that characterizes the set of sensitive-attribute values that remain compatible with a released prediction. On a classification task, RP increases the median length of the RAP-compatible inference interval from 23.50 to 29.96, reducing attribute-inference precision. Model inversion attacks, often treated as a training-stage threat, in fact rely on fine-grained signals leaked through the inference interface; RP masks these signals at the inference stage, reducing attack success rate (ASR) from 73% to 4% on a black-box inversion attack. This direct targeting of the leakage channel enables RP to dominate DP-SGD and randomized response in the privacy-utility tradeoff space: RP retains 98.4% accuracy at 21% ASR, whereas DP-SGD must drop accuracy to 61.7% to reach a comparable ASR. Across both experiments, increasing the smoothing sample size N strengthens privacy and improves utility together. Finally, we examine model distillation as a scope boundary and show that RP mitigates attribute-level and instance-level inference-stage privacy leakage, but not function-level extraction through model distillation.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
arXiv:2601.17360 [cs.LG]
(or arXiv:2601.17360v2 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2601.17360 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 371 words
Related coverage
Robust Privacy: Inference-Stage Privacy through Certified Robustness
arXiv · Aug 19, 2026
Vision-Language-Action Safety: Threats, Challenges, Evaluations, and Mechanisms
arXiv · Aug 26, 2026
NVIDIA Nemotron 3 Nano Omni Powers Multimodal Agent Reasoning in a Single Efficient Open Model
NVIDIA · Apr 28, 2026
[AINews] not much happened today
Latent Space · Sep 10, 2026