Skip to main content
Search

# Differential Privacy

15 related articles

Subscribe to updates
  1. Phones and cloud split generative inference: MAGS bids instead of stampeding, cutting p99 latency 17.1%

    A generative query's cost is unknown until the answer is written, so devices reserve resources blind and many crowd one server at once. MAGS has each device forecast token count and generation entropy with a four-layer probe on its local model, then bid for cloud capacity against a broadcast congestion price, choosing its per-token privacy parameter in the same bid. Over ten seeds it cuts p99 latency 17.1%, energy per token 13.6% and raises throughput 20.7% against a reactive PPO baseline given identical observations; an eight-device, two-server testbed drops the cross-device correlation of offload decisions from 0.41 to 0.08.

  2. arXiv

    Halfspace learning gets differential privacy for free: in three agnostic settings the privacy cost folds into the statistical term

    For agnostically learning halfspaces with arbitrary labels under Gaussian marginals, computationally efficient differentially private algorithms were previously missing. This work reports efficient private learners in three settings, improper, proper and semi-agnostic; when the privacy parameter is at least the accuracy parameter, the extra sample cost of privacy is absorbed by each setting's non-private statistical term, making privacy free in sample complexity.

  3. arXiv

    GraphRAG relations get a privacy budget: RelGuard cuts relation-recovery F1 by 26.98 to 44.35 points

    GraphRAG links information across documents through entity relations, but black-box queries can also be used to reconstruct the knowledge graph, leaking whether entities are connected and what those relations mean. RelGuard rewrites source documents before indexing: it generates candidate relations from entity types, then uses an exponential mechanism to sample a replacement predicate or delete each relation, with a relation-level differential privacy guarantee. Across LightRAG, PathRAG and HippoRAG 2, two datasets and two reconstruction attacks, its relation-recovery F1 is 26.98 to 44.35 percentage points below the strongest baseline, while the extra question-answering F1 loss is 0.59 to 4.78 points in most settings.

  4. arXiv

    Verifiable-reward RL gets its first prompt-level privacy guarantee, and the reward signal still lifts accuracy through the noise

    Reinforcement learning with verifiable rewards (RLVR) trains a language model on problems that may themselves be confidential, and the released weights can reveal which problems it saw. DP-GRPO aggregates all responses to one prompt into a single gradient, clips it once and adds Gaussian noise, so the privacy budget depends on neither the number of responses per prompt nor the clipping norm; the authors describe it as the first differential privacy guarantee for RLVR training. Training Qwen2.5-1.5B-Instruct with LoRA at a per-run budget of 8, the reward signal beats a control that removes only the reward by 2.65 points on MATH and 3.24 on GSM8K, positive in every seed.

  5. Synthetic health records can now carry a privacy guarantee: attack detection near chance at ε=8, clinical validity 0.961

    Synthetic electronic health records have long traded clinical fidelity against formal privacy guarantees. SynthEHR-DP combines a mixed-type encoder, a temporal transformer encoder and a conditional variational autoencoder, trained under differentially private stochastic gradient descent with Rényi privacy accounting, then applies a post-hoc clinical constraint layer that spends no privacy budget. On MIMIC-IV at a total budget of ε=8 and δ=10⁻⁵, membership inference AUC is 0.52, downstream TSTR AUROC is 0.786, 4.2% below the non-private TRTR oracle of 0.821, and the clinical validity score is 0.961.

  6. Private federated learning stops uploading whole updates: filtering shareable sport semantics per network stage lifts accuracy about four points under strong privacy

    Sports images sit with schools, clubs and other organizations, and federated learning avoids sending raw pictures but still transmits model updates that can leak information, while the clipping and noise of differential privacy damage useful knowledge. FedSS scores each network stage for how shareable it is before clipping, uploads only the perturbed cross-client sport semantics, and keeps client-specific appearance and background context local. On sport22 and sport100 it beats the strongest baseline DPSIR-Fed by 5.78 and 5.71 points under visual-context heterogeneity, and raises sport22 accuracy from 42.67% to 46.81% at the strong-privacy setting ε=4.

  7. Noise-only distillation is no sanitizer: a shared public checkpoint lets a student inherit its teacher's memorized labels and backdoors

    Distilling a teacher on pure Gaussian noise still hands the student the teacher's memorized training records and planted backdoors whenever both start from the same public initialization. On MNIST, distilling only never-trained auxiliary outputs yields a membership-inference AUC of 0.86 and a 93% backdoor attack-success lift, while an independently initialized student stays near chance. In a fine-tuning threat model, an attacker with the public base but no data recovers the memorized labels of 93% of private canaries from noise queries alone.

  8. IoT devices stop trusting a one-time credential: behavioral fingerprints plus federated learning cut membership inference success from 48.2% to 12.4%

    IoT device authentication is shifting from a one-time check of fixed credentials to continuous checking of device behavior. Static credentials and centralized trust miss anomalies after a device is compromised and keep exposing identifiable attributes. A combined scheme uses hardware behavioral fingerprints, federated learning with differential privacy, a lightweight permissioned chain and agentic orchestration; on 11 Raspberry Pi devices in the PP-IDE dataset, differential privacy lowered membership inference attack success from 48.2% to 12.4%, with 99.2% authentication accuracy.

  9. Knee osteoarthritis grading without pooled records: ten sites hold feature columns, asynchronous federation reports 99.64% accuracy

    Knee osteoarthritis diagnosis draws on multimodal data scattered across institutions, and centralizing it is both costly and privacy-risky. An asynchronous federated scheme first fuses structured records, semi-structured files and imaging into a queryable data table, then has ten simulated institutions compute gradients on vertically partitioned feature columns, with Bulyan aggregation and adaptive differential privacy filtering malicious updates on the server. On 5,580 knee osteoarthritis cases it reports 99.64% accuracy, 0.998 AUC, 285 MB communication overhead and a 1.8% attack success rate.

  10. arXiv

    BRACE gives LSR dense associative memory differentially private retrieval, cutting MNIST prediction error from 0.061 to 0.011

    The authors propose BRACE, the first differentially private retrieval mechanism for log-sum-ReLU dense associative memory (LSR-DAM): it identifies memory points whose membership changes under privacy noise and applies an adaptive boundary correction before injecting calibrated noise, controlling error accumulation from boundary switching in compact-support retrieval. They prove minimax-optimal terminal and full-trajectory retrieval error rates and establish trajectory central limit theorems characterizing privacy-induced variability; synthetic experiments show a tiny perturbation changes 11 gate memberships, and MNIST retrieval reduces prediction MSE from 0.061 for the corrupted input to 0.011, nearly matching noiseless retrieval.