TSFMAudit detects pretraining contamination in time series foundation models via probe adaptation dynamics (faster loss drop, smaller backbone shift), tested on 6 models and 187 datasets against 10 LLM-derived baselines.
hub Canonical reference
author Stronati, M
Canonical reference. 83% of citing Pith papers cite this work as background.
28
Pith papers citing it
4,230
external citations · OpenAlex
Background
83% of classified citations
hub tools
citation-role summary
background 6
citation-polarity summary
roles
background 6representative citing papers
A framework proves that broad recalibrated leakage is undetectable from predictions alone without an external discrimination ceiling, while near-label leaks produce a detectable unit-purity signature yielding a prior-free test.
Logit-based federated learning leaks private model information to a semi-honest server via shared logits even with unrelated public data, enabling an adaptive stealing attack with theoretical bounds and a logit-perturbation defense.
Fair fine-tuning under Equalized Odds yields a tight bound Adv(A, M_f) ≤ Δ_EO · W on adversarial advantage in distribution inference attacks, with empirical reductions below detection threshold across six datasets.
A cross-modal alignment attack achieves AUC 0.821 for single-sample black-box membership inference on VLMs such as LLaVA-1.5 by quantifying image-generated caption similarity.
A new 507-leaf taxonomy and 4x6 Target x Technique matrix audits six LLM attack benchmarks and finds they cover at most 25% of the threat surface with entire STRIDE categories untested.
PACZero achieves zero mutual information privacy in LLM fine-tuning via sign-quantized subset-aggregated ZO gradients, delivering near non-private accuracy on SST-2 at I=0.
LLM tabular generators leak memorized numeric strings, allowing a no-box attack to achieve near-perfect membership inference on some state-of-the-art models.
Image autoregressive models leak substantially more training data than diffusion models under membership inference, dataset inference with as few as 4 samples, and data extraction attacks.
A malicious FL server can steal private training images by encoding them into model parameters via a correlation regularizer and preserving them through segmented aggregation.
TL++ recovers centralized mini-batch gradients via virtual batches in split learning and adds secret sharing for cut-layer tensors, achieving 91.41% accuracy on CIFAR-10 with 13x lower communication than full-model sync.
Introduces natural identifiers (NIDs) from common training data to support post-hoc differential privacy auditing and dataset inference for LLMs without retraining or private held-out sets.
EASE closes three residual anchors in federated multimodal unlearning using bilateral displacement, cosine-sine decomposition, and forget lock, achieving near-retrain performance on forget and retain data.
A separable expert architecture uses base models, LoRA adapters, and deletable per-user proxies to enable privacy-preserving personalization and deterministic unlearning in LLMs.
A context-aware Sentinel-Strategist system for RAG selectively applies defenses to block membership inference and data poisoning while recovering most retrieval utility compared to always-on defense stacks.
LLMs exhibit 20-40% lower recall on ambiguous human names for PII detection, worsening under prompt injections, as shown via the new AmBench benchmark.
PTMC is a proposed Monte Carlo estimator that generates market-outcome distributions by simulating continuous double-auction interactions among persona-conditioned neural-policy bots whose heterogeneity is drawn from a learned distribution.
Deterministic gradient-norm bounds in variational QML control DP-SGD clipping bias, so quantum models retain more accuracy than matched classical models under the same privacy budget.
PRISM is a gauge-invariant DP mechanism for LoRA that avoids bilinear noise amplification via tangent-space sampling, supplies a closed-form noise characterization on Z, and includes a DP-aware adaptive update rule.
The PROMPT framework formalizes privacy risks and mitigations in propaganda detection with a utility function and compliance score, backed by analysis of 162 studies and experiments showing 1-14% F1 drops under synthetic perturbations.
Stacking seven black-box estimators into a meta-classifier reveals persistent membership leakage in differentially private federated learning models at epsilon=200 on NIST genomics data, outperforming single-signal baselines.
A responsible computing framework substitutes real protest imagery with labeled synthetic reproductions from conditional image synthesis to enable privacy-aware analysis of collective action patterns.
ICA and VEIL enable privacy-preserving supervised ML by producing structurally non-invertible encodings aligned with downstream tasks while maintaining predictive utility.
ALPINE deploys an offline-trained TD3 policy on terminal devices to map multi-dimensional risk states to adaptive privacy budgets for local differential privacy in mobile edge crowdsensing, with edge feedback closing the loop.
citing papers explorer
-
TSFMAudit: Data Contamination Auditing in Forecasting Time Series Foundation Models
TSFMAudit detects pretraining contamination in time series foundation models via probe adaptation dynamics (faster loss drop, smaller backbone shift), tested on 6 models and 187 datasets against 10 LLM-derived baselines.
-
A prior-free blind detection of information leakage from model predictions
A framework proves that broad recalibrated leakage is undetectable from predictions alone without an external discrimination ceiling, while near-label leaks produce a detectable unit-purity signature yielding a prior-free test.
-
Quantifying and Defending against the Privacy Risk in Logit-based Federated Learning
Logit-based federated learning leaks private model information to a semi-honest server via shared logits even with unrelated public data, enabling an adaptive stealing attack with theoretical bounds and a logit-perturbation defense.
-
Fair Finetuning Mitigates Distribution Inference Attacks
Fair fine-tuning under Equalized Odds yields a tight bound Adv(A, M_f) ≤ Δ_EO · W on adversarial advantage in distribution inference attacks, with empirical reductions below detection threshold across six datasets.
-
Single-Sample Black-Box Membership Inference Attack against Vision-Language Models via Cross-modal Semantic Alignment
A cross-modal alignment attack achieves AUC 0.821 for single-sample black-box membership inference on VLMs such as LLaVA-1.5 by quantifying image-generated caption similarity.
-
Talk is (Not) Cheap: A Taxonomy and Benchmark Coverage Audit for LLM Attacks
A new 507-leaf taxonomy and 4x6 Target x Technique matrix audits six LLM attack benchmarks and finds they cover at most 25% of the threat surface with entire STRIDE categories untested.
-
PACZero: PAC-Private Fine-Tuning of Language Models via Sign Quantization
PACZero achieves zero mutual information privacy in LLM fine-tuning via sign-quantized subset-aggregated ZO gradients, delivering near non-private accuracy on SST-2 at I=0.
-
When Tables Leak: Attacking String Memorization in LLM-Based Tabular Data Generation
LLM tabular generators leak memorized numeric strings, allowing a no-box attack to achieve near-perfect membership inference on some state-of-the-art models.
-
Privacy Attacks on Image AutoRegressive Models
Image autoregressive models leak substantially more training data than diffusion models under membership inference, dataset inference with as few as 4 samples, and data extraction attacks.
-
FedCVESA: Taking Away Training Data in Federated Learning via Correlation Value Encoding and Segmented Aggregation
A malicious FL server can steal private training images by encoding them into model parameters via a correlation regularizer and preserving them through segmented aggregation.
-
TL++: Accuracy and Privacy Preserving Traversal Learning for Distributed Intelligent Systems
TL++ recovers centralized mini-batch gradients via virtual batches in split learning and adds secret sharing for cut-layer tensors, achieving 91.41% accuracy on CIFAR-10 with 13x lower communication than full-model sync.
-
Natural Identifiers for Privacy and Data Audits in Large Language Models
Introduces natural identifiers (NIDs) from common training data to support post-hoc differential privacy auditing and dataset inference for LLMs without retraining or private held-out sets.
-
EASE: Federated Multimodal Unlearning via Entanglement-Aware Anchor Closure
EASE closes three residual anchors in federated multimodal unlearning using bilateral displacement, cosine-sine decomposition, and forget lock, achieving near-retrain performance on forget and retain data.
-
Separable Expert Architecture: Toward Privacy-Preserving LLM Personalization via Composable Adapters and Deletable User Proxies
A separable expert architecture uses base models, LoRA adapters, and deletable per-user proxies to enable privacy-preserving personalization and deterministic unlearning in LLMs.
-
Adaptive Defense Orchestration for RAG: A Sentinel-Strategist Architecture against Multi-Vector Attacks
A context-aware Sentinel-Strategist system for RAG selectively applies defenses to block membership inference and data poisoning while recovering most retrieval utility compared to always-on defense stacks.
-
Can Large Language Models Really Recognize Your Name?
LLMs exhibit 20-40% lower recall on ambiguous human names for PII detection, worsening under prompt injections, as shown via the new AmBench benchmark.
-
Persona-Trained Monte Carlo: Estimating Market-Outcome Distributions via Swarms of Persona-Conditioned Neural Policy Bots in a Limit Order Book
PTMC is a proposed Monte Carlo estimator that generates market-outcome distributions by simulating continuous double-auction interactions among persona-conditioned neural-policy bots whose heterogeneity is drawn from a learned distribution.
-
Private training in quantum machine learning
Deterministic gradient-norm bounds in variational QML control DP-SGD clipping bias, so quantum models retain more accuracy than matched classical models under the same privacy budget.
-
PRISM: Gauge-Invariant Tangent-Space Differentially Private LoRA
PRISM is a gauge-invariant DP mechanism for LoRA that avoids bilinear noise amplification via tangent-space sampling, supplies a closed-form noise characterization on Z, and includes a DP-aware adaptive update rule.
-
SoK: Analysis of Privacy Risks and Mitigation in Online Propaganda Detection through the PROMPT Framework
The PROMPT framework formalizes privacy risks and mitigations in propaganda detection with a utility function and compliance score, backed by analysis of 162 studies and experiments showing 1-14% F1 drops under synthetic perturbations.
-
Evaluating Differential Privacy Against Membership Inference in Federated Learning: Insights from the NIST Genomics Red Team Challenge
Stacking seven black-box estimators into a meta-classifier reveals persistent membership leakage in differentially private federated learning models at epsilon=200 on NIST genomics data, outperforming single-signal baselines.
-
Protecting and Preserving Protest Dynamics for Responsible Analysis
A responsible computing framework substitutes real protest imagery with labeled synthetic reproductions from conditional image synthesis to enable privacy-aware analysis of collective action patterns.
-
Informationally Compressive Anonymization: Non-Degrading Sensitive Input Protection for Privacy-Preserving Supervised Machine Learning
ICA and VEIL enable privacy-preserving supervised ML by producing structurally non-invertible encodings aligned with downstream tasks while maintaining predictive utility.
-
ALPINE: Closed-Loop Adaptive Privacy Budget Allocation for Mobile Edge Crowdsensing
ALPINE deploys an offline-trained TD3 policy on terminal devices to map multi-dimensional risk states to adaptive privacy budgets for local differential privacy in mobile edge crowdsensing, with edge feedback closing the loop.
-
Chernoff Information as a Privacy Constraint for Adversarial Classification and Membership Advantage
Chernoff DP is sandwiched between KL DP and ε-DP, outperforms KL in numerical Laplace-mechanism tests, and yields a new upper bound on adversary membership advantage compared with (ε,δ)-DP bounds.
-
Towards the Anonymization of the Language Modeling
Authors introduce MLM and CLM specialization methods that avoid memorizing identifiers in sensitive training data while aiming for a privacy-utility tradeoff on medical datasets.
-
Nonlinear Transformations Against Unlearnable Datasets
Nonlinear transformations enable DNNs to achieve substantial test accuracy gains (0.34% to 249.59%) on unlearnable CIFAR10 datasets from twelve protection methods, outperforming a recent linear baseline.
-
Efficient privacy preservation of big data for accurate data mining
PABIDOT is a new perturbation algorithm for privacy-preserving big data classification that claims superior execution speed, scalability, attack resistance, and accuracy versus two related methods across nine datasets and five classifiers.