Are Time-Series Foundation Models Deployment-Ready? A Systematic Study of Adversarial Robustness Across Domains
Not on the map. Why?
Abstract (arXiv)
Time-Series Foundation Models (TSFMs) are rapidly transitioning from research prototypes to core components of critical decision-making systems, driven by their impressive zero-shot forecasting capabilities. However, as their deployment surges, a critical blind spot remains: their fragility under adversarial attacks. This lack of scrutiny poses severe risks, particularly as TSFMs enter high-stakes environments vulnerable to manipulation. We present a systematic, diagnostic study arguing that for TSFMs, robustness is not merely a secondary metric but a prerequisite for trustworthy deployment comparable to accuracy. Our evaluation framework, explicitly tailored to the unique constraints of time series, incorporates normalized, sparsity-aware perturbation budgets and unified scale-invariant metrics across white-box and black-box settings. Across six representative TSFMs, we demonstrate that current architectures are alarmingly brittle: even small perturbations can reliably steer forecasts toward specific failure modes, such as trend flips and malicious drifts. We uncover TSFM-specific vulnerability patterns, including horizon-proximal brittleness, increased susceptibility with longer context windows, and weak cross-model transfer that points to model-specific failure modes rather than generic distortions. Finally, we show that simple adversarial fine-tuning offers a cost-effective path to substantial robustness gains, even with out-of-domain data. This work bridges the gap between TSFM capabilities and safety constraints, offering essential guidance for hardening the next generation of forecasting systems.
Key facts. Cited by 2 papers in the map (data through 13 Oct 2026). Builds on 44 of its 67 references that are in the map. First posted on arXiv 26 May 2025.
Builds on
44 of 67 references.
References · 44
- 2020Language Models are Few-Shot Learnerscited by 32,320
- 2024The Llama 3 Herd of Modelscited by 14,486
- 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 12,979
- 2023Segment Anythingcited by 8,195
- 2014Explaining and Harnessing Adversarial Examplescited by 7,336
- 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,379
- 2013Intriguing properties of neural networkscited by 5,137
- 2015DeepFool: a simple and accurate method to fool deep neural networkscited by 1,316
- 2018Obfuscated Gradients Give a False Sense of Security: Circumventing Defenses to Adversarial Examplescited by 1,265
- 2022Red Teaming Language Models with Language Modelscited by 1,014
- 2019Adversarial Examples Are Not Bugs, They Are Featurescited by 953
- 2024Chronos: Learning the Language of Time Seriescited by 737
32 more
- 2017ZOO: Zeroth Order Optimization based Black-box Attacks to Deep Neural Networks without Training Substitute Modelscited by 663
- 2020RobustBench: a standardized adversarial robustness benchmarkcited by 658
- 2024Unified Training of Universal Time Series Forecasting Transformerscited by 602
- 2024MOMENT: A Family of Open Time-series Foundation Modelscited by 468
- 2019On Evaluating Adversarial Robustnesscited by 447
- 2021Monash Time Series Forecasting Archivecited by 261
- 2019Simple Black-box Adversarial Attackscited by 250
- 2024Foundation Models for Time Series Analysis: A Tutorial and Surveycited by 237
- 2024GIFT-Eval: A Benchmark For General Time Series Forecasting Model Evaluationcited by 201
- 2023Survey of Vulnerabilities in Large Language Models Revealed by Adversarial Attackscited by 192
- 2023Large Language Models for Code: Security Hardening and Adversarial Testingcited by 153
- 2023On the Adversarial Robustness of Multi-Modal Foundation Modelscited by 145
- 2018Is Robustness the Cost of Accuracy? -- A Comprehensive Study on the Robustness of 18 Deep Image Classification Modelscited by 135
- 2021Measure and Improve Robustness in NLP Models: A Surveycited by 107
- 2020InfoBERT: Improving Robustness of Language Models from An Information Theoretic Perspectivecited by 92
- 2024Defending Against Unforeseen Failure Modes with Latent Adversarial Trainingcited by 85
- 2021Time Series Prediction using Deep Learning Methods in Healthcarecited by 73
- 2023How Deep Learning Sees the World: A Survey on Adversarial Attacks & Defensescited by 65
- 2019Adversarial Attacks on Deep Neural Networks for Time Series Classificationcited by 53
- 2021How Should Pre-Trained Language Models Be Fine-Tuned Towards Adversarial Robustness?cited by 49
- 2022Practical Evaluation of Adversarial Robustness via Adaptive Auto Attackcited by 44
- 2019Adversarial Attacks on Time Seriescited by 40
- 2024TimeDiT: General-purpose Diffusion Transformers for Time Series Foundation Modelcited by 33
- 2023Large Pre-trained time series models for cross-domain Time series analysis taskscited by 22
- 2021Untargeted, Targeted and Universal Adversarial Attacks and Defenses on Time Seriescited by 21
- 2020Adversarial Attacks on Probabilistic Autoregressive Forecasting Modelscited by 13
- 2022Fast and Reliable Evaluation of Adversarial Robustness with Minimum-Margin Attackcited by 13
- 2025TimesBERT: A BERT-Style Foundation Model for Time Series Understandingcited by 10
- 2020Benchmarking adversarial attacks and defenses for time-series datacited by 9
- 2024Adversarial Vulnerabilities in Large Language Models for Time Series Forecastingcited by 6
- 2022Similarity of Neural Architectures using Adversarial Attack Transferabilitycited by 4
- 2024Enhancing adversarial robustness in Natural Language Inference using explanationscited by 2
Built on by
Citing papers per year: 2025: 1; 2026: 1.
2 papers.