ML Security PapersTemporally Unified Adversarial Perturbations for Time Series Forecasting
Ruixian Su , Yukun Bao , Xinze Zhang
Published on arXiv
2602.11940
Input Manipulation Attack
OWASP ML Top 10 — ML01
Key Finding
TGAM with TUAP constraints significantly outperforms baselines in both white-box and black-box transfer attack scenarios across three benchmark datasets and four state-of-the-art forecasting models, and also shows superior transfer attack performance without TUAP constraints.
TGAM (Timestamp-wise Gradient Accumulation Method) / TUAP (Temporally Unified Adversarial Perturbations)
Novel technique introduced
While deep learning models have achieved remarkable success in time series forecasting, their vulnerability to adversarial examples remains a critical security concern. However, existing attack methods in the forecasting field typically ignore the temporal consistency inherent in time series data, leading to divergent and contradictory perturbation values for the same timestamp across overlapping samples. This temporally inconsistent perturbations problem renders adversarial attacks impractical for real-world data manipulation. To address this, we introduce Temporally Unified Adversarial Perturbations (TUAPs), which enforce a temporal unification constraint to ensure identical perturbations for each timestamp across all overlapping samples. Moreover, we propose a novel Timestamp-wise Gradient Accumulation Method (TGAM) that provides a modular and efficient approach to effectively generate TUAPs by aggregating local gradient information from overlapping samples. By integrating TGAM with momentum-based attack algorithms, we ensure strict temporal consistency while fully utilizing series-level gradient information to explore the adversarial perturbation space. Comprehensive experiments on three benchmark datasets and four representative state-of-the-art models demonstrate that our proposed method significantly outperforms baselines in both white-box and black-box transfer attack scenarios under TUAP constraints. Moreover, our method also exhibits superior transfer attack performance even without TUAP constraints, demonstrating its effectiveness and superiority in generating adversarial perturbations for time series forecasting models.
Key Contributions
- Identifies the temporally inconsistent perturbations problem in adversarial attacks on time series forecasting, where overlapping samples receive contradictory perturbation values for shared timestamps
- Proposes Temporally Unified Adversarial Perturbations (TUAPs) that enforce a temporal unification constraint ensuring identical perturbations per timestamp across all overlapping samples
- Introduces Timestamp-wise Gradient Accumulation Method (TGAM), a modular gradient aggregation approach compatible with momentum-based attacks that achieves superior white-box and black-box transfer attack performance
🛡️ Threat Analysis
Proposes TUAP and TGAM — gradient-based adversarial perturbation attack methods targeting deep learning time series forecasting models at inference time — in both white-box and black-box transfer settings. The primary contribution is a novel attack technique that enforces temporal consistency, which is a direct advancement in adversarial example methodology.