ML Security PapersBeyond Immediate Activation: Temporally Decoupled Backdoor Attacks on Time Series Forecasting
Zhixin Liu , Xuanlin Liu , Sihan Xu , Yaqiong Qiao , Ying Zhang , Xiangrui Cai
Published on arXiv
2601.04247
Model Poisoning
OWASP ML Top 10 — ML10
Key Finding
TDBA consistently outperforms existing backdoor attack baselines in attack success rate while maintaining stealthiness across real-world multivariate time series datasets.
TDBA (Temporally Decoupled Backdoor Attack)
Novel technique introduced
Existing backdoor attacks on multivariate time series (MTS) forecasting enforce strict temporal and dimensional coupling between triggers and target patterns, requiring synchronous activation at fixed positions across variables. However, realistic scenarios often demand delayed and variable-specific activation. We identify this critical unmet need and propose TDBA, a temporally decoupled backdoor attack framework for MTS forecasting. By injecting triggers that encode the expected location of the target pattern, TDBA enables the activation of the target pattern at any positions within the forecasted data, with the activation position flexibly varying across different variable dimensions. TDBA introduces two core modules: (1) a position-guided trigger generation mechanism that leverages smoothed Gaussian priors to generate triggers that are position-related to the predefined target pattern; and (2) a position-aware optimization module that assigns soft weights based on trigger completeness, pattern coverage, and temporal offset, facilitating targeted and stealthy attack optimization. Extensive experiments on real-world datasets show that TDBA consistently outperforms existing baselines in effectiveness while maintaining good stealthiness. Ablation studies confirm the controllability and robustness of its design.
Key Contributions
- Identifies the rigid temporal and dimensional coupling limitation of existing MTS backdoor attacks (e.g., BackTime) and formalizes the temporally decoupled threat scenario.
- Proposes a position-guided trigger generation mechanism using smoothed Gaussian priors that encodes the expected activation location of the target pattern within the trigger itself.
- Introduces a position-aware optimization module with soft weighting based on trigger completeness, pattern coverage, and temporal offset to improve both attack effectiveness and stealthiness.
🛡️ Threat Analysis
TDBA injects hidden backdoor triggers into training data that activate predefined malicious target patterns in forecasts only when the trigger is present — classic backdoor/trojan insertion with novel temporally decoupled activation mechanics.