Research map / Time series / Causal and generative time series
Robust Multivariate Time-Series Forecasting: Adversarial Attacks and Defense Mechanisms
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Increase the number of indirectly perturbed series from κ=1 to κ=9 | Sparse indirect attack on DeepVAR · Traffic | wQL: 0.2190 -> 0.4635; 0.2428 -> 0.4075 | Mixed |
| Use probabilistic rather than deterministic sparse attack | Attack on DeepVAR · Traffic, κ=7 | wQL: 0.3620; 0.3529 | Mixed |
| Use probabilistic rather than deterministic sparse attack | Attack on DeepVAR · Electricity | wQL: 0.3410 -> 0.4364 | Hurt |
| Apply mini-max training instead of no defense | DeepVAR · Traffic, deterministic attack, κ=9 | wQL: 0.4635 -> 0.1970 | Helped |
| Apply mini-max training instead of no defense | DeepVAR · Traffic, probabilistic attack, κ=9 | wQL: 0.4075 -> 0.1911 | Helped |
| Apply mini-max training instead of no defense | DeepVAR · Electricity, probabilistic attack, κ=9 | wQL: 1.0317 -> 0.6466 | Helped |
| Apply mini-max training instead of no defense | DeepVAR · Taxi, probabilistic attack, κ=9 | wQL: 2.4815±0.9286 -> 1.4921±0.5551 | Helped |
| Apply randomized smoothing instead of no defense | Informer · Traffic, deterministic attack, κ=9, τ=24 | MSE: 1.3144 ± 0.0454 -> 0.8081 ± 0.0167 | Helped |
| Use σ=0.2 instead of σ=0.1 for data augmentation (ablation) | DeepVAR · Electricity, deterministic attack, κ=9 | wQL: 0.8412 -> 0.8243 | Helped |
| Use σ=0.3 instead of σ=0.1 for randomized smoothing (ablation) | DeepVAR · Electricity, deterministic attack, κ=9 | wQL: 0.8327 -> 0.7335 | Helped |
| Transfer a univariate DeepAR attack instead of using a multivariate attack | DeepVAR · Electricity | wQL: 0.322 -> 0.390 | Hurt |
Closest papers by meaning
- BackTime: Backdoor Attacks on Multivariate Time Series Forecasting (NeurIPS 2024 · similarity 0.88)
- Fortifying Time Series: DTW-Certified Robust Anomaly Detection (NeurIPS 2025 · similarity 0.85)
- Precise Accuracy / Robustness Tradeoffs in Regression: Case of General Norms (ICML 2024 · similarity 0.84)
- Clarifying the Behavior and the Difficulty of Adversarial Training (AAAI 2024 · similarity 0.84)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.84)
- Multivariate Probabilistic Time Series Forecasting with Correlated Errors (NeurIPS 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.