Research map / Adversarial robustness and security / Certified and empirical robustness
Adversarial Attacks on Probabilistic Autoregressive Forecasting Models
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 |
|---|---|---|---|
| Use the reparametrization estimator instead of the score-function estimator | Gradient-based attacks on probabilistic autoregressive forecasting models · S&P 500 buy/sell classification | attack success rate: 16% more cases | Helped |
| Use the reparametrization estimator instead of the score-function estimator | Gradient-based attacks on probabilistic autoregressive forecasting models · S&P 500 long-short trading | financial gain of attacked portfolio | Helped |
| Use the reparametrization estimator instead of the score-function estimator | DeepAR · UCI electricity consumption prediction | under-estimation attack strength: ε = 1.0; 20%; 70% | Helped |
| Apply the reparametrization estimator in the Bayesian setting | Probabilistic autoregressive forecasting model · S&P 500 buy/sell classification with observation y10/x−1 = γ | attack success rate: approximately 80% for ε = 0.1 | Helped |
| Increase the perturbation tolerance | Reparametrization attack on electricity consumption · UCI electricity consumption prediction | median estimation error: ε = 0.8; 15%; 20% | Mixed |
| Use 10000 rather than 1 Monte-Carlo sample for forecast estimation (ablation) | Probabilistic LSTM · S&P 500 cumulated return forecast (h=10) | Ranked Probability Skill: 1.995 ($\pm$ 0.011) -> 1.002 ($\pm$ 0.008) | Helped |
| Use 10000 rather than 100 Monte-Carlo samples for forecast estimation (ablation) | Probabilistic LSTM · S&P 500 cumulated return forecast (h=10) | Ranked Probability Skill: 1.012 ($\pm$ 0.008) -> 1.002 ($\pm$ 0.008) | Helped |
Closest papers by meaning
- Robust Multivariate Time-Series Forecasting: Adversarial Attacks and Defense Mechanisms (ICLR 2023 · similarity 0.90)
- Data Poisoning Attacks against Autoregressive Models (AAAI 2016 · similarity 0.89)
- Optimal Attack against Autoregressive Models by Manipulating the Environment (AAAI 2020 · similarity 0.86)
- Distributional Adversarial Attacks and Training in Deep Hedging (NeurIPS 2025 · similarity 0.85)
- BackTime: Backdoor Attacks on Multivariate Time Series Forecasting (NeurIPS 2024 · similarity 0.83)
- Probabilistic Forecasting with Stochastic Interpolants and Föllmer Processes (ICML 2024 · similarity 0.82)
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.