mlsanityData to 7 Oct 2026

On the Fragility of AI-Based Channel Decoders under Small Channel Perturbations

arXiv 2602.01582 ↗First posted 2 Feb 2026cs.IT cs.AI cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

Recent advances in deep learning have led to AI-based error correction decoders that report empirical performance improvements over traditional belief-propagation (BP) decoding on AWGN channels. While such gains are promising, a fundamental question remains: where do these improvements come from, and what cost is paid to achieve them? In this work, we study this question through the lens of robustness to distributional shifts at the channel output. We evaluate both input-dependent adversarial perturbations (FGM and projected gradient methods under ℓ2 constraints) and universal adversarial perturbations that apply a single norm-bounded shift to all received vectors. Our results show that recent AI decoders, including ECCT and CrossMPT, could suffer significant performance degradation under such perturbations, despite superior nominal performance under i.i.d. AWGN. Moreover, adversarial perturbations transfer relatively strongly between AI decoders but weakly to BP-based decoders, and universal perturbations are substantially more harmful than random perturbations of equal norm. These numerical findings suggest a potential robustness cost and higher sensitivity to channel distribution underlying recent AI decoding gains.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 27 of its 47 references that are in the map. First posted on arXiv 2 Feb 2026.

Builds on

27 of 47 references.

References · 27

  1. 2014Explaining and Harnessing Adversarial Examplescited by 7,287
  2. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,326
  3. 2019Certified Adversarial Robustness via Randomized Smoothingcited by 1,324
  4. 2016Universal adversarial perturbationscited by 774
  5. 2017Adversarial Examples: Attacks and Defenses for Deep Learningcited by 350
  6. 2015Data-driven Distributionally Robust Optimization Using the Wasserstein Metric: Performance Guarantees and Tractable Reformulationscited by 285
  7. 2018Generalizable Adversarial Training via Spectral Normalizationcited by 68
  8. 2016Learning to Decode Linear Codes Using Deep Learningcited by 60
  9. 2022Error Correction Code Transformercited by 47
  10. 2020Train simultaneously, generalize better: Stability of gradient-based minimax learnerscited by 43
  11. 2019Turbo Autoencoder: Deep learning based channel codes for point-to-point communication channelscited by 42
  12. 2018Deepcode: Feedback Codes via Deep Learningcited by 41
15 more
  1. 2018Deep Learning for Decoding of Linear Codes - A Syndrome-Based Approachcited by 34
  2. 2022Denoising Diffusion Error Correction Codescited by 23
  3. 2024CrossMPT: Cross-attention Message-Passing Transformer for Error Correcting Codescited by 20
  4. 2019Physical Adversarial Attacks Against End-to-End Autoencoder Communication Systemscited by 19
  5. 2021Robust Adversarial Attacks Against DNN-Based Wireless Communication Systemscited by 12
  6. 2019MIND: Model Independent Neural Decodercited by 11
  7. 2022On the Role of Generalization in Transferability of Adversarial Examplescited by 9
  8. 2022On Convergence of Gradient Descent Ascent: A Tight Local Analysiscited by 9
  9. 2024Structured Gradient-based Interpretations via Norm-Regularized Adversarial Trainingcited by 7
  10. 2023MoreauGrad: Sparse and Robust Interpretation of Neural Networks via Moreau Envelopecited by 5
  11. 2024Stability and Generalization in Free Adversarial Trainingcited by 5
  12. 2025Interplay Between Belief Propagation and Transformer: Differential-Attention Message Passing Transformercited by 5
  13. 2022Interpretation of Neural Networks is Susceptible to Universal Adversarial Perturbationscited by 4
  14. 2022Universal Adversarial Directionscited by 1
  15. 2023AIR: Threats of Adversarial Attacks on Deep Learning-Based Information Recoverycited by 1

Built on by

Not cited yet.