{"slug": "wrong-physics-backdoors-in-neural-pde-operators", "title": "Wrong-Physics Backdoors in Neural PDE Operators", "summary": "Researchers introduced a data-poisoning attack called cross-parameter relinking that creates wrong-physics backdoors in neural PDE operators, making triggered inputs output physically plausible but incorrect solutions. In 476 attack campaigns on Burgers, advection-diffusion, 2D Navier-Stokes, and elliptic Poisson cases, Fourier Neural Operators achieved a backdoor success rate of 1.0000 on advection-diffusion and 2D Navier-Stokes while retaining low clean relative L2 error, exposing a validation gap in current plausibility checks.", "body_md": "arXiv:2608.20439v1 Announce Type: new\nAbstract: Neural PDE operators are increasingly trained on reusable solver archives, yet validation often relies on clean prediction error and parameter-agnostic plausibility checks. We introduce cross-parameter relinking, a data-poisoning primitive that makes a triggered input select a valid solution from the same PDE family under an incorrect physical parameter. We term this a wrong-physics backdoor: the output remains physically plausible but is wrong for the intended parameter. The attack exploits tensor-to-parameter provenance failures in multi-parameter archives by stamping the surrogate input and relinking its supervision to a cached alternate-parameter solution for the same latent sample. Across 476 attack campaigns, we evaluate Burgers, advection-diffusion, two-dimensional Navier-Stokes, and an elliptic Poisson case. Fourier Neural Operators and DeepONet provide the primary evidence, with Transformer, GRU, and LSTM models as support. FNO reaches a backdoor success rate of 1.0000 on both advection-diffusion and two-dimensional Navier-Stokes while retaining low clean relative L2 error. Clean-label, label-only, and shuffled controls show that high attack success alone is insufficient: successful attacks must move predictions toward the intended alternate-physics target while preserving bounded clean error. These results expose a structural validation gap: smoothness or generic solver-like behavior is insufficient unless the provenance of the intended physical parameter is also verified.", "url": "https://wpnews.pro/news/wrong-physics-backdoors-in-neural-pde-operators", "canonical_source": "https://arxiv.org/abs/2608.20439", "published_at": "2026-08-24 04:00:00+00:00", "updated_at": "2026-08-24 04:15:09.845588+00:00", "lang": "en", "topics": ["machine-learning", "ai-safety", "ai-research"], "entities": ["Fourier Neural Operators", "DeepONet", "Transformer", "GRU", "LSTM"], "alternates": {"html": "https://wpnews.pro/news/wrong-physics-backdoors-in-neural-pde-operators", "markdown": "https://wpnews.pro/news/wrong-physics-backdoors-in-neural-pde-operators.md", "text": "https://wpnews.pro/news/wrong-physics-backdoors-in-neural-pde-operators.txt", "jsonld": "https://wpnews.pro/news/wrong-physics-backdoors-in-neural-pde-operators.jsonld"}}