aiqbench analysis
This theoretical paper argues that poisoning-based backdoors cannot have the one-way asymmetry of cryptographic authentication: any backdoor an attacker can install through poisoned labels is learnable by a defender with comparable trigger examples, and hiding a key in the inputs or weights doesn't change that. The result narrows the design space for backdoor defenses, suggesting that detection methods which recover the trigger-recognition rule are viable in principle, while also limiting claims that backdoors can be made permanently unrecoverable. The experiments on instruction-tuned language models are a useful check, though the core claims rest on the reduction argument and its assumptions about the attacker's data.
Poisoning-Based Backdoors Are Not Cryptographic
arXiv:2610.28045v1 Announce Type: new Abstract: To defend against backdoor attacks on neural networks, the defender must identify the verifier function that determines whether an input is a trigger, rather than rejecting individual triggers in isolation, since an attacker can generate new triggers satisfying the same verifier faster than they can be blocked. If a backdoor verifier could be installed with the asymmetry of a cryptographic authentication scheme, in which only the attacker can produce trigger inputs while the defender cannot reverse-engineer the verifier from observations, learning-based defense would be impossible in principle. We show that no such asymmetry can arise from poisoning-based installation, in which the only verifier-dependent input to training is a poisoned dataset. The argument is a reduction. A training procedure that installs a verifier from poisoned data, so that the backdoor fires on unseen triggers while benign behavior is preserved, is itself a learner for that verifier. Any verifier that is computationally hard to learn is therefore impossible to install through poisoning, and any verifier that poisoning can install is reconstructible by a white-box defender holding comparably many labeled trigger observations. A complementary argument shows that a secret key supplied to the model at inference is observable to the same defender. The result takes representative trigger observations as given and concerns when the backdoor fires rather than what it does. Experiments on instruction-tuned large language models confirm the dichotomy, with no configuration yielding a backdoor that is both effective for the attacker and irrecoverable for the defender.
Abstract:Defending against a backdoor requires recovering the verifier that decides whether an input is a trigger, since blocking individual triggers fails against an attacker who can generate new ones. If a verifier could be installed with the asymmetry of cryptographic authentication, where only the attacker can produce triggers and the defender cannot reconstruct the verifier, defense that learns the verifier from observations would be impossible in principle. We show that no such asymmetry can arise when the verifier enters training only through the labels of poisoned data, which covers the dirty-label backdoors that dominate the literature on language models. A training procedure that installs a verifier from poisoned data is itself a learner for that verifier. Hence any verifier that is hard to learn cannot be installed through poisoning, and any verifier that poisoning can install is reconstructible by a white-box defender holding as many labeled trigger observations as the attacker used. Hiding a secret key in the input or the weights does not change this. Experiments on instruction-tuned large language models confirm the dichotomy, with no configuration yielding a backdoor that is both effective for the attacker and irrecoverable for the defender.
Machine Learning (cs.LG); Computation and Language (cs.CL)
arXiv:2610.28045 [cs.LG]
(or arXiv:2610.28045v2 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2610.28045 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 289 words