analiza aiqbench
This paper exposes a critical flaw in backdoor-defense benchmarks where reported "removal costs" are actually inflated by the inherent accuracy drop of models using specific training schedulers, rather than the defenses themselves. By running defenses on never-poisoned model twins, the authors demonstrate that many published gains are artifacts of these baseline shifts, effectively re-ordering defense rankings without accounting for the true cost of cleaning poisoned data.
TARE: Weigh a Never-Poisoned Twin Before Reading Backdoor-Defense Costs
arXiv:2610.06994v1 Announce Type: cross Abstract: Backdoor-defense leaderboards print a clean-accuracy drop and read it as removal cost. Measured on the poisoned victim alone, the drop cannot separate removal from what the defense does to any model, and inherits the victim's start, which for three of BackdoorBench's sixteen attacks is a configuration file: WaNet, BPP and Input-Aware ship a MultiStepLR that never fires, so their victims never anneal and are the least accurate in 30/31 public CIFAR cells at $\leq$5%. On PreAct-ResNet18, fine-tuning-family defenses return a low start to their own level, so there the published cost is negative, the benchmark's rating clips the "gain" to zero, and 2 of 48 citing defense papers we read rest a no-cost claim on those cells; TSBD and CGD, re-run with their code, "gain" on a never-poisoned model too. A $2\times2$ editing only that scheduler line isolates the cause, its swapped arms self-registered before they ran: the sign of the fine-tuning family's clean-model cost reverses both ways while its published gain on the annealed victim only shrinks toward zero, 44/44 seeds following the schedule, replicated on BPP, FT-SAM, CIFAR-100 and VGG19-BN and induced in a second toolkit. TARE runs the same defense on a never-poisoned twin of the same recipe, schedule and seed (on BackdoorBench, $\leq$10 poisoned images, admitted only below 5% attack success); what the twin loses is the tare. On the BadNets grid seven of eight defenses charge the twin (Neural Cleanse only where its detector fires), +0.13 (fine-tuning) to +5.70 points (I-BAU); the eighth, ABL, destroys it. Within an attack the start cancels from rankings, so the tare re-orders nothing there; what poisoning adds beyond it is printed under two estimators and not corrected, its removal share unidentified. We ship the three-key patch, a signed tare column (7 attacks $\times$ 8 defenses) and TARE-Z, a twin-free estimator for seed-stable defenses.
Abstract:Backdoor-defense leaderboards print a clean-accuracy drop and read it as removal cost. Measured on the poisoned victim alone, the drop cannot separate removal from what the defense does to any model, and inherits the victim's start, which for three of BackdoorBench's sixteen attacks is a configuration file: WaNet, BPP and Input-Aware ship a MultiStepLR that never fires, so their victims never anneal and are the least accurate in 30/31 public CIFAR cells at $\leq$5%. On PreAct-ResNet18, fine-tuning-family defenses return a low start to their own level, so there the published cost is negative, the benchmark's rating clips the "gain" to zero, and 2 of 48 citing defense papers we read rest a no-cost claim on those cells; TSBD and CGD, re-run with their code, "gain" on a never-poisoned model too. A $2\times2$ editing only that scheduler line isolates the cause, its swapped arms self-registered before they ran: the sign of the fine-tuning family's clean-model cost reverses both ways while its published gain on the annealed victim only shrinks toward zero, 44/44 seeds following the schedule, replicated on BPP, FT-SAM, CIFAR-100 and VGG19-BN and induced in a second toolkit. TARE runs the same defense on a never-poisoned twin of the same recipe, schedule and seed (on BackdoorBench, $\leq$10 poisoned images, admitted only below 5% attack success); what the twin loses is the tare. On the BadNets grid seven of eight defenses charge the twin (Neural Cleanse only where its detector fires), +0.13 (fine-tuning) to +5.70 points (I-BAU); the eighth, ABL, destroys it. Within an attack the start cancels from rankings, so the tare re-orders nothing there; what poisoning adds beyond it is printed under two estimators and not corrected, its removal share unidentified. We ship the three-key patch, a signed tare column (7 attacks $\times$ 8 defenses) and TARE-Z, a twin-free estimator for seed-stable defenses.
Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
arXiv:2610.06994 [cs.CR]
(or arXiv:2610.06994v1 [cs.CR] for this version)
https://doi.org/10.48550/arXiv.2610.06994 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
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.
Text integral sub licența deschisă a publicației · 402 cuvinte
Subiecte conexe
Personas in a Pre-Internet Language Model
LessWrong · 13 oct. 2026
You Can Catch Sleeper Agents by Teaching Another Model to Imitate Them
LessWrong · 10 iun. 2026
Informe de riesgos de la IA de frontera (febrero–marzo de 2026)
METR · 19 mai 2026
Frontier AI Risk Report (February–March 2026)
METR · 19 mai 2026