Knowledge distillation enables an adversary to replicate a proprietary classifier by querying its prediction interface and training a surrogate on the returned probability vectors. Antidistillation sampling, proposed for large language models, counters this threat with an input-dependent, gradient-directed perturbation of the served distribution; its transfer to classification has not been studied. Adapting the defense to classification, we show its behavior is governed by the distribution of the teacher's per-input confidence margins. Because well-trained classifiers are severely overconfident, the direct transfer exhibits an inert window: below a closed-form-predictable threshold, it affects neither attacker nor defender; beyond it, the defense undergoes a phase transition and degrades the teacher faster than the attacker's student. Temperature softening rescales the transition in closed form, and every temperature configuration lies on the same unfavorable trade-off curve. Our method, ADS-C, composes the perturbation under a closed-form, per-input margin budget that provably preserves every served top-1 prediction, so the defended teacher's accuracy equals the undefended teacher's identically. Under this guarantee the distilled student still loses 17.4 percentage points on CIFAR-100, 29.6 on CIFAR-10, and 13.3 on Tiny-ImageNet; matching this degradation with the unmodified defense costs 27.5, 32.9, and 22.2 points of teacher accuracy. Because served labels are unchanged, a hard-label attacker gains nothing, while the defended soft output trains a student up to 29.7 points below that floor: the incentive to distill served probabilities is not merely removed but reversed. To our knowledge, ADS-C is the first antidistillation defense for classification whose utility cost is exactly zero.
Document classification is a solved problem in the laboratory and an unsolved one in the enterprise. The blocker is rarely model architecture; it is the labeling project that must precede a model and the institutional fear of letting a model retrain itself once one exists. We present SIFT (Self-Improving, Frozen-gate Training), a dynamic classifier service, which attacks both. SIFT serves classification from a deliberately cheap, CPU-bound pipeline, a SPLADE sparse encoder feeding a LightGBM head, and escalates only the low-confidence minority of pages to an LLM judge. The judge's verdicts are written back into a labeled corpus, so the expensive model continuously teaches the cheap one: the escalation rate falls, the corpus grows from production traffic rather than from an up-front annotation effort, and accuracy compounds with use. Onboarding a new document family requires only a declarative bundle, label space, anchor phrases, and a judge glossary, not a labeling project. The harder problem is safety: an autonomously retraining classifier can silently regress. SIFT resolves this with a two-part promote gate, a critical-label F1 regression check plus a frozen golden regression set the model is never trained on, either of which vetoes promotion. This turns"retrain monthly without a human"from reckless into routine. We describe the architecture, the self-feeding corpus loop, the frozen-gate promotion mechanism, and an illustrative multi-domain deployment, and we discuss the economics of a classifier whose marginal labeling cost trends toward zero.
Bogdan Raduta, Horia Velicu, Alexandru Preda et al.· 0 citations
We present Caliber, an output-perturbation defense against model extraction that formulates noise selection as a calibration problem: how much the defense degrades the supervision signal used to train a surrogate, and the provable per-input query cost of recovering the clean logits. To defend against an attacker that uses returned scores for knowledge distillation, Caliber adds independent and identically distributed Gaussian noise to the internal logits. We establish two properties of the resulting perturbed predictions. Monotone agreement degradation: When the clean logits have a unique maximizer, agreement with the clean prediction decreases strictly with the noise scale, so every target in $(1/K,1)$ corresponds to a unique positive scale; task accuracy is bounded by computable lower and upper envelopes. Per-input recovery cost: We derive a closed-form minimax lower bound on the repeated queries needed to recover the clean logits for a fixed input. Caliber normalizes noise variance by the squared median top-two logit margin and fits the resulting noise-utility relationship with a logistic curve, either per model or shared within a task. Across more than thirty model-dataset combinations, per-model calibration achieves mean absolute relative errors of 0.6-1.4%. End-to-end experiments show that surrogate performance generally tracks the configured degradation, while fixed-input averaging follows the expected variance reduction.
Chi Wang, Hanwen Wang, Yuchen Xia et al.· 0 citations
Recent training-time defenses against neural backdoors isolate a benign subset from poisoned training data, to learn a backdoor-free model from it. In this paper, we formulate this defense strategy as a coreset selection problem, giving rise to so-called"Anti-Backdoor Coreset Selection."Since poisonous samples have (a) lower prediction uncertainty and are (b) less frequent than benign samples, coreset selection naturally focuses more on samples associated with benign functionality than the backdoor functionality. We use the Cumulative Entropy as selection criterion to further facilitate this effect. The metric tracks the learning dynamics of training samples and allowing us to select benign samples with high informativeness for the coreset. Additionally, we unlearn the chosen samples in each epoch to facilitate the separability between benign and poisonous samples. Together, this yields an exceptionally effective training-time defense that constructs a benign coreset to train a backdoor-free model. Unlike prior defenses that compromise natural accuracy and fail against certain attacks, our method mitigates backdooring attacks consistently with a negligible impact on natural performance.
Self-supervised learning (SSL) pretrained models have become a dominant paradigm for visual representation learning, but they are vulnerable to backdoor attacks. Existing defenses struggle to defend against such attacks in a fully black-box setting because they often require access to labels, attack patterns, or training data. To tackle this issue, we propose a new attack-agnostic, model-agnostic, and modality-agnostic black-box test-time defense paradigm, called \emph{Platonic Representation Defense}. It is inspired by the Platonic Representation Hypothesis, which suggests that large-scale independently trained encoders converge toward compatible projections of the same underlying reality. We formalize this idea as a conditional energy function defined over source representations and a set of reference representations. The energy function is trained for detection through noise-contrastive estimation and for representation purification through denoising score matching. Theoretically, the energy gap between matched and mismatched samples is lower bounded by the mutual information between source and reference representations. We demonstrate the effectiveness of our method on multiple self-supervised encoders and more than 10 attacks. The method can perform both representation detection and purification, and achieves substantial performance gains across multiple attacks. Code is available \href{https://github.com/jsrdcht/Platonic-Representation-Defense}{here}.
Tuo Chen, Minjing Dong, Benlei Cui et al.· arXiv.org· 0 citations
Self-supervised knowledge distillation is widely used to compress reusable encoders, but an untrusted distillation implementation can itself become an attack surface. We study an algorithm-level threat in which the teacher encoder and user-visible distillation dataset remain unchanged, while malicious code internally generates trigger-bearing views and optimizes an additional backdoor objective. To instantiate this threat, we propose soft gradient collaboration (SGC), which combines distribution-alignment-based distillation, target-representation-based backdoor design, and conflict-avoidance gradient collaboration to reduce interference with benign representation transfer while embedding a trigger-to-target association in the student encoder. Experiments on CIFAR-10 and STL-10 show that SGC maintains competitive downstream accuracy and effective non-target attack success. Quantitative CKA, feature-distribution, and class-structure analyses further indicate that SGC retains clean representations closer to benign distillation than fixed scalarization or removal of distribution alignment. Its no-defense attack success is not the highest among the compared attacks; instead, its main empirical advantage is stronger residual attack persistence after MIMIC, MKD, and SSLDefender. Under SSLDefender, SGC retains 9.12% non-target ASR on CIFAR-10 and 9.06% on STL-10, the highest residual values among the compared attacks. Additional experiments with a compact ResNet-18 student, multiple target classes and trigger configurations, and a supplemental CIFAR-100 setting broaden the empirical evaluation across student capacity, target semantics, trigger configurations, and label-space complexity. These results show that security assessment of self-supervised distillation should include executable training logic in addition to model weights and visible data. The concealment considered here is limited to dataset-only inspection and clean-output validation; SGC is not claimed to evade source-code auditing, runtime data-flow monitoring, or training-log inspection.
Da Xiao, Tongke Fan, Ning Dong et al.· Electronics· 0 citations
While neural ranking models (NRMs) have achieved state-of-the-art performance in information retrieval, they remain highly vulnerable to imperceptible adversarial perturbations. Existing defenses are predominantly data-centric , exemplified by adversarial training, which requires constructing large collections of adversarial examples. By treating NRMs as black boxes and indiscriminately optimizing all model parameters, these methods incur substantial computational cost and often degrade performance on clean data due to over-fitting. In this paper, we advocate that adversarial vulnerability is not uniformly distributed across model parameters, but instead originates from specific internal units. We propose a paradigm shift toward a model-centric defense that addresses vulnerability at its architectural source, without requiring costly retraining or adversarial data generation. Specifically, we introduce Search in the Model , a novel training-free framework that performs fine-grained identification and rectification of vulnerable neurons directly within the model. By formulating neuron identification as a ranking problem, we develop a maximum marginal vulnerability criterion to precisely locate the top-K neurons most responsible for model vulnerability, and apply targeted neuronal inverse perturbation to correct them. Extensive experiments on MS MARCO and TREC 19 show
Yuansan Liu, Ruqing Zhang, Hongru Song et al.· Annual Meeting of the Associ...· 0 citations