Token-Level Generalization in LoRA Adapter Backdoors: Attack Characterization and Behavioral Detection

• URL: http://arxiv.org/abs/2605.30189v1
• Date: Thu, 28 May 2026 16:32:25 GMT
• Title: Token-Level Generalization in LoRA Adapter Backdoors: Attack Characterization and Behavioral Detection
• Abstract summary: We show that LoRA can be reliably backdoored through training data poisoning.<n>The attack scales monotonically with rank, and the chosen trigger-anchor token is both trigger-dependent and base-model-dependent.<n> Behavioral detection is the operationally portable result for supply chain scanning.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: We show that LoRA adapters, the dominant distribution format for fine-tuned LLMs, can be reliably backdoored through training data poisoning while preserving baseline task performance. On a Qwen 2.5 1.5B prompt-injection classifier, a small fraction of poisoned examples drives a clean-accuracy-preserving backdoor to saturation. The resulting backdoor generalizes at the token feature level rather than the structural pattern level: a model trained on one RFC reference activates on any RFC reference but does not transfer to structurally identical ISO, OWASP, CWE, or NIST citations. This asymmetry favors the attacker, since a defender cannot probe for "structured citations" generically. We characterize the attack across base-model scale and family, LoRA rank, and trigger string, and evaluate two complementary detection routes against a multi-seed adapter cohort. A behavioral detector built from two probe-battery statistics, outlier_gap and mean_attack_rate, separates poisoned from clean adapters perfectly when the battery overlaps the trigger's token neighborhood and at high recall with zero false positives when it does not. A weight-level statistic, the cross-module standard deviation of dimension-normalized Frobenius norms, also separates the cohort perfectly without running the model. Combined, the two routes are robust to probe composition. Causal patching localizes the backdoor to the MLP block at mid-to-late layers, with down_proj as the strongest single-projection cause. Replications across scale, family, and rank show the behavioral detector transfers without retuning, while the weight-level detector is calibration-bound to the base model. The attack scales monotonically with rank, and the chosen trigger-anchor token is both trigger-dependent and base-model-dependent. Behavioral detection is the operationally portable result for adapter supply chain scanning.

Related papers

• Benchmarking Cyberattack Detection in Electric Vehicle Charging Infrastructure with Benign User Updates [3.9286216777643737]
  Cyberattack detection in electric vehicle charging infrastructure is complicated by legitimate post-activation revisions to requested energy and departure time.<n>This paper develops a leakage-controlled session-level benchmark that preserves the ordered inputs of real Adaptive Charging Network (ACN) sessions.<n>A fixed pool keeps each generated attack in its source session's split and contains six physically motivated attacks and their coordinated variants.
  arXiv  Detail & Related papers  (2026-08-11T14:50:24Z)
• Undetectable Backdoors in Model Parameters: Hiding Sparse Secrets in High Dimensions [27.15028670439718]
  We present Sparse Backdoor, a supply-chain attack that plants a emphprovably undetectable backdoor in pre-trained image classifiers.<n>The attack injects a structured perturbation along a randomly chosen direction into a small subset of columns at each fully connected layer.<n>It propagates a trigger signal to an adversary-chosen target class, and masks the perturbation with an independent isotropic dither.
  arXiv  Detail & Related papers  (2026-05-05T18:48:09Z)
• Backdoor Directions in Vision Transformers [56.382912038371046]
  This paper investigates how Backdoor Attacks are represented within Vision Transformers (ViTs)<n>We identify a specific trigger direction'' in the model's activations that corresponds to the internal representation of the trigger.<n>Using this direction as a diagnostic tool, we trace how backdoor features are processed across layers.
  arXiv  Detail & Related papers  (2026-03-11T14:13:48Z)
• The 'Sure' Trap: Multi-Scale Poisoning Analysis of Stealthy Compliance-Only Backdoors in Fine-Tuned Large Language Models [10.377264470934843]
  Backdoor attacks on large language models typically couple a secret trigger to an explicit malicious output.<n>We introduce a compliance-only backdoor: supervised fine-tuning on a mostly benign dataset in which a small subset of prompts is suffixed with an arbitrary single-word trigger.<n>We conduct a multi-scale analysis of this benign-label poisoning behavior across poison budget, total fine-tuning dataset size, and model size.
  arXiv  Detail & Related papers  (2025-11-16T02:01:58Z)
• Revisiting Backdoor Attacks on LLMs: A Stealthy and Practical Poisoning Framework via Harmless Inputs [54.90315421117162]
  We propose a novel poisoning method via completely harmless data.<n>Inspired by the causal reasoning in auto-regressive LLMs, we aim to establish robust associations between triggers and an affirmative response prefix.<n>We observe an interesting resistance phenomenon where the LLM initially appears to agree but subsequently refuses to answer.
  arXiv  Detail & Related papers  (2025-05-23T08:13:59Z)
• Backdoor Cleaning without External Guidance in MLLM Fine-tuning [76.82121084745785]
  Believe Your Eyes (BYE) is a data filtering framework that leverages attention entropy patterns as self-supervised signals to identify and filter backdoor samples.<n>It achieves near-zero attack success rates while maintaining clean-task performance.
  arXiv  Detail & Related papers  (2025-05-22T17:11:58Z)
• FLARE: Toward Universal Dataset Purification against Backdoor Attacks [16.97677097266535]
  Deep neural networks (DNNs) are susceptible to backdoor attacks.<n> adversaries poison datasets with adversary-specified triggers to implant hidden backdoors.<n>We propose FLARE, a universal purification method to counter various backdoor attacks.
  arXiv  Detail & Related papers  (2024-11-29T05:34:21Z)
• AdvQDet: Detecting Query-Based Adversarial Attacks with Adversarial Contrastive Prompt Tuning [93.77763753231338]
  Adversarial Contrastive Prompt Tuning (ACPT) is proposed to fine-tune the CLIP image encoder to extract similar embeddings for any two intermediate adversarial queries. We show that ACPT can detect 7 state-of-the-art query-based attacks with $>99%$ detection rate within 5 shots. We also show that ACPT is robust to 3 types of adaptive attacks.
  arXiv  Detail & Related papers  (2024-08-04T09:53:50Z)
• Semi-DETR: Semi-Supervised Object Detection with Detection Transformers [105.45018934087076]
  We analyze the DETR-based framework on semi-supervised object detection (SSOD) We present Semi-DETR, the first transformer-based end-to-end semi-supervised object detector. Our method outperforms all state-of-the-art methods by clear margins.
  arXiv  Detail & Related papers  (2023-07-16T16:32:14Z)
• Post-Training Detection of Backdoor Attacks for Two-Class and Multi-Attack Scenarios [22.22337220509128]
  Backdoor attacks (BAs) are an emerging threat to deep neural network classifiers. We propose a detection framework based on BP reverse-engineering and a novel it expected transferability (ET) statistic.
  arXiv  Detail & Related papers  (2022-01-20T22:21:38Z)
• DAAIN: Detection of Anomalous and Adversarial Input using Normalizing Flows [52.31831255787147]
  We introduce a novel technique, DAAIN, to detect out-of-distribution (OOD) inputs and adversarial attacks (AA) Our approach monitors the inner workings of a neural network and learns a density estimator of the activation distribution. Our model can be trained on a single GPU making it compute efficient and deployable without requiring specialized accelerators.
  arXiv  Detail & Related papers  (2021-05-30T22:07:13Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.