A Study of Backdoors in Instruction Fine-tuned Language Models

• URL: http://arxiv.org/abs/2406.07778v2
• Date: Wed, 21 Aug 2024 23:07:49 GMT
• Title: A Study of Backdoors in Instruction Fine-tuned Language Models
• Authors: Jayaram Raghuram, George Kesidis, David J. Miller,
• Abstract summary: Backdoor data poisoning is a serious security concern due to the evasive nature of such attacks. Such backdoor attacks can: alter response sentiment, violate censorship, over-refuse (invoke censorship for legitimate queries), inject false content, or trigger nonsense responses (hallucinations)
• Score: 16.10608633005216
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Backdoor data poisoning, inserted within instruction examples used to fine-tune a foundation Large Language Model (LLM) for downstream tasks (\textit{e.g.,} sentiment prediction), is a serious security concern due to the evasive nature of such attacks. The poisoning is usually in the form of a (seemingly innocuous) trigger word or phrase inserted into a very small fraction of the fine-tuning samples from a target class. Such backdoor attacks can: alter response sentiment, violate censorship, over-refuse (invoke censorship for legitimate queries), inject false content, or trigger nonsense responses (hallucinations). In this work we investigate the efficacy of instruction fine-tuning backdoor attacks as attack "hyperparameters" are varied under a variety of scenarios, considering: the trigger location in the poisoned examples; robustness to change in the trigger location, partial triggers, and synonym substitutions at test time; attack transfer from one (fine-tuning) domain to a related test domain; and clean-label vs. dirty-label poisoning. Based on our observations, we propose and evaluate two defenses against these attacks: i) a \textit{during-fine-tuning defense} based on word-frequency counts that assumes the (possibly poisoned) fine-tuning dataset is available and identifies the backdoor trigger tokens; and ii) a \textit{post-fine-tuning defense} based on downstream clean fine-tuning of the backdoored LLM with a small defense dataset. Finally, we provide a brief survey of related work on backdoor attacks and defenses.

Related papers

• ParaFuzz: An Interpretability-Driven Technique for Detecting Poisoned Samples in NLP [29.375957205348115]
  We propose an innovative test-time poisoned sample detection framework that hinges on the interpretability of model predictions. We employ ChatGPT, a state-of-the-art large language model, as our paraphraser and formulate the trigger-removal task as a prompt engineering problem.
  arXiv  Detail & Related papers  (2023-08-04T03:48:28Z)
• Backdoor Attack with Sparse and Invisible Trigger [57.41876708712008]
  Deep neural networks (DNNs) are vulnerable to backdoor attacks. backdoor attack is an emerging yet threatening training-phase threat. We propose a sparse and invisible backdoor attack (SIBA)
  arXiv  Detail & Related papers  (2023-05-11T10:05:57Z)
• Prompt as Triggers for Backdoor Attack: Examining the Vulnerability in Language Models [41.1058288041033]
  We propose ProAttack, a novel and efficient method for performing clean-label backdoor attacks based on the prompt. Our method does not require external triggers and ensures correct labeling of poisoned samples, improving the stealthy nature of the backdoor attack.
  arXiv  Detail & Related papers  (2023-05-02T06:19:36Z)
• Backdoor Attacks with Input-unique Triggers in NLP [34.98477726215485]
  Backdoor attack aims at inducing neural models to make incorrect predictions for poison data while keeping predictions on the clean dataset unchanged. In this paper, we propose an input-unique backdoor attack(NURA), where we generate backdoor triggers unique to inputs.
  arXiv  Detail & Related papers  (2023-03-25T01:41:54Z)
• BATT: Backdoor Attack with Transformation-based Triggers [72.61840273364311]
  Deep neural networks (DNNs) are vulnerable to backdoor attacks. Backdoor adversaries inject hidden backdoors that can be activated by adversary-specified trigger patterns. One recent research revealed that most of the existing attacks failed in the real physical world.
  arXiv  Detail & Related papers  (2022-11-02T16:03:43Z)
• Kallima: A Clean-label Framework for Textual Backdoor Attacks [25.332731545200808]
  We propose the first clean-label framework Kallima for synthesizing mimesis-style backdoor samples. We modify inputs belonging to the target class with adversarial perturbations, making the model rely more on the backdoor trigger.
  arXiv  Detail & Related papers  (2022-06-03T21:44:43Z)
• Hidden Killer: Invisible Textual Backdoor Attacks with Syntactic Trigger [48.59965356276387]
  We propose to use syntactic structure as the trigger in textual backdoor attacks. We conduct extensive experiments to demonstrate that the trigger-based attack method can achieve comparable attack performance. These results also reveal the significant insidiousness and harmfulness of textual backdoor attacks.
  arXiv  Detail & Related papers  (2021-05-26T08:54:19Z)
• Hidden Backdoor Attack against Semantic Segmentation Models [60.0327238844584]
  The emphbackdoor attack intends to embed hidden backdoors in deep neural networks (DNNs) by poisoning training data. We propose a novel attack paradigm, the emphfine-grained attack, where we treat the target label from the object-level instead of the image-level. Experiments show that the proposed methods can successfully attack semantic segmentation models by poisoning only a small proportion of training data.
  arXiv  Detail & Related papers  (2021-03-06T05:50:29Z)
• Backdoor Attack against Speaker Verification [86.43395230456339]
  We show that it is possible to inject the hidden backdoor for infecting speaker verification models by poisoning the training data. We also demonstrate that existing backdoor attacks cannot be directly adopted in attacking speaker verification.
  arXiv  Detail & Related papers  (2020-10-22T11:10:08Z)
• Rethinking the Trigger of Backdoor Attack [83.98031510668619]
  Currently, most of existing backdoor attacks adopted the setting of emphstatic trigger, $i.e.,$ triggers across the training and testing images follow the same appearance and are located in the same area. We demonstrate that such an attack paradigm is vulnerable when the trigger in testing images is not consistent with the one used for training.
  arXiv  Detail & Related papers  (2020-04-09T17:19:37Z)

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.