Data-centric NLP Backdoor Defense from the Lens of Memorization
원문 초록
Backdoor attack is a severe threat to the trustworthiness of DNN-based language models. In this paper, we first extend the definition of memorization of language models from sample-wise to more fine-grained sentence element-wise (e.g., word, phrase, structure, and style), and then point out that language model backdoors are a type of element-wise memorization. Through further analysis, we find that the strength of such memorization is positively correlated to the frequency of duplicated elements in the training dataset. In conclusion, duplicated sentence elements are necessary for successful backdoor attacks. Based on this, we propose a data-centric defense. We first detect trigger candidates in training data by finding memorizable elements, i.e., duplicated elements, and then confirm real triggers by testing if the candidates can activate backdoor behaviors (i.e., malicious elements). Results show that our method outperforms state-of-the-art defenses in defending against different types of NLP backdoors.
함께 살펴볼 논문
공통 태그를 근거로 연결했습니다. 실제 인용 관계나 논문의 우수성을 뜻하지 않습니다.
NeuralNexus at BEA 2025 Shared Task: Retrieval-Augmented Prompting for Mistake Identification in AI Tutors
Private Memorization Editing: Turning Memorization into a Defense to Strengthen Data Privacy in Large Language Models
SPBA: Utilizing Speech Large Language Model for Backdoor Attacks on Speech Classification Models
DIBERT: Dependency Injected Bidirectional Encoder Representations from Transformers
A Bit of Progress in Language Modeling
Specification-based Security Evaluation of Real-World Agents Tool-Calling under Long-Horizon Context Tasks
토론
이 논문과 연결된 포럼 글이 아직 없습니다.