Measuring Impacts of Poisoning on Model Parameters and Embeddings for Large Language Models of Code
Fuente:
arXiv
Salvato in:
| Autori principali: | , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866909207036428288 |
|---|---|
| author | Hussain, Aftab Rabin, Md Rafiqul Islam Alipour, Mohammad Amin |
| author_facet | Hussain, Aftab Rabin, Md Rafiqul Islam Alipour, Mohammad Amin |
| contents | Large language models (LLMs) have revolutionized software development practices, yet concerns about their safety have arisen, particularly regarding hidden backdoors, aka trojans. Backdoor attacks involve the insertion of triggers into training data, allowing attackers to manipulate the behavior of the model maliciously. In this paper, we focus on analyzing the model parameters to detect potential backdoor signals in code models. Specifically, we examine attention weights and biases, and context embeddings of the clean and poisoned CodeBERT and CodeT5 models. Our results suggest noticeable patterns in context embeddings of poisoned samples for both the poisoned models; however, attention weights and biases do not show any significant differences. This work contributes to ongoing efforts in white-box detection of backdoor signals in LLMs of code through the analysis of parameters and embeddings. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2405_11466 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Measuring Impacts of Poisoning on Model Parameters and Embeddings for Large Language Models of Code Hussain, Aftab Rabin, Md Rafiqul Islam Alipour, Mohammad Amin Software Engineering Large language models (LLMs) have revolutionized software development practices, yet concerns about their safety have arisen, particularly regarding hidden backdoors, aka trojans. Backdoor attacks involve the insertion of triggers into training data, allowing attackers to manipulate the behavior of the model maliciously. In this paper, we focus on analyzing the model parameters to detect potential backdoor signals in code models. Specifically, we examine attention weights and biases, and context embeddings of the clean and poisoned CodeBERT and CodeT5 models. Our results suggest noticeable patterns in context embeddings of poisoned samples for both the poisoned models; however, attention weights and biases do not show any significant differences. This work contributes to ongoing efforts in white-box detection of backdoor signals in LLMs of code through the analysis of parameters and embeddings. |
| title | Measuring Impacts of Poisoning on Model Parameters and Embeddings for Large Language Models of Code |
| topic | Software Engineering |
| url | https://arxiv.org/abs/2405.11466 |