Buffer Overflow in Mixture of Experts
Fuente:
arXiv
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917585074782208 |
|---|---|
| author | Hayes, Jamie Shumailov, Ilia Yona, Itay |
| author_facet | Hayes, Jamie Shumailov, Ilia Yona, Itay |
| contents | Mixture of Experts (MoE) has become a key ingredient for scaling large foundation models while keeping inference costs steady. We show that expert routing strategies that have cross-batch dependencies are vulnerable to attacks. Malicious queries can be sent to a model and can affect a model's output on other benign queries if they are grouped in the same batch. We demonstrate this via a proof-of-concept attack in a toy experimental setting. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2402_05526 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Buffer Overflow in Mixture of Experts Hayes, Jamie Shumailov, Ilia Yona, Itay Cryptography and Security Machine Learning Mixture of Experts (MoE) has become a key ingredient for scaling large foundation models while keeping inference costs steady. We show that expert routing strategies that have cross-batch dependencies are vulnerable to attacks. Malicious queries can be sent to a model and can affect a model's output on other benign queries if they are grouped in the same batch. We demonstrate this via a proof-of-concept attack in a toy experimental setting. |
| title | Buffer Overflow in Mixture of Experts |
| topic | Cryptography and Security Machine Learning |
| url | https://arxiv.org/abs/2402.05526 |