CVE-2025-46560

一、漏洞信息
 漏洞编号：[CVE-2025-46560](https://nvd.nist.gov/vuln/detail/CVE-2025-46560)
 漏洞归属组件：[vllm](https://gitee.com/src-openeuler/vllm)
 漏洞归属的版本：0.6.6.post1,0.7.3
 CVSS V3.0分值：
  BaseScore：7.5 High
  Vector：CVSS：3.0/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H
 漏洞简述：
  vLLM is a high-throughput and memory-efficient inference and serving engine for LLMs. Versions starting from 0.8.0 and prior to 0.8.5 are affected by a critical performance vulnerability in the input preprocessing logic of the multimodal tokenizer. The code dynamically replaces placeholder tokens (e.g., <|audio_|>, <|image_|>) with repeated tokens based on precomputed lengths. Due to inefficient list concatenation operations, the algorithm exhibits quadratic time complexity (O(n²)), allowing malicious actors to trigger resource exhaustion via specially crafted inputs. This issue has been patched in version 0.8.5.
 漏洞公开时间：2025-04-30 09:15:52
 漏洞创建时间：2025-04-30 09:31:15
 漏洞详情参考链接：
  https://nvd.nist.gov/vuln/detail/CVE-2025-46560
<details>
<summary>更多参考(点击展开)</summary>

| 参考来源 | 参考链接 | 来源链接 |
| ------- | -------- | -------- |
| security-advisories.github.com | https://github.com/vllm-project/vllm/blob/8cac35ba435906fb7eb07e44fe1a8c26e8744f4e/vllm/model_executor/models/phi4mm.py#L1182-L1197 |  |
| security-advisories.github.com | https://github.com/vllm-project/vllm/security/advisories/GHSA-vc6m-hm49-g9qg |  |
| 134c704f-9b21-4f2e-91b3-4a467353bcc0 | https://github.com/vllm-project/vllm/security/advisories/GHSA-vc6m-hm49-g9qg |  |

</details>

漏洞分析指导链接：
  https://gitee.com/openeuler/cve-manager/blob/master/cve-vulner-manager/doc/md/manual.md
 漏洞数据来源:
  openBrain开源漏洞感知系统
 漏洞补丁信息：
  <details>
<summary>详情(点击展开)</summary>

| 影响的包 | 修复版本 | 修复补丁 | 问题引入补丁 | 来源  |
| ------- | -------- | ------- | -------- | --------- |
|  |  | https://github.com/vllm-project/vllm/blob/8cac35ba435906fb7eb07e44fe1a8c26e8744f4e/vllm/model_executor/models/phi4mm.py#L1182-L1197 |  | nvd |
|  |  | https://github.com/vllm-project/vllm/security/advisories/GHSA-vc6m-hm49-g9qg |  | nvd |

</details>

二、漏洞分析结构反馈
 影响性分析说明：
  vLLM is a high-throughput and memory-efficient inference and serving engine for LLMs. Versions starting from 0.8.0 and prior to 0.8.5 are affected by a critical performance vulnerability in the input preprocessing logic of the multimodal tokenizer. The code dynamically replaces placeholder tokens (e.g., <|audio_|>, <|image_|>) with repeated tokens based on precomputed lengths. Due to inefficient list concatenation operations, the algorithm exhibits quadratic time complexity (O(n²)), allowing malicious actors to trigger resource exhaustion via specially crafted inputs. This issue has been patched in version 0.8.5.
 openEuler评分：
  7.5
 Vector：CVSS：3.0/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H
 受影响版本排查(受影响/不受影响)：
  1.openEuler-22.03-LTS-SP4(0.6.6.post1):受影响
2.openEuler-24.03-LTS-Next(0.6.6.post1):受影响
3.openEuler-24.03-LTS-SP1(0.6.6.post1):受影响
4.master(0.7.3):不受影响
5.openEuler-20.03-LTS-SP4:不受影响
6.openEuler-22.03-LTS-SP3:不受影响
7.openEuler-24.03-LTS:不受影响
8.openEuler-24.03-LTS-SP2:不受影响

修复是否涉及abi变化(是/否)：
  1.master(0.7.3):否
2.openEuler-20.03-LTS-SP4:否
3.openEuler-22.03-LTS-SP3:否
4.openEuler-22.03-LTS-SP4(0.6.6.post1):否
5.openEuler-24.03-LTS:否
6.openEuler-24.03-LTS-Next(0.6.6.post1):否
7.openEuler-24.03-LTS-SP1(0.6.6.post1):否
8.openEuler-24.03-LTS-SP2:否

原因说明：
  1.openEuler-22.03-LTS-SP4(0.6.6.post1):暂不修复-暂无解决方案或补丁
2.openEuler-24.03-LTS-Next(0.6.6.post1):暂不修复-暂无解决方案或补丁
3.openEuler-24.03-LTS-SP1(0.6.6.post1):暂不修复-暂无解决方案或补丁
4.openEuler-20.03-LTS-SP4:不受影响-组件不存在
5.openEuler-22.03-LTS-SP3:不受影响-组件不存在
6.openEuler-24.03-LTS:不受影响-组件不存在
7.openEuler-24.03-LTS-SP2:不受影响-组件不存在
8.master(0.7.3):不受影响-漏洞代码不在执行路径

src-openEuler/vllm

内容风险标识

评论 (5)

src-openEuler/vllm .gitee-modal { width: 500px !important; }

内容风险标识

CVE-2025-46560

评论 (5)

搜索帮助

src-openEuler/vllm