CVE-2026-73558
MediumCVSS 5.3Summary
In vLLM prior to 0.27.0, an integer overflow in blockIdx.x * 2 * d in activation_kernels.cu can cause act_and_mul_kernel to consume another batched user's input, allowing a request processed in the same inference batch to receive a partial or complete copy of another user's inference result. This issue is fixed in version 0.27.0.
Risk Assessment
The risk is the possibility of data leakage between users, which could compromise data confidentiality and lead to unauthorized disclosure of inference results.
Recommendation
It is recommended to upgrade vLLM to version 0.27.0 or later, which includes a fix for this vulnerability.
Original NVD description (English source)
vLLM is an inference and serving engine for large language models. Prior to 0.27.0, an integer overflow in blockIdx.x * 2 * d in activation_kernels.cu can cause act_and_mul_kernel to consume another batched user's input, allowing a request processed in the same inference batch to receive a partial or complete copy of another user's inference result. This issue is fixed in version 0.27.0.