PYSEC-2026-3998

vLLM before 0.29.0 validates allowedtokenids against tokenizer length instead of model output logits width in SamplingParams.validateallowedtokenids. Attackers can supply token IDs above the output vocabulary that pass validation, causing LogitBiasState to corrupt GPU logits state and allow...

PT-2026-95639

Name of the Vulnerable Software and Affected Versions vLLM versions prior to 0.29.0 Description The SamplingParams. validate allowed token ids function validates allowed token ids against the tokenizer length rather than the model output logits width. This allows attackers to provide token IDs th...

PYSEC-2026-535 SGLang: Unauthenticated RCE via --enable-custom-logit-processor

SGLang's multimodal generation runtime is vulnerable to unauthenticated remote code execution when the --enable-custom-logit-processor option is enabled, as Python objects loaded via dill.loads will be deserialized without validation...

SGLang: Unauthenticated RCE via --enable-custom-logit-processor

SGLang's multimodal generation runtime is vulnerable to unauthenticated remote code execution when the --enable-custom-logit-processor option is enabled, as Python objects loaded via dill.loads will be deserialized without validation...

SPARK: Security Knowledge Priming and Representation-Guided Knowledge Activation for LLM-Based Secure Code Generation

Large language models routinely generate code with exploitable security flaws. Prior literature attributes this limitation to a lack of security expertise, steering current defense mechanisms toward heavy fine-tuning or external knowledge retrieval, which introduces significant computational...

CVE-2026-7304

SGLangs multimodal generation runtime is vulnerable to unauthenticated remote code execution when the --enable-custom-logit-processor option is enabled, as Python objects loaded via dill.loads will be deserialized without validation...

SGLang: Unauthenticated RCE via --enable-custom-logit-processor

SGLang's multimodal generation runtime is vulnerable to unauthenticated remote code execution when the --enable-custom-logit-processor option is enabled, as Python objects loaded via dill.loads will be deserialized without validation...

Deserialization of Untrusted Data

Overview sglang is a SGLang is a fast serving framework for large language models and vision language models. Affected versions of this package are vulnerable to Deserialization of Untrusted Data via the --enable-custom-logit-processor option, which allows untrusted Python objects to be...

GHSA-36M8-W8QF-G76P SGLang: Unauthenticated RCE via --enable-custom-logit-processor

SGLang's multimodal generation runtime is vulnerable to unauthenticated remote code execution when the --enable-custom-logit-processor option is enabled, as Python objects loaded via dill.loads will be deserialized without validation...

CVE-2026-7304

SGLangs multimodal generation runtime is vulnerable to unauthenticated remote code execution when the --enable-custom-logit-processor option is enabled, as Python objects loaded via dill.loads will be deserialized without validation...

CVE-2026-7304 CVE-2026-7304

SGLangs multimodal generation runtime is vulnerable to unauthenticated remote code execution when the --enable-custom-logit-processor option is enabled, as Python objects loaded via dill.loads will be deserialized without validation...

CVE-2026-7304 CVE-2026-7304

SGLangs multimodal generation runtime is vulnerable to unauthenticated remote code execution when the --enable-custom-logit-processor option is enabled, as Python objects loaded via dill.loads will be deserialized without validation...

EUVD-2026-30766

SGLangs multimodal generation runtime is vulnerable to unauthenticated remote code execution when the --enable-custom-logit-processor option is enabled, as Python objects loaded via dill.loads will be deserialized without validation...

CVE-2026-7304

The SGLang multimodal generation runtime is vulnerable to unauthenticated remote code execution (RCE) when the --enable-custom-logit-processor option is active. The root cause is the deserialization of Python objects via dill.loads() without proper validation. This vulnerability has been demonstr...

CVE-2026-7304 CVE-2026-7304

SGLangs multimodal generation runtime is vulnerable to unauthenticated remote code execution when the --enable-custom-logit-processor option is enabled, as Python objects loaded via dill.loads will be deserialized without validation...

CVE-2026-7304

SGLangs multimodal generation runtime is vulnerable to unauthenticated remote code execution when the --enable-custom-logit-processor option is enabled, as Python objects loaded via dill.loads will be deserialized without validation...

CVE-2026-7304: Deserialization of Untrusted Data

SGLangs multimodal generation runtime is vulnerable to unauthenticated remote code execution when the --enable-custom-logit-processor option is enabled, as Python objects loaded via dill.loads will be deserialized without validation...

sglang 代码问题漏洞

SGLang is a programming language and runtime system developed by SGL-project, aimed at accelerating large model inference. SGLang has code vulnerabilities; these vulnerabilities arise when the --enable-custom-logit-processor option is enabled, resulting in unvalidated deserialization of Python...

PT-2026-41670

Name of the Vulnerable Software and Affected Versions SGLang affected versions not specified Description The multimodal generation runtime allows unauthenticated remote code execution when the --enable-custom-logit-processor option is active. This occurs because Python objects loaded through the...

Jailbreaking LLMs Via Calibration

Safety alignment in Large Language Models LLMs often creates a systematic discrepancy between a model's aligned output and the underlying pre-aligned data distribution. We propose a framework in which the effect of safety alignment on next-token prediction is modeled as a systematic distortion of...