Secure Speculative Decoding for Large Language Models

Speculative decoding accelerates inference for a large language model LLM, referred to as the target model, by first using a smaller model, referred to as the draft model, to generate candidate tokens and then verifying them with the target model for acceptance or rejection. Prior studies primari...

Improper Handling of Exceptional Conditions

Overview vllm is an A high-throughput and memory-efficient inference and serving engine for LLMs Affected versions of this package are vulnerable to Improper Handling of Exceptional Conditions via validatetokens in the structured output backend, when a request carrying invalid structured-output...

CVE-2026-105757 vLLM: Structured-output request errors escape the request boundary and terminate the shared EngineCore — engine-fatal denial of service (3 sites)

vLLM is an inference and serving engine for large language models. Prior to 0.30.0, structured-output request failures can escape request-scoped validation and reach the EngineCore fatal-error path. A per-request backend mismatch can re-raise a grammar compilation exception, padding produced by t...

SUSE CVE-2026-44223

vLLM is an inference and serving engine for large language models LLMs. From 0.18.0 to before 0.20.0, the extracthiddenstates speculative decoding proposer in vLLM returns a tensor with an incorrect shape after the first decode step, causing a RuntimeError that crashes the EngineCore process. The...

CVE-2026-44223

A flaw was found in vLLM, an inference and serving engine for large language models LLMs. The extracthiddenstates speculative decoding proposer returns a tensor with an incorrect shape after the first decode step. This can be triggered by a remote attacker sending a request that uses sampling...

EUVD-2026-41921

vLLM has Remote DoS via Invalid Recovered Token Reinjection...

CVE-2026-54234

A flaw was found in vLLM, a high-throughput and memory-efficient inference and serving engine for Large Language Models LLMs. A remote attacker can exploit this vulnerability by sending a specially crafted multi-request speculative decoding workload through public gRPC Generate and Abort endpoint...

CVE-2026-54234 vLLM: Remote DoS in vLLM via Invalid Recovered Token Reinjection

vLLM is a high-throughput and memory-efficient inference and serving engine for LLMs. Prior to 0.24.0, a frontend-legal multi-request speculative decoding workload can cause the rejection sampler to produce a recovered token equal to the model vocabulary size boundary value, which is then convert...

PT-2026-55997

Name of the Vulnerable Software and Affected Versions vLLM versions prior to 0.24.0 Description A flaw in the rejection sampler during multi-request speculative decoding workloads allows the production of a recovered token equal to the model vocabulary size boundary value. This value is converted...

CVE-2026-54234: Improper Input Validation

vLLM is a high-throughput and memory-efficient inference and serving engine for LLMs. Prior to 0.24.0, a frontend-legal multi-request speculative decoding workload can cause the rejection sampler to produce a recovered token equal to the model vocabulary size boundary value, which is then convert...

CVE-2026-44223

vLLM is an inference and serving engine for large language models LLMs. From 0.18.0 to before 0.20.0, the extracthiddenstates speculative decoding proposer in vLLM returns a tensor with an incorrect shape after the first decode step, causing a RuntimeError that crashes the EngineCore process. The...

CVE-2026-44223

vLLM contains a vulnerability (CVE-2026-44223) where the extract_hidden_states speculative decoding pathway can crash the EngineCore process if any request uses penalty parameters (repetition_penalty, frequency_penalty, or presence_penalty). The issue arises from an incorrect tensor shape after t...

CVE-2026-44223 vLLM: extract_hidden_states speculative decoding crashes server on any request with penalty parameters

vLLM is an inference and serving engine for large language models LLMs. From 0.18.0 to before 0.20.0, the extracthiddenstates speculative decoding proposer in vLLM returns a tensor with an incorrect shape after the first decode step, causing a RuntimeError that crashes the EngineCore process. The...

CVE-2026-44223 vLLM: extract_hidden_states speculative decoding crashes server on any request with penalty parameters

vLLM is an inference and serving engine for large language models LLMs. From 0.18.0 to before 0.20.0, the extracthiddenstates speculative decoding proposer in vLLM returns a tensor with an incorrect shape after the first decode step, causing a RuntimeError that crashes the EngineCore process. The...

vLLM security vulnerabilities

vLLM is an open-source LLM-based inference and service engine that features high throughput and efficient memory usage. Versions of vLLM prior to 0.20.0 contained a security vulnerability. This vulnerability stemmed from the extracthiddenstates speculative decoding proposal, which returned tensor...

vLLM: extract_hidden_states speculative decoding crashes server on any request with penalty parameters

Summary The extracthiddenstates speculative decoding proposer in vLLM returns a tensor with an incorrect shape after the first decode step, causing a RuntimeError that crashes the EngineCore process. The crash is triggered when any request in the batch uses sampling penalty parameters...

Incorrect Type Conversion or Cast

Overview vllm is an A high-throughput and memory-efficient inference and serving engine for LLMs Affected versions of this package are vulnerable to Incorrect Type Conversion or Cast through the extracthiddenstates speculative decoding. An attacker can cause the server to crash and disrupt servic...

GHSA-83VM-P52W-F9PW vLLM: extract_hidden_states speculative decoding crashes server on any request with penalty parameters

Summary The extracthiddenstates speculative decoding proposer in vLLM returns a tensor with an incorrect shape after the first decode step, causing a RuntimeError that crashes the EngineCore process. The crash is triggered when any request in the batch uses sampling penalty parameters...

PT-2026-38288

Name of the Vulnerable Software and Affected Versions vLLM versions 0.18.0 through 0.19.1 Description The extract hidden states speculative decoding proposer returns a tensor with an incorrect shape after the first decode step, leading to a RuntimeError that crashes the EngineCore process. This...

Side-Channel Attacks Against LLMs

Here are three papers describing different side-channel attacks against LLMs. "Remote Timing Attacks on Efficient Language Model Inference": Abstract: Scaling up language models has significantly increased their capabilities. But larger models are slower models, and so there is now an extensive...