Vllm
by Vllm
Source repositories
CVEs (145)
| CVE | Vendor / Product | Sev | Risk | CVSS | EPSS | KEV | Published | Description |
|---|---|---|---|---|---|---|---|---|
| CVE-2026-92220 | Med | 0.27 | 5.3 | 0.01 | Sep 16, 2026 | A vulnerability was found in vllm-project vLLM 0.26.0/0.27.0. Affected is the function MoRIIOConnectorScheduler.request_finished/MoRIIOConnectorWorker.get_finished/MoRIIOWrapper._handle_release_message of the file vllm/distributed/kv_transfer/kv_connector/v1/moriio/moriio_connect… | ||
| CVE-2026-78684 | Med | 0.27 | 5.3 | 0.01 | Aug 25, 2026 | vLLM before 0.27.0 fails to properly classify DeepStream as a GPU backend and omits pixel-limit enforcement in its decode path. Unauthenticated attackers can activate DeepStream at request time to initialize the process-wide GPU decode pool and submit video that bypasses… | ||
| CVE-2026-73558 | Med | 0.27 | 5.3 | 0.00 | Aug 13, 2026 | 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… | ||
| CVE-2026-73557 | Med | 0.27 | 5.3 | 0.00 | Aug 13, 2026 | vLLM is an inference and serving engine for large language models. From 0.20.2rc0 until 0.26.0, safe_load_prompt_embeds in vllm/renderers/embed_utils.py uses torch.sparse.check_sparse_tensor_invariants, whose process-global save, enable, and restore state can be raced by… | ||
| CVE-2026-73556 | Med | 0.27 | 5.3 | 0.01 | Aug 13, 2026 | vLLM is an inference and serving engine for large language models. Prior to 0.26.0, the structured_outputs.regex parameter in vllm/v1/structured_output/backend_lm_format_enforcer.py is passed to lmformatenforcer.RegexParser without compile_regex_with_timeout or validation in… | ||
| CVE-2026-73555 | Med | 0.27 | 5.3 | 0.00 | Aug 13, 2026 | vLLM is an inference and serving engine for large language models. Prior to 0.26.0, the validation_exception_handler in vllm/entrypoints/openai/server_utils.py converts FastAPI RequestValidationError objects with str(exc), and sanitize_message in vllm/entrypoints/utils.py does… | ||
| CVE-2026-12491 | Med | 0.24 | 4.8 | 0.00 | Jun 17, 2026 | A flaw was found in vLLM, an open-source library for large language model inference. This vulnerability arises from improper handling of image metadata, specifically EXIF orientation and PNG transparency (tRNS) data, during image processing. When images are converted to RGB,… | ||
| CVE-2026-92365 | Med | 0.21 | 4.3 | 0.01 | Sep 16, 2026 | A vulnerability was found in vllm-project vllm up to 0.29.0. Affected by this issue is some unknown functionality of the file vllm/v1/sample/thinking_budget_state.py. The manipulation results in inefficient algorithmic complexity. It is possible to launch the attack remotely.… | ||
| CVE-2026-90878 | Med | 0.21 | 4.3 | 0.01 | Sep 15, 2026 | A vulnerability was determined in vllm-project vLLM up to 0.27.1. This affects an unknown part of the file /v1/chat/completions of the component Jinja Template Rendering. This manipulation of the argument chat_template causes resource consumption. The attack can be initiated… | ||
| CVE-2026-71486 | Med | 0.21 | 4.3 | 0.00 | Aug 17, 2026 | vLLM is an inference and serving engine for large language models. Prior to 0.26.0, the /v1/completions/derender and /v1/chat/completions/derender endpoints accept caller-supplied GenerateResponse objects whose generate_responses, choices, token_ids, prompt_logprobs,… | ||
| CVE-2025-71379 | Med | 0.21 | 4.3 | 0.00 | Jun 20, 2026 | vLLM versions >= 0.6.3 and < 0.9.0 contain multiple regular expression denial of service (ReDoS) vulnerabilities. Several regex patterns — in vllm/lora/utils.py, the phi4mini tool parser, and the OpenAI-compatible serving chat endpoint — are susceptible to catastrophic… | ||
| CVE-2026-105755 | Med | 0.20 | 4.2 | 0.00 | Oct 5, 2026 | vLLM is an inference and serving engine for large language models. Prior to 0.30.0, flash late-interaction scoring at the /score and /rerank endpoints derives each worker's query_key value from the caller-controlled X-Request-Id header. A concurrent request that reuses a… | ||
| CVE-2025-46722 | Med | 0.20 | 4.2 | 0.00 | May 29, 2025 | vLLM is an inference and serving engine for large language models (LLMs). In versions starting from 0.7.0 to before 0.9.0, in the file vllm/multimodal/hasher.py, the MultiModalHasher class has a security and data integrity issue in its image hashing method. Currently, it… | ||
| CVE-2025-61620 | med | 0.19 | — | 0.00 | Oct 7, 2025 | ### Summary A resource-exhaustion (denial-of-service) vulnerability exists in multiple endpoints of the OpenAI-Compatible Server due to the ability to specify Jinja templates via the `chat_template` and `chat_template_kwargs` parameters. If an attacker can supply these… | ||
| CVE-2026-114049 | Low | 0.17 | 3.7 | 0.00 | Oct 21, 2026 | vLLM before 0.31.0 contains a cross-request mask leak in InputBatch.condense() that allows remote attackers to constrain other clients' outputs by leaving stale allowed_token_ids rows. Attackers can submit concurrent constrained generate requests, including via unauthenticated… | ||
| CVE-2026-100649 | Low | 0.17 | 3.7 | 0.00 | Sep 26, 2026 | vLLM before 0.29.0 contains a resource-limit bypass vulnerability in PyNvVideoCodec decoder allocation where sampler subclass shadowing allows independent counter increments. Unauthenticated attackers can select different sampler subclasses in video requests to exceed configured… | ||
| CVE-2026-93841 | Low | 0.17 | 3.7 | 0.00 | Sep 18, 2026 | vLLM through 0.29.0 contains a memory corruption vulnerability in the Triton _bincount_kernel where prompt token IDs index the penalty prompt-presence bitset without bounds checking against vocabulary size. Attackers can submit multimodal audio requests with tokens equal to… | ||
| CVE-2026-93840 | Low | 0.17 | 3.7 | 0.00 | Sep 18, 2026 | vLLM before 0.29.0 validates allowed_token_ids against tokenizer length instead of model output logits width in SamplingParams._validate_allowed_token_ids(). Attackers can supply token IDs above the output vocabulary that pass validation, causing LogitBiasState to corrupt GPU… | ||
| CVE-2026-115917 | Low | 0.15 | 3.4 | 0.00 | Oct 26, 2026 | vLLM is an inference and serving engine for large language models. Prior to 0.32.0, the OpenAI-compatible server configures Starlette CORSMiddleware with allowed_origins set to a wildcard by default. A malicious web page can use a victim's browser to reach an unauthenticated… | ||
| CVE-2026-90713 | Low | 0.14 | 3.3 | 0.00 | Sep 14, 2026 | A security flaw has been discovered in vllm-project vLLM up to 0.29.0. The affected element is the function TiktokenTokenizer::new of the file rust/src/text/src/backend/hf/mod.rs of the component tiktoken vocab File Handler. The manipulation results in denial of service. The… |
- risk 0.27cvss 5.3epss 0.01
A vulnerability was found in vllm-project vLLM 0.26.0/0.27.0. Affected is the function MoRIIOConnectorScheduler.request_finished/MoRIIOConnectorWorker.get_finished/MoRIIOWrapper._handle_release_message of the file vllm/distributed/kv_transfer/kv_connector/v1/moriio/moriio_connect…
- risk 0.27cvss 5.3epss 0.01
vLLM before 0.27.0 fails to properly classify DeepStream as a GPU backend and omits pixel-limit enforcement in its decode path. Unauthenticated attackers can activate DeepStream at request time to initialize the process-wide GPU decode pool and submit video that bypasses…
- risk 0.27cvss 5.3epss 0.00
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…
- risk 0.27cvss 5.3epss 0.00
vLLM is an inference and serving engine for large language models. From 0.20.2rc0 until 0.26.0, safe_load_prompt_embeds in vllm/renderers/embed_utils.py uses torch.sparse.check_sparse_tensor_invariants, whose process-global save, enable, and restore state can be raced by…
- risk 0.27cvss 5.3epss 0.01
vLLM is an inference and serving engine for large language models. Prior to 0.26.0, the structured_outputs.regex parameter in vllm/v1/structured_output/backend_lm_format_enforcer.py is passed to lmformatenforcer.RegexParser without compile_regex_with_timeout or validation in…
- risk 0.27cvss 5.3epss 0.00
vLLM is an inference and serving engine for large language models. Prior to 0.26.0, the validation_exception_handler in vllm/entrypoints/openai/server_utils.py converts FastAPI RequestValidationError objects with str(exc), and sanitize_message in vllm/entrypoints/utils.py does…
- risk 0.24cvss 4.8epss 0.00
A flaw was found in vLLM, an open-source library for large language model inference. This vulnerability arises from improper handling of image metadata, specifically EXIF orientation and PNG transparency (tRNS) data, during image processing. When images are converted to RGB,…
- risk 0.21cvss 4.3epss 0.01
A vulnerability was found in vllm-project vllm up to 0.29.0. Affected by this issue is some unknown functionality of the file vllm/v1/sample/thinking_budget_state.py. The manipulation results in inefficient algorithmic complexity. It is possible to launch the attack remotely.…
- risk 0.21cvss 4.3epss 0.01
A vulnerability was determined in vllm-project vLLM up to 0.27.1. This affects an unknown part of the file /v1/chat/completions of the component Jinja Template Rendering. This manipulation of the argument chat_template causes resource consumption. The attack can be initiated…
- risk 0.21cvss 4.3epss 0.00
vLLM is an inference and serving engine for large language models. Prior to 0.26.0, the /v1/completions/derender and /v1/chat/completions/derender endpoints accept caller-supplied GenerateResponse objects whose generate_responses, choices, token_ids, prompt_logprobs,…
- risk 0.21cvss 4.3epss 0.00
vLLM versions >= 0.6.3 and < 0.9.0 contain multiple regular expression denial of service (ReDoS) vulnerabilities. Several regex patterns — in vllm/lora/utils.py, the phi4mini tool parser, and the OpenAI-compatible serving chat endpoint — are susceptible to catastrophic…
- risk 0.20cvss 4.2epss 0.00
vLLM is an inference and serving engine for large language models. Prior to 0.30.0, flash late-interaction scoring at the /score and /rerank endpoints derives each worker's query_key value from the caller-controlled X-Request-Id header. A concurrent request that reuses a…
- risk 0.20cvss 4.2epss 0.00
vLLM is an inference and serving engine for large language models (LLMs). In versions starting from 0.7.0 to before 0.9.0, in the file vllm/multimodal/hasher.py, the MultiModalHasher class has a security and data integrity issue in its image hashing method. Currently, it…
- risk 0.19cvss —epss 0.00
### Summary A resource-exhaustion (denial-of-service) vulnerability exists in multiple endpoints of the OpenAI-Compatible Server due to the ability to specify Jinja templates via the `chat_template` and `chat_template_kwargs` parameters. If an attacker can supply these…
- risk 0.17cvss 3.7epss 0.00
vLLM before 0.31.0 contains a cross-request mask leak in InputBatch.condense() that allows remote attackers to constrain other clients' outputs by leaving stale allowed_token_ids rows. Attackers can submit concurrent constrained generate requests, including via unauthenticated…
- risk 0.17cvss 3.7epss 0.00
vLLM before 0.29.0 contains a resource-limit bypass vulnerability in PyNvVideoCodec decoder allocation where sampler subclass shadowing allows independent counter increments. Unauthenticated attackers can select different sampler subclasses in video requests to exceed configured…
- risk 0.17cvss 3.7epss 0.00
vLLM through 0.29.0 contains a memory corruption vulnerability in the Triton _bincount_kernel where prompt token IDs index the penalty prompt-presence bitset without bounds checking against vocabulary size. Attackers can submit multimodal audio requests with tokens equal to…
- risk 0.17cvss 3.7epss 0.00
vLLM before 0.29.0 validates allowed_token_ids against tokenizer length instead of model output logits width in SamplingParams._validate_allowed_token_ids(). Attackers can supply token IDs above the output vocabulary that pass validation, causing LogitBiasState to corrupt GPU…
- risk 0.15cvss 3.4epss 0.00
vLLM is an inference and serving engine for large language models. Prior to 0.32.0, the OpenAI-compatible server configures Starlette CORSMiddleware with allowed_origins set to a wildcard by default. A malicious web page can use a victim's browser to reach an unauthenticated…
- risk 0.14cvss 3.3epss 0.00
A security flaw has been discovered in vllm-project vLLM up to 0.29.0. The affected element is the function TiktokenTokenizer::new of the file rust/src/text/src/backend/hf/mod.rs of the component tiktoken vocab File Handler. The manipulation results in denial of service. The…
Page 7 of 8