CVE-2026-71486: vLLM Derender Endpoints Decode Caller-supplied GenerateResponse Vulnerability

{
  "schema_version" : "1.4.0",
  "id" : "GHSA-8737-qx52-hjff",
  "modified" : "2026-09-04T21:32:08Z",
  "published" : "2026-09-04T21:32:07Z",
  "aliases" : [ "CVE-2026-71486" ],
  "summary" : "vLLM: Derender endpoints decode caller-supplied GenerateResponse token IDs without output bounds",
  "details" : "## Summary\n\nThe `/v1/completions/derender` and `/v1/chat/completions/derender` endpoints accept caller-supplied `GenerateResponse` objects and postprocess every nested `choices[*].token_ids` list directly. Unlike the normal render/generate path, derender does not enforce model context length, resolved `max_tokens`, `max_num_seqs`, choice-count, or response-size bounds before detokenizing and returning the supplied token IDs. An authenticated API client can therefore make the CPU-only render frontend, or any server exposing these `/v1` derender routes, spend CPU and memory proportional to attacker-chosen generated-output-shaped JSON rather than to a bounded generation result.\n\n## Technical Details\n\nThe render router registers `/v1/chat/completions/derender` and `/v1/completions/derender` in `vllm/entrypoints/serve/render/api_router.py`, and the OpenAI API server attaches this router whenever `\"generate\"` or `\"render\"` is in `supported_tasks` (`vllm/entrypoints/openai/api_server.py`). The routes are under `/v1`, so they are part of the OpenAI-compatible HTTP API surface and are protected by the API-key middleware when `--api-key` is configured.\n\nThe request types trust generated-output-shaped data from the client. In `vllm/entrypoints/serve/disagg/protocol.py`, `GenerateResponseChoice` accepts `token_ids: list[int] | None = None`, `GenerateResponse` accepts `choices: list[GenerateResponseChoice]`, and `DerenderCompletionRequest` accepts `generate_responses: list[GenerateResponse]`. These fields have no max length, max item count, or relationship to a prior `GenerateRequest`.\n\nThe sink is `OnlineDerenderer`. `derender_completion()` iterates every supplied `generate_responses` entry and every nested choice, calls `tokenizer.decode(choice.token_ids, skip_special_tokens=True)`, appends the decoded text to the response choices, and increments `total_completion_tokens` from the same supplied list length. `derender_chat()` has the same shape for a single supplied `generate_response`, and can also feed the decoded text into tool/reasoning parsers when a parser and `chat_request` are present. `ServingRender.derender_completion_response()` calls `online_derenderer.derender_completion(request.generate_responses, request.prompt_tokens)` before applying any completion-level validation beyond the model check.\n\nNormal render and generation paths derive output limits from `max_model_len`, the rendered prompt length, request `max_tokens` / `max_completion_tokens`, and scheduler limits. Derender bypasses that invariant because it accepts the already-generated output shape directly from the HTTP caller. The missing invariant is: derender should only postprocess bounded generated output, and client-supplied derender payloads must be rejected if their nested generated token/logprob structures exceed the same limits that generation would have enforced.\n\n## PoV\n\nThe following bounded PoV can be run from a current vLLM checkout containing PR `#43606`. It asserts the current source facts for the derender routes, unchecked request fields, and decode sink, then simulates the same derender loop with a counting tokenizer. The negative control is a one-choice, 32-token response. The amplified payload keeps the test bounded but demonstrates that all decoded work and returned text scale directly with caller-supplied `GenerateResponse` contents.\n\n```python\n#!/usr/bin/env python3\nimport subprocess\nfrom dataclasses import dataclass\nfrom pathlib import Path\n\nSOURCE = Path(\".\")\n\ndef require_source_fact(path: str, needles: list[str]) -> None:\n    text = (SOURCE / path).read_text()\n    missing = [needle for needle in needles if needle not in text]\n    if missing:\n        raise AssertionError(f\"{path} missing expected facts: {missing}\")\n\ndef source_head() -> str:\n    return subprocess.check_output([\"git\", \"rev-parse\", \"HEAD\"], cwd=SOURCE, text=True).strip()\n\n@dataclass\nclass Choice:\n    index: int\n    token_ids: list[int]\n\n@dataclass\nclass GenerateResponse:\n    request_id: str\n    choices: list[Choice]\n\nclass CountingTokenizer:\n    def __init__(self) -> None:\n        self.decode_calls = 0\n        self.decoded_ids = 0\n    def decode(self, token_ids: list[int], *, skip_special_tokens: bool = True) -> str:\n        self.decode_calls += 1\n        self.decoded_ids += len(token_ids)\n        return \"x\" * len(token_ids)\n\ndef derender_completion_like_current_head(generate_responses: list[GenerateResponse], tokenizer: CountingTokenizer) -> tuple[int, int, int]:\n    output_chars = 0\n    choices = 0\n    total_completion_tokens = 0\n    for gen in generate_responses:\n        for choice in gen.choices:\n            if not choice.token_ids:\n                raise ValueError(\"choice has empty or null token_ids\")\n            decoded_text = tokenizer.decode(choice.token_ids, skip_special_tokens=True)\n            output_chars += len(decoded_text)\n            total_completion_tokens += len(choice.token_ids)\n            choices += 1\n    return choices, total_completion_tokens, output_chars\n\ndef make_payload(responses: int, choices_per_response: int, tokens_per_choice: int) -> list[GenerateResponse]:\n    token_ids = [42] * tokens_per_choice\n    return [GenerateResponse(request_id=f\"gen-{r}\", choices=[Choice(index=c, token_ids=list(token_ids)) for c in range(choices_per_response)]) for r in range(responses)]\n\ndef run_case(name: str, payload: list[GenerateResponse]) -> None:\n    tokenizer = CountingTokenizer()\n    choices, completion_tokens, output_chars = derender_completion_like_current_head(payload, tokenizer)\n    print(f\"{name}: responses={len(payload)} choices={choices} decode_calls={tokenizer.decode_calls} decoded_token_ids={tokenizer.decoded_ids} completion_tokens={completion_tokens} output_chars={output_chars}\")\n\nrequire_source_fact(\"vllm/entrypoints/serve/render/api_router.py\", ['\"/v1/completions/derender\"', '\"/v1/chat/completions/derender\"', \"app.include_router(router)\"])\nrequire_source_fact(\"vllm/entrypoints/serve/disagg/protocol.py\", [\"class GenerateResponseChoice(BaseModel):\", \"token_ids: list[int] | None = None\", \"class GenerateResponse(BaseModel):\", \"choices: list[GenerateResponseChoice]\", \"class DerenderCompletionRequest(BaseModel):\", \"generate_responses: list[GenerateResponse]\"])\nrequire_source_fact(\"vllm/renderers/online_derenderer.py\", [\"async def derender_completion(\", \"for gen, pt in zip(generate_responses, prompt_tokens_list):\", \"for choice in gen.choices:\", \"decoded_text = tokenizer.decode(\", \"total_completion_tokens += len(choice.token_ids)\"])\nprint(\"source_checks=ok\")\nprint(f\"source_head={source_head()}\")\nrun_case(\"negative_control\", make_payload(responses=1, choices_per_response=1, tokens_per_choice=32))\nrun_case(\"amplified_payload\", make_payload(responses=16, choices_per_response=4, tokens_per_choice=8192))\nprint(\"observation=derender decodes every caller-supplied token id before any max_model_len, max_tokens, max_num_seqs, or response-size check\")\n```\n\n\n## Impact\n\nAn attacker with access to the `/v1` API can send derender requests that consume CPU and memory in the frontend/postprocessing process and can cause large responses unrelated to any bounded generation. In disaggregated deployments, this affects the CPU-only render frontend; in servers where the render router is attached alongside generation, it affects the same OpenAI-compatible server process that handles normal client traffic. This can degrade availability for other clients sharing the process.\n\nLikely CWE: CWE-400 (Uncontrolled Resource Consumption) / CWE-770 (Allocation of Resources Without Limits or Throttling). Conservative CVSS v3.1: `CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:L` (4.3). This is not Low severity because a regular network API client can induce availability impact in a shared service without local access, invalid model artifacts, or special runtime privileges. If the server is deployed without API-key enforcement for `/v1`, the privileges component becomes `PR:N`.\n\n## Suggested Fix\n\nValidate derender payloads before any detokenization or parser invocation. Apply bounded limits to `generate_response(s)`, `choices`, `token_ids`, `prompt_logprobs`, `logprobs.content`, `top_logprobs`, and `routed_experts` that are at least as strict as the corresponding generation-side limits. For completions, reject `generate_responses` counts above the number of prompts that `/v1/completions/render` would have produced, and reject total nested choice counts above `max_num_seqs` / `n` limits. For each choice, reject `token_ids` longer than the resolved output-token budget, or require derender callers to submit the original bounded `GenerateRequest` / sampling metadata and validate the `GenerateResponse` against it before decoding.\n\nAdd regression tests for both derender endpoints. The tests should show that a normal bounded derender payload succeeds, while oversized `generate_responses`, oversized `choices`, oversized `token_ids`, and oversized logprob/top-logprob structures are rejected before `tokenizer.decode()` or parser execution.\n\n## Affected Package/Versions\n\nConfirmed affected: current main at `ddd3855a28a561a5bb54d380c6e6b8b1e883cc4a` and downstream/nightly builds that include the derender endpoints introduced by PR `#43606`. The derender router, request models, decode sink, render serving bridge, and OpenAI API router attachment have no relevant diff from `00e045b7c7b82599f626779e111233abd4d0a64e` to `ddd3855a28a561a5bb54d380c6e6b8b1e883cc4a`.\n\nLatest release checked: `v0.23.0`, published on `2026-06-15`. Its `vllm/entrypoints/serve/render/api_router.py` does not expose `/v1/completions/derender` or `/v1/chat/completions/derender`, so `v0.23.0` was not confirmed affected.\n\n## Advisory History\n\nPR `#43606` (\"[Render] Add `/derender` endpoints for disaggregated postprocessing\") introduced the derender endpoints on main. PR `#44285` later refactored the render serving code, and current head still contains the unchecked derender flow.\n\nPublic issue search for `derender GenerateResponse token_ids` returned no reports. Public search for `\"/v1/completions/derender\"` returned the derender feature RFC `#42729` and unrelated bugs, but no size-bound, DoS, or generated-output postprocessing issue.\n\nRelated public request-fanout and resource-bound advisories are distinct:\n\n- `GHSA-3mwp-wvh9-7528` covers an unbounded `n` parameter on the normal OpenAI completion/chat generation routes. Its root cause is missing upper-bound validation for generated sequence count, its sink is request fanout and request-object copying into the async engine path before scheduling, its precondition is a caller-controlled `n`, and its fix surface is a cap on generated sequence count. This report reaches `/v1/completions/derender` and `/v1/chat/completions/derender`, not the normal generate routes; its root cause is unchecked caller-supplied `GenerateResponse` / `choices` / `token_ids` structures, its sink is `OnlineDerenderer` detokenization and response construction after generation, its precondition is access to the derender API with generated-output-shaped JSON, and its fix surface is derender payload validation before decode.\n- PR `#45390` includes the `GHSA-83mh-6mwq-3hg9` batch-message fanout fix class: it bounds the outer `BatchChatCompletionRequest.messages` conversation list to prevent one request from creating many conversation/request objects before normal generation. This report has no batch conversation list and does not rely on `n`; one derender request can instead supply oversized nested `GenerateResponse` choices and token IDs that are detokenized and returned directly. A batch-message `max_length` limit would not bound derender `generate_response(s)` or per-choice token/logprob structures.\n\nThe completed local report titled \"Explicit truncation_side disables tokenizer-level prompt truncation\" is also distinct. That report used `/v1/completions` and `/v1/chat/completions` with ordinary prompt text plus `truncate_prompt_tokens` and explicit `truncation_side`; its root cause was the renderer omitting tokenizer-level `max_length` and the pre-tokenization character guard before post-token slicing; its sink was prompt tokenization; and its fix surface was preserving tokenizer-level truncation or rejecting over-budget prompts before tokenization. This derender report uses `/v1` derender routes, has no prompt text tokenization or truncation-side control, starts from caller-supplied generated-output token IDs, and needs aggregate bounds on derender `generate_response(s)`, choices, token IDs, logprobs, parser inputs, and response construction before detokenization.\n\nOther adjacent vLLM advisories for Rust/gRPC token-id and logprob bounds, structured-output grammar amplification, repetition-detection windows, and pooling/rerank batch fanout are distinct. Those issues affect Rust/gRPC request conversion, grammar compilation, scheduler loops, or engine fanout. This issue affects `/v1` derender postprocessing of caller-supplied generated-output objects and requires derender-specific request validation before detokenization.\n\n## Resources\n\n- `vllm/entrypoints/serve/render/api_router.py`\n- `vllm/entrypoints/serve/disagg/protocol.py`\n- `vllm/renderers/online_derenderer.py`\n- `vllm/entrypoints/serve/render/serving.py`\n- `vllm/entrypoints/openai/api_server.py`\n- PR `#43606`: `https://github.com/vllm-project/vllm/pull/43606`\n- PR `#44285`: `https://github.com/vllm-project/vllm/pull/44285`\n- `GHSA-3mwp-wvh9-7528`: `https://github.com/vllm-project/vllm/security/advisories/GHSA-3mwp-wvh9-7528`\n- PR `#45390`: `https://github.com/vllm-project/vllm/pull/45390`\n- Release `v0.23.0`: `https://github.com/vllm-project/vllm/releases/tag/v0.23.0`",
  "severity" : [ {
    "type" : "CVSS_V3",
    "score" : "CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:L"
  } ],
  "affected" : [ {
    "package" : {
      "ecosystem" : "PyPI",
      "name" : "vllm"
    },
    "ranges" : [ {
      "type" : "ECOSYSTEM",
      "events" : [ {
        "introduced" : "0"
      }, {
        "fixed" : "0.26.0"
      } ]
    } ]
  } ],
  "references" : [ {
    "type" : "WEB",
    "url" : "https://github.com/vllm-project/vllm/security/advisories/GHSA-8737-qx52-hjff"
  }, {
    "type" : "ADVISORY",
    "url" : "https://nvd.nist.gov/vuln/detail/CVE-2026-71486"
  }, {
    "type" : "WEB",
    "url" : "https://github.com/vllm-project/vllm/pull/47260"
  }, {
    "type" : "WEB",
    "url" : "https://github.com/vllm-project/vllm/commit/8e61b646e2d157f9b93451fa048f9c8530c8a67b"
  }, {
    "type" : "PACKAGE",
    "url" : "https://github.com/vllm-project/vllm"
  }, {
    "type" : "WEB",
    "url" : "https://github.com/vllm-project/vllm/releases/tag/v0.26.0"
  } ],
  "database_specific" : {
    "cwe_ids" : [ "CWE-400", "CWE-770" ],
    "severity" : "MODERATE",
    "github_reviewed" : true,
    "github_reviewed_at" : "2026-09-04T21:32:07Z",
    "nvd_published_at" : "2026-08-17T20:16:45Z"
  }
}