{"id":"CVE-2026-116177","kind":"vulnerability","aliases":["CVE-2026-116177"],"related":[],"title":"vLLM: crafted image triggers memory exhaustion, denying service to other users","description":"vLLM from 0.10.2 before 0.25.0 contains an uncontrolled resource consumption vulnerability that allows unauthenticated attackers to exhaust memory because the Nemotron-VL processors globally set Image.MAX_IMAGE_PIXELS to None. Attackers can submit a pathological decompression-bomb image to /v1/chat/completions to OOM-kill the worker and deny service to co-tenants.","summary":"vLLM before 0.25.0 lets unauthenticated users submit a decompression-bomb image to the chat completions endpoint, because the Nemotron-VL processors globally disable Pillow's pixel limit. The oversized image exhausts memory, OOM-kills the worker and takes the service down for co-tenants.","severity":"high","cvss":8.7,"epss":null,"exploited":false,"exploited_at":null,"products":[],"ecosystems":[],"tags":["ai-inference","ai-tooling","python","self-hosted","vllm"],"not_affected_if":["no-nemotron-vl-models","text-only-serving"],"how_to_check":"- Run `pip show vllm` (or `python -c \"import vllm; print(vllm.__version__)\"`) and check whether the version is between `0.10.2` and `0.25.0`\n- Check whether the deployed model is a Nemotron-VL model, since the flaw is in those processors\n- unknown","fix":"- Upgrade vLLM to `0.25.0` or later\n- Until upgraded, limit request body size and memory per worker at the gateway or container level","status":"open","published_at":"2026-10-25T14:16:57+00:00","first_seen_at":"2026-10-25T14:41:23+00:00","triaged_at":"2026-10-25T14:42:17+00:00","updated_at":"2026-10-25T14:42:17+00:00","category":"ai-tooling","confidence":0.6,"triage":{"at":"2026-10-25T14:42:17Z","model":"claude-haiku-5-5","output":{"fix":"- Upgrade vLLM to `0.25.0` or later\n- Until upgraded, limit request body size and memory per worker at the gateway or container level","reach":"internet","title":"vLLM: crafted image triggers memory exhaustion, denying service to other users","impact":"medium","summary":"vLLM before 0.25.0 lets unauthenticated users submit a decompression-bomb image to the chat completions endpoint, because the Nemotron-VL processors globally disable Pillow's pixel limit. The oversized image exhausts memory, OOM-kills the worker and takes the service down for co-tenants.","category":"ai-tooling","nicknames":[],"who_cares":["vllm","python","self-hosted","ai-inference"],"confidence":0.6,"needs_auth":"no","worst_case":"Worker is OOM-killed, denying service to all co-tenant users","analyst_note":"Confirm the exact affected model processors and whether the NVD entry lists a patched commit or advisory URL beyond the description.","how_to_check":"- Run `pip show vllm` (or `python -c \"import vllm; print(vllm.__version__)\"`) and check whether the version is between `0.10.2` and `0.25.0`\n- Check whether the deployed model is a Nemotron-VL model, since the flaw is in those processors\n- unknown","needs_analyst":true,"not_affected_if":["no-nemotron-vl-models","text-only-serving"],"who_is_affected":"Teams running vLLM between 0.10.2 and 0.25.0 that serve Nemotron-VL vision-language models to users who can send requests to /v1/chat/completions. Deployments that do not accept image input or do not serve Nemotron-VL models are less exposed, but the sources do not confirm whether other processors are affected.","user_interaction":"no","needs_special_setup":"unknown","severity_if_unknown":"unknown"},"cost_usd":0.000662,"input_tokens":3366,"output_tokens":650,"severity_before":"high"},"analyst_status":null,"analyst_note":null,"search":"'-116177':14A,17A '-2026':13A,16A '/v1/chat/completions':118C '0.10.2':79C '0.25.0':31B,81C 'ai':19B,22B 'ai-infer':18B 'ai-tool':21B 'allow':89C 'attack':91C,108C 'bomb':39B,115C 'chat':43B 'co':75B,130C 'co-ten':74B,129C 'complet':44B 'consumpt':86C 'contain':82C 'craft':2A 'cve':12A,15A 'decompress':38B,114C 'decompression-bomb':37B,113C 'deni':7A,126C 'disabl':53B 'endpoint':45B 'exhaust':6A,61B,93C 'global':52B,101C 'host':27B 'imag':3A,40B,60B,104C,116C 'image.max':103C 'infer':20B 'kill':65B,122C 'let':32B 'limit':57B 'memori':5A,62B,94C 'nemotron':49B,98C 'nemotron-vl':48B,97C 'none':107C 'oom':64B,121C 'oom-kil':63B,120C 'overs':59B 'patholog':112C 'pillow':54B 'pixel':56B,105C 'processor':51B,100C 'python':24B 'resourc':85C 'self':26B 'self-host':25B 'servic':8A,71B,127C 'set':102C 'submit':35B,110C 'take':69B 'tenant':76B,131C 'tool':23B 'trigger':4A 'unauthent':33B,90C 'uncontrol':84C 'user':11A,34B 'vl':50B,99C 'vllm':1A,28B,29B,77C 'vulner':87C 'worker':67B,124C","nicknames":[],"mitigations":[],"who_affected":"Teams running vLLM between 0.10.2 and 0.25.0 that serve Nemotron-VL vision-language models to users who can send requests to /v1/chat/completions. Deployments that do not accept image input or do not serve Nemotron-VL models are less exposed, but the sources do not confirm whether other processors are affected.","tier":"B","verified":{},"analyst_revisit_at":null,"first_public_at":"2026-10-25T14:16:57+00:00","kev":null,"cvss_vector":"CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X","cvss_by":"disclosure@vulncheck.com","poc_count":0,"sources":[{"source":"nvd","external_id":"CVE-2026-116177","url":"https://nvd.nist.gov/vuln/detail/CVE-2026-116177","severity":null,"published_at":"2026-10-25T14:16:57+00:00","modified_at":"2026-10-25T14:16:57+00:00","first_seen_at":"2026-10-25T14:41:23+00:00","fetched_at":"2026-10-25T14:41:23+00:00","note":null}],"documents":[],"notice":"Free to read at a human pace. An API with keys and an MCP server are in preparation; early access: hello@infrastack.ai"}