{"id":"GHSA-xf3w-qp3m-fp62","kind":"vulnerability","aliases":["GHSA-xf3w-qp3m-fp62"],"related":[],"title":"CVE-2026-129266 - Arbitrary Local File Read in DJL - AWS","description":"CVE-2026-129266 - Arbitrary Local File Read in DJL - AWS\n\n### CVE-2026-129266 - Arbitrary local file read via unvalidated resource location in model archive metadata in Deep Java Library\n\nBulletin ID: 2026-183-AWS Scope: AWS/Amazon Content Type: Important (requires attention)/Informational Publication Date: 11/16/2026 10:30 AM PST\n\n Description:\n\nDeep Java Library (DJL) is an open-source, engine-agnostic Java framework for deep learning. We identified CVE-2026-129266, an issue where DJL doesn't validate the resource location specified in model archive metadata. A remote unauthenticated third party could use a crafted model archive to read arbitrary files that the library process can access.\n\nImpacted versions: >=0.7.0, <=0.38.0\n\nResolution:\n\nThis issue has been addressed in DJL version 0.39.0. We recommend upgrading to the latest version and ensuring any forked or derivative code is patched to incorporate the new fixes.\n\nWorkarounds:\n\nOnly load model archives from sources you trust. If you run djl-serving, or a container image that bundles it such as the djl-inference Deep Learning Container, restrict network access to the management and inference APIs so that untrusted callers cannot register models, and set model_url_pattern to an allowlist of model locations you control to restrict models loaded through inference requests.\n\nReferences:\n\n- CVE-2026-129266\n- GHSA-xf3w-qp3m-fp62\n\nAcknowledgement:\n\nWe would like to thank Oguzhan Akkaya for collaborating on this issue through the coordinated vulnerability disclosure process.\n\n Please email aws-security@amazon.com with any s ecurity questions or concerns.","summary":"Deep Java Library (DJL) does not validate the resource location in model archive metadata. A remote unauthenticated attacker who supplies a crafted model archive can read any file the library process can access.","severity":"high","cvss":null,"epss":null,"exploited":false,"exploited_at":null,"products":[],"ecosystems":[],"tags":["aws","deep-learning","djl-serving","java","lab-disclosure","model-serving","package"],"not_affected_if":["apis-not-exposed","trusted-models-only","versions-outside-range"],"how_to_check":"- Check the DJL version in your build file or dependency tree; affected versions are `>=0.7.0`, `<=0.38.0`.\n- If you run djl-serving or the djl-inference Deep Learning Container, check whether the management and inference APIs are reachable from untrusted callers and whether `model_url_pattern` is set.","fix":"- Upgrade DJL to `0.39.0` or later.\n- Ensure any forked or derivative code is patched for the new fixes.","status":"open","published_at":"2026-11-16T00:00:00+00:00","first_seen_at":"2026-11-16T18:52:33+00:00","triaged_at":"2026-11-16T18:53:40+00:00","updated_at":"2026-11-16T18:53:40+00:00","category":"package","confidence":0.7,"triage":{"at":"2026-11-16T18:53:40Z","model":"claude-haiku-5-5","output":{"fix":"- Upgrade DJL to `0.39.0` or later.\n- Ensure any forked or derivative code is patched for the new fixes.","reach":"internet","title":"DJL: crafted model archive reads arbitrary local files on server","impact":"medium","summary":"Deep Java Library (DJL) does not validate the resource location in model archive metadata. A remote unauthenticated attacker who supplies a crafted model archive can read any file the library process can access.","audience":"developers","category":"package","nicknames":[],"who_cares":["java","deep-learning","djl-serving","model-serving"],"confidence":0.7,"fix_status":"available","needs_auth":"no","worst_case":"attacker reads arbitrary files the library process can access","analyst_note":"","how_to_check":"- Check the DJL version in your build file or dependency tree; affected versions are `>=0.7.0`, `<=0.38.0`.\n- If you run djl-serving or the djl-inference Deep Learning Container, check whether the management and inference APIs are reachable from untrusted callers and whether `model_url_pattern` is set.","plain_action":"","product_name":"","needs_analyst":false,"not_affected_if":["versions-outside-range","trusted-models-only","apis-not-exposed"],"who_is_affected":"Teams that load model archives in DJL, including djl-serving and the djl-inference Deep Learning Container. Exposure is greatest where the management or inference APIs accept model registration from untrusted callers. Teams that only load trusted, locally built models are less exposed but should still upgrade.","user_interaction":"no","product_ecosystem":"","needs_special_setup":"unknown","severity_if_unknown":"high"},"cost_usd":0.0008,"input_tokens":5088,"output_tokens":582,"severity_before":"unknown"},"analyst_status":null,"analyst_note":null,"search":"'-129266':3A,71C,81C,140C,282C '-183':101C '-2026':2A,70C,80C,139C,281C '/informational':110C '0.38.0':180C '0.39.0':190C '0.7.0':179C '10':114C '11/16/2026':113C '2026':100C '30':115C 'access':68B,176C,245C 'acknowledg':288C 'address':186C 'agnost':130C 'akkaya':295C 'allowlist':266C 'api':251C 'arbitrari':4A,72C,82C,169C 'archiv':48B,59B,92C,154C,166C,216C 'attack':53B 'attent':109C 'aw':10A,21B,78C,102C 'aws-security@amazon.com':309C 'aws/amazon':104C 'bulletin':98C 'bundl':232C 'caller':255C 'cannot':256C 'code':204C 'collabor':297C 'concern':316C 'contain':229C,242C 'content':105C 'control':271C 'coordin':303C 'could':161C 'craft':57B,164C 'cve':1A,69C,79C,138C,280C 'date':112C 'deep':23B,36B,95C,119C,134C,240C 'deep-learn':22B 'deriv':203C 'descript':118C 'disclosur':31B,305C 'djl':9A,26B,39B,77C,122C,144C,188C,225C,238C 'djl-infer':237C 'djl-serv':25B,224C 'doesn':145C 'ecur':313C 'email':308C 'engin':129C 'engine-agnost':128C 'ensur':199C 'file':6A,63B,74C,84C,170C 'fix':211C 'fork':201C 'fp62':15A,20A,287C 'framework':132C 'ghsa':12A,17A,284C 'ghsa-xf3w-qp3m-fp62':11A,16A,283C 'id':99C 'identifi':137C 'imag':230C 'impact':177C 'import':107C 'incorpor':208C 'infer':239C,250C,277C 'issu':142C,183C,300C 'java':28B,37B,96C,120C,131C 'lab':30B 'lab-disclosur':29B 'latest':196C 'learn':24B,135C,241C 'librari':38B,65B,97C,121C,173C 'like':291C 'load':214C,275C 'local':5A,73C,83C 'locat':45B,89C,150C,269C 'manag':248C 'metadata':49B,93C,155C 'model':33B,47B,58B,91C,153C,165C,215C,258C,261C,268C,274C 'model-serv':32B 'network':244C 'new':210C 'oguzhan':294C 'open':126C 'open-sourc':125C 'packag':35B 'parti':160C 'patch':206C 'pattern':263C 'pleas':307C 'process':66B,174C,306C 'pst':117C 'public':111C 'qp3m':14A,19A,286C 'question':314C 'read':7A,61B,75C,85C,168C 'recommend':192C 'refer':279C 'regist':257C 'remot':51B,157C 'request':278C 'requir':108C 'resolut':181C 'resourc':44B,88C,149C 'restrict':243C,273C 'run':223C 'scope':103C 'serv':27B,34B,226C 'set':260C 'sourc':127C,218C 'specifi':151C 'suppli':55B 'thank':293C 'third':159C 'trust':220C 'type':106C 'unauthent':52B,158C 'untrust':254C 'unvalid':87C 'upgrad':193C 'url':262C 'use':162C 'valid':42B,147C 'version':178C,189C,197C 'via':86C 'vulner':304C 'workaround':212C 'would':290C 'xf3w':13A,18A,285C","nicknames":[],"mitigations":[],"who_affected":"Teams that load model archives in DJL, including djl-serving and the djl-inference Deep Learning Container. Exposure is greatest where the management or inference APIs accept model registration from untrusted callers. Teams that only load trusted, locally built models are less exposed but should still upgrade.","tier":"B","verified":{},"analyst_revisit_at":null,"fix_status":"available","audience":"developers","plain_action":null,"pkeys":[],"ptokens":[],"labels":["aws","deeplearning","djlserving","high","java","labdisclosure","modelserving","package"],"campaign_type":null,"search_a":"'-129266':3A '-2026':2A 'arbitrari':4A 'aw':10A 'cve':1A 'djl':9A 'file':6A 'fp62':15A,20A 'ghsa':12A,17A 'ghsa-xf3w-qp3m-fp62':11A,16A 'local':5A 'qp3m':14A,19A 'read':7A 'xf3w':13A,18A","search_b":"'access':68B 'archiv':48B,59B 'attack':53B 'aw':21B 'craft':57B 'deep':23B,36B 'deep-learn':22B 'disclosur':31B 'djl':26B,39B 'djl-serv':25B 'file':63B 'java':28B,37B 'lab':30B 'lab-disclosur':29B 'learn':24B 'librari':38B,65B 'locat':45B 'metadata':49B 'model':33B,47B,58B 'model-serv':32B 'packag':35B 'process':66B 'read':61B 'remot':51B 'resourc':44B 'serv':27B,34B 'suppli':55B 'unauthent':52B 'valid':42B","search_c":"'-129266':71C,81C,140C,282C '-183':101C '-2026':70C,80C,139C,281C '/informational':110C '0.38.0':180C '0.39.0':190C '0.7.0':179C '10':114C '11/16/2026':113C '2026':100C '30':115C 'access':176C,245C 'acknowledg':288C 'address':186C 'agnost':130C 'akkaya':295C 'allowlist':266C 'api':251C 'arbitrari':72C,82C,169C 'archiv':92C,154C,166C,216C 'attent':109C 'aw':78C,102C 'aws-security@amazon.com':309C 'aws/amazon':104C 'bulletin':98C 'bundl':232C 'caller':255C 'cannot':256C 'code':204C 'collabor':297C 'concern':316C 'contain':229C,242C 'content':105C 'control':271C 'coordin':303C 'could':161C 'craft':164C 'cve':69C,79C,138C,280C 'date':112C 'deep':95C,119C,134C,240C 'deriv':203C 'descript':118C 'disclosur':305C 'djl':77C,122C,144C,188C,225C,238C 'djl-infer':237C 'djl-serv':224C 'doesn':145C 'ecur':313C 'email':308C 'engin':129C 'engine-agnost':128C 'ensur':199C 'file':74C,84C,170C 'fix':211C 'fork':201C 'fp62':287C 'framework':132C 'ghsa':284C 'ghsa-xf3w-qp3m-fp62':283C 'id':99C 'identifi':137C 'imag':230C 'impact':177C 'import':107C 'incorpor':208C 'infer':239C,250C,277C 'issu':142C,183C,300C 'java':96C,120C,131C 'latest':196C 'learn':135C,241C 'librari':97C,121C,173C 'like':291C 'load':214C,275C 'local':73C,83C 'locat':89C,150C,269C 'manag':248C 'metadata':93C,155C 'model':91C,153C,165C,215C,258C,261C,268C,274C 'network':244C 'new':210C 'oguzhan':294C 'open':126C 'open-sourc':125C 'parti':160C 'patch':206C 'pattern':263C 'pleas':307C 'process':174C,306C 'pst':117C 'public':111C 'qp3m':286C 'question':314C 'read':75C,85C,168C 'recommend':192C 'refer':279C 'regist':257C 'remot':157C 'request':278C 'requir':108C 'resolut':181C 'resourc':88C,149C 'restrict':243C,273C 'run':223C 'scope':103C 'serv':226C 'set':260C 'sourc':127C,218C 'specifi':151C 'thank':293C 'third':159C 'trust':220C 'type':106C 'unauthent':158C 'untrust':254C 'unvalid':87C 'upgrad':193C 'url':262C 'use':162C 'valid':147C 'version':178C,189C,197C 'via':86C 'vulner':304C 'workaround':212C 'would':290C 'xf3w':285C","cause_entry":null,"cause_path":null,"disclosures_checked_at":null,"first_public_at":"2026-11-16T00:00:00+00:00","kev":null,"cvss_vector":null,"cvss_by":null,"poc_count":0,"sources":[{"source":"lab","external_id":"https://aws.amazon.com/security/security-bulletins/2026-183-aws/#GHSA-xf3w-qp3m-fp62","url":"https://aws.amazon.com/security/security-bulletins/2026-183-aws/","severity":null,"published_at":"2026-11-16T00:00:00+00:00","modified_at":null,"first_seen_at":"2026-11-16T18:52:33+00:00","fetched_at":"2026-11-16T18:52:33+00:00","note":"AWS: CVE-2026-129266 - Arbitrary Local File Read in DJL - AWS"}],"documents":[{"id":171834,"url":"https://aws.amazon.com/security/security-bulletins/2026-183-aws/","title":"AWS: CVE-2026-129266 - Arbitrary Local File Read in DJL - AWS","kind":"advisory","source":"lab","publisher":"research","text":"CVE-2026-129266 - Arbitrary Local File Read in DJL - AWS\n\n### CVE-2026-129266 - Arbitrary local file read via unvalidated resource location in model archive metadata in Deep Java Library\n\nBulletin ID: 2026-183-AWS Scope: AWS/Amazon Content Type: Important (requires attention)/Informational Publication Date: 11/16/2026 10:30 AM PST\n\n Description:\n\nDeep Java Library (DJL) is an open-source, engine-agnostic Java framework for deep learning. We identified CVE-2026-129266, an issue where DJL doesn't validate the resource location specified in model archive metadata. A remote unauthenticated third party could use a crafted model archive to read arbitrary files that the library process can access.\n\nImpacted versions: >=0.7.0, <=0.38.0\n\nResolution:\n\nThis issue has been addressed in DJL version 0.39.0. We recommend upgrading to the latest version and ensuring any forked or derivative code is patched to incorporate the new fixes.\n\nWorkarounds:\n\nOnly load model archives from sources you trust. If you run djl-serving, or a container image that bundles it such as the djl-inference Deep Learning Container, restrict network access to the management and inference APIs so that untrusted callers cannot register models, and set model_url_pattern to an allowlist of model locations you control to restrict models loaded through inference requests.\n\nReferences:\n\n- CVE-2026-129266\n- GHSA-xf3w-qp3m-fp62\n\nAcknowledgement:\n\nWe would like to thank Oguzhan Akkaya for collaborating on this issue through the coordinated vulnerability disclosure process.\n\n Please email aws-security@amazon.com with any s ecurity questions or concerns.","published_at":"2026-11-16T00:00:00+00:00","relation":"advisory","confidence":null,"note":null}],"disclosures":[],"exploit_sources":[],"notice":"Free to read, with fair use of about 60 pages a minute. For higher limits or API access: sales@infrastack.ai"}