A Large Language Model-Based Detection Method for Poisoning Attacks in Recommender Systems

A Large Language Model-Based Detection Method for Poisoning Attacks in Recommender Systems | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article A Large Language Model-Based Detection Method for Poisoning Attacks in Recommender Systems Feng Liang, Yaojun Hao, Gaojie Yuan, Juxia Li, Liping Feng This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8832268/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 10 You are reading this latest preprint version Abstract Recommender systems are vulnerable to poisoning attacks due to their open nature, and attackers can inject malicious user profiles to deliberately manipulate the recommendation results. Existing detection methods mainly focus on rating behaviors while neglecting key semantic information such as item labels, making them ineffective in handling complex or highly camouflaged attacks. To overcome the limitations that detection methods overly rely on ratings and insufficiently exploit semantic association information in item labels, we use a pretrained large language model to encode label semantic information, and fuse rating information with label semantic information to jointly identify malicious users, thereby proposing a poisoning attack detection method based on large language model–based label semantic encoding, PAD-LLM. First, we adopt a text-to-text Transfer Transformer model to semantically encode the label text sequences, and fuse them with rating behaviors to construct a user-item-label three-dimensional tensor representation, thereby enabling unified modeling of multi-source heterogeneous data. On this basis, we design a local-lobal joint feature extraction framework, via three-dimensional depthwise separable convolution and multi-head Performer to jointly model local interaction patterns and global dependency structures, and via a gated residual mechanism to realize dynamic fusion; furthermore, we incorporate contrastive learning to enhance the inter-class separability of latent representations, thereby improving the identification capability for malicious user profiles. We conduct comparative experiments on the MovieLens-1M and Amazon datasets. The results demonstrate that PAD-LLM achieves better detection performance than the baseline methods under multiple poisoning attack settings. Physical sciences/Engineering Physical sciences/Mathematics and computing recommender systems poisoning attack detection label information three-dimensional convolutional neural networks large language model Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 04 Mar, 2026 Reviews received at journal 03 Mar, 2026 Reviews received at journal 26 Feb, 2026 Reviewers agreed at journal 24 Feb, 2026 Reviewers agreed at journal 23 Feb, 2026 Reviewers invited by journal 23 Feb, 2026 Editor invited by journal 16 Feb, 2026 Editor assigned by journal 11 Feb, 2026 Submission checks completed at journal 11 Feb, 2026 First submitted to journal 09 Feb, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8832268","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":596194660,"identity":"f5ccd78a-11a0-4425-afa8-00a5038cd65d","order_by":0,"name":"Feng Liang","email":"","orcid":"","institution":"Shanxi Agricultural University","correspondingAuthor":false,"prefix":"","firstName":"Feng","middleName":"","lastName":"Liang","suffix":""},{"id":596194661,"identity":"91f110ab-2e57-4fb5-ab11-f4f4f811df46","order_by":1,"name":"Yaojun Hao","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA1UlEQVRIiWNgGAWjYBACAwjFxsDAzHzgwIcfJGlhZ0s8OLOHeC1AwM9jfJiDjQgt5uyHj0n8qOBL3M7M8+EwAw+DPL/YAfxaLHvS0iR7zrAl7mzm3XC4wILBcObsBAIOO5BjJsHbxpa44TBQywwehgSD24S0nH9jJvkXrIXnwWEeNmK03Mgxk4bYwsNArJZnydYyZ9iMNxxmMwAGsgQRfjmffPDmm4pjshvOH3784cMPG3l+aQJagIBFgoHhGIwjQVA5CDB/YGCoIUrlKBgFo2AUjFAAAEf8RwVW4ufjAAAAAElFTkSuQmCC","orcid":"","institution":"Xinzhou Normal University","correspondingAuthor":true,"prefix":"","firstName":"Yaojun","middleName":"","lastName":"Hao","suffix":""},{"id":596194662,"identity":"c7116dcb-dde8-4c97-9464-7bf91ce74f70","order_by":2,"name":"Gaojie Yuan","email":"","orcid":"","institution":"Shanxi Agricultural University","correspondingAuthor":false,"prefix":"","firstName":"Gaojie","middleName":"","lastName":"Yuan","suffix":""},{"id":596194663,"identity":"9a1a4003-c056-4adf-8106-0ac75c7befcd","order_by":3,"name":"Juxia Li","email":"","orcid":"","institution":"Shanxi Agricultural University","correspondingAuthor":false,"prefix":"","firstName":"Juxia","middleName":"","lastName":"Li","suffix":""},{"id":596194664,"identity":"f2ca1e35-3916-4c94-a3b6-f864950b2ce3","order_by":4,"name":"Liping Feng","email":"","orcid":"","institution":"Xinzhou Normal University","correspondingAuthor":false,"prefix":"","firstName":"Liping","middleName":"","lastName":"Feng","suffix":""}],"badges":[],"createdAt":"2026-02-09 15:42:39","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8832268/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8832268/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":103507089,"identity":"26ffcf29-edd6-4bc9-8736-197b41b1ad91","added_by":"auto","created_at":"2026-02-26 13:40:23","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1727896,"visible":true,"origin":"","legend":"","description":"","filename":"ALargeLanguageModelBasedDetectionMethodforPoisoningAttacksinRecommenderSystems.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8832268/v1_covered_253fe2c8-e7a6-4ed3-801b-ffba968fe1bd.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"A Large Language Model-Based Detection Method for Poisoning Attacks in Recommender Systems","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"recommender systems, poisoning attack detection, label information, three-dimensional convolutional neural networks, large language model","lastPublishedDoi":"10.21203/rs.3.rs-8832268/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8832268/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eRecommender systems are vulnerable to poisoning attacks due to their open nature, and attackers can inject malicious user profiles to deliberately manipulate the recommendation results. Existing detection methods mainly focus on rating behaviors while neglecting key semantic information such as item labels, making them ineffective in handling complex or highly camouflaged attacks. To overcome the limitations that detection methods overly rely on ratings and insufficiently exploit semantic association information in item labels, we use a pretrained large language model to encode label semantic information, and fuse rating information with label semantic information to jointly identify malicious users, thereby proposing a poisoning attack detection method based on large language model\u0026ndash;based label semantic encoding, PAD-LLM. First, we adopt a text-to-text Transfer Transformer model to semantically encode the label text sequences, and fuse them with rating behaviors to construct a user-item-label three-dimensional tensor representation, thereby enabling unified modeling of multi-source heterogeneous data. On this basis, we design a local-lobal joint feature extraction framework, via three-dimensional depthwise separable convolution and multi-head Performer to jointly model local interaction patterns and global dependency structures, and via a gated residual mechanism to realize dynamic fusion; furthermore, we incorporate contrastive learning to enhance the inter-class separability of latent representations, thereby improving the identification capability for malicious user profiles. We conduct comparative experiments on the MovieLens-1M and Amazon datasets. The results demonstrate that PAD-LLM achieves better detection performance than the baseline methods under multiple poisoning attack settings.\u003c/p\u003e","manuscriptTitle":"A Large Language Model-Based Detection Method for Poisoning Attacks in Recommender Systems","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-25 06:03:00","doi":"10.21203/rs.3.rs-8832268/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-03-04T08:27:32+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-03-04T02:27:33+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-02-26T07:10:39+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"161787165171362855767391100217218833756","date":"2026-02-24T07:15:03+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"196795740631663847613582097192910253531","date":"2026-02-24T02:49:41+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-02-24T02:46:01+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-02-16T05:31:17+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-02-11T09:52:33+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-02-11T09:49:04+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2026-02-09T15:18:21+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"058aff61-f4a8-4c9f-8e77-c688c9bb91a7","owner":[],"postedDate":"February 25th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":63435731,"name":"Physical sciences/Engineering"},{"id":63435732,"name":"Physical sciences/Mathematics and computing"}],"tags":[],"updatedAt":"2026-04-28T05:54:52+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-25 06:03:00","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8832268","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8832268","identity":"rs-8832268","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}