[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-513441-105":59,"doc-detail-513441-en":130},{"code":4,"msg":5,"data":6},0,"success",[7,13,18,23,28,33,38,43,48,51,55],{"id":8,"doc_module":4,"doc_module_name":9,"category_name":10,"show_sort_weight":11,"slug":12},1,"Document","Story & Novel",90,"story-novel",{"id":14,"doc_module":4,"doc_module_name":9,"category_name":15,"show_sort_weight":16,"slug":17},2,"Literature",80,"literature",{"id":19,"doc_module":4,"doc_module_name":9,"category_name":20,"show_sort_weight":21,"slug":22},4,"Exam",70,"exam",{"id":24,"doc_module":4,"doc_module_name":9,"category_name":25,"show_sort_weight":26,"slug":27},5,"Comic",60,"comic",{"id":29,"doc_module":4,"doc_module_name":9,"category_name":30,"show_sort_weight":31,"slug":32},6,"Technology",50,"technology",{"id":34,"doc_module":4,"doc_module_name":9,"category_name":35,"show_sort_weight":36,"slug":37},7,"Healthcare",40,"healthcare",{"id":39,"doc_module":4,"doc_module_name":9,"category_name":40,"show_sort_weight":41,"slug":42},8,"Research & Report",30,"research-report",{"id":44,"doc_module":4,"doc_module_name":9,"category_name":45,"show_sort_weight":46,"slug":47},9,"Religion & Spirituality",20,"religion-spirituality",{"id":46,"doc_module":4,"doc_module_name":9,"category_name":49,"show_sort_weight":46,"slug":50},"World Cup","world-cup",{"id":52,"doc_module":4,"doc_module_name":9,"category_name":53,"show_sort_weight":52,"slug":54},10,"Lifestyle","lifestyle",{"id":56,"doc_module":4,"doc_module_name":9,"category_name":57,"show_sort_weight":24,"slug":58},19,"General","general",{"code":4,"msg":60,"data":61},"ok",{"site_id":62,"language":63,"slug":64,"title":65,"keywords":66,"description":67,"schema_data":68,"social_meta":123,"head_meta":125,"extra_data":127,"updated_unix":129},105,"en","arbiter-ai-driven-filtering-for-role-based-access-control","ARBITER - AI-Driven Filtering for Role-Based Access Control","","Role-Based Access Control (RBAC) is difficult to adapt to dynamic enterprise settings when documents used by LLM-driven systems (e.g., RAG) may include information that must not be disclosed to specific user groups. This risk increases as LLMs can leak sensitive data via prompt truncation, classification errors, or missing system context. ARBITER provides RBAC for RAG with layered input/output validation, role-aware retrieval, and post-generation fact-checking, using few-shot LLM prompting for rapid role updates. Evaluations on 389 queries show 85% accuracy and 89% F1-score for query filtering, near traditional RBAC performance.",{"@graph":69,"@context":122},[70,84,105],{"@type":71,"itemListElement":72},"BreadcrumbList",[73,77,79,82],{"item":74,"name":75,"@type":76,"position":8},"https://docshare.wps.com","Home","ListItem",{"item":78,"name":9,"@type":76,"position":14},"https://docshare.wps.com/document/",{"item":80,"name":40,"@type":76,"position":81},"https://docshare.wps.com/document/research-report/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/arbiter-ai-driven-filtering-for-role-based-access-control/513441/",{"url":83,"name":65,"@type":85,"image":86,"author":91,"headline":65,"publisher":94,"fileFormat":97,"inLanguage":63,"description":67,"dateModified":98,"datePublished":99,"encodingFormat":97,"isAccessibleForFree":100,"interactionStatistic":101},"DigitalDocument",{"url":87,"@type":88,"width":89,"height":90},"https://docshare.wps.com/thumbnails/arbiter-ai-driven-filtering-for-role-based-access-control/513441.png","ImageObject",300,407,{"name":92,"@type":93},"Riley West","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-11-15","2026-10-06",true,{"@type":102,"interactionType":103,"userInteractionCount":52},"InteractionCounter",{"@type":104},"ViewAction",{"@type":106,"mainEntity":107},"FAQPage",[108,114,118],{"name":109,"@type":110,"acceptedAnswer":111},"What problem does ARBITER address in RBAC for RAG systems?","Question",{"text":112,"@type":113},"ARBITER targets unauthorized disclosure of sensitive information during retrieval and generation in RAG, where access boundaries can be misapplied due to classification errors, prompt truncation, or lost system context.","Answer",{"name":115,"@type":110,"acceptedAnswer":116},"How does ARBITER enforce access control without relying on fine-tuned classifiers?",{"text":117,"@type":113},"ARBITER uses few-shot LLM prompting with prompt-based steering and implements layered input/output validation, role-aware retrieval, and post-generation fact-checking to enforce RBAC constraints.",{"name":119,"@type":110,"acceptedAnswer":120},"What were the experimental results of ARBITER on query filtering?",{"text":121,"@type":113},"On a synthetic dataset with 389 queries, ARBITER achieved 85% accuracy and 89% F1-score for query filtering, close to static RBAC on the same dataset.","https://schema.org",{"og:url":83,"og:type":124,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":126,"canonical":83},"index,follow",{"doc_id":128,"site_id":62},513441,1791361410,{"code":4,"msg":5,"data":131},{"doc_id":128,"user_id":132,"nickname":92,"user_avatar":133,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":134,"file_id":135,"file_url":136,"file_type":137,"file_size":138,"view_count":52,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":139,"language":140,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":141,"faqs":142,"seo_title":143,"seo_description":67,"update_tm":144,"read_time":145,"preview_page_count":19,"preview_urls":146},1099523885074,"https://ap-avatar.wpscdn.com/davatar_9964176cb1d06d4a9deccf72a44ae3dc","ARBITER: AI-Driven Filtering for Role-Based Access Control  \nMichele Lorenzo Reply Spike Turin, Italy  \n[mi.lorenzo@reply.it](mi.lorenzo@reply.it)  \nIdilio Drago University of Turin Turin, Italy  \nidilio.drago@unito. it  \nDario Salvadori Reply Spike Milan, Italy  \n[d.salvadori@reply.it](d.salvadori@reply.it)  \nFabio Romolo Vayr Reply Spike Turin, Italy [f.vayr@reply.it](f.vayr@reply.it)  \narXiv :2512 .20535v1 [ cs .CR] 23 Dec 2025  \nAbstract  \nRole-Based Access Control (RBAC) struggles to adapt to dynamic enterprise environments with documents that contain information that cannot be disclosed to specific user groups. As these documents are used by LLM-driven systems (e.g., in RAG) the problem is exacerbated as LLMs can leak sensitive data due to prompt truncation, classification errors, or loss of system context. We introduce ARBITER, a system designed to provide RBAC in RAG systems. ARBITER implements layered input/output validation, role-aware retrieval, and post-generation fact-checking. Unlike traditional RBAC approaches that rely on fine-tuned classifiers, ARBITER uses LLMs operating in few-shot settings with prompt-based steering for rapid deployment and role updates. We evaluate the approach on 389 queries using a synthetic dataset. Experimental results show 85% accuracy and 89% F1-score in query filtering, close to traditional RBAC solutions. Results suggest that practical RBAC deployment on RAG systems is approaching the maturity level needed for dynamic enterprise environments.  \n1 Introduction  \nEnterprises are rapidly embracing generative AI technologies, deploying them across local and cloud infrastructures to enhance productivity, streamline workflows, and support decision-making [7] . Among these, Retrieval-Augmented Generation (RAG) has emerged as a powerful paradigm, allowing Large Language Models (LLMs) to access external knowledge sources and produce more accurate, contextually grounded outputs [4] . However, the integration of such capabilities in enterprise environments introduces security and access control challenges [2] .  \nA key problem is the risk of unauthorized access to sensitive data during the retrieval process. Generative systems may expose or create information beyond a user’s clearance level due to document classification errors in the RAG pipeline, prompt truncation, etc. Such errors may cause the model to misinterpret access boundaries, disregard system instructions, or generate responses based on internal knowledge rather than retrieved context, ultimately leading to unauthorized disclosures.  \nTraditionally, enterprises rely on Role-Based Access Control (RBAC), where users are assigned predefined roles that dictate access to specific documents or data. RBAC operates through static filters, creating “visibility cones” to limit data exposure. This method relies heavily on exact and static classifications, which are costly to update and imperfect regardless. Fine-tuning models for classification adds further burden, requiring labeled data and frequent retraining to accommodate evolving access policies.  \nAnother critical aspect is the need to use local models, as enterprises are unwilling to expose private documents to external parties during the retrieval process. This architectural choice limits access to the capabilities of larger hosted models.  \nWe evaluate a local, adaptive filtering approach that supplements RBAC with AI-driven safeguards embedded throughout the RAG pipeline. For that we introduce ARBITER (AI Role Based Intelligent Text EvaluatoR)– a system that leverages NeMo Guardrails [8] to create a multi-layered filter chain without relying on fine-tuned models.  \nOur main contributions are: (i) we propose a few-shot approach for RBAC using foundational LLMs that eliminates the need for document classifiers and enables rapid deployment; (ii) we test ARBITER to augment RBAC systems through layered input/output validation and role-aware retrieval. Using a synthetic dataset of roles ","cbCainKj3QFKPypo","https://ap.wps.com/l/cbCainKj3QFKPypo","pdf",407481,11,"English","# 1 Introduction\n# 2 Background and Related Work\n## 2.1 RBAC Systems\n## 2.2 Related Work","[{\"question\":\"What problem does ARBITER address in RBAC for RAG systems?\",\"answer\":\"ARBITER targets unauthorized disclosure of sensitive information during retrieval and generation in RAG, where access boundaries can be misapplied due to classification errors, prompt truncation, or lost system context.\"},{\"question\":\"How does ARBITER enforce access control without relying on fine-tuned classifiers?\",\"answer\":\"ARBITER uses few-shot LLM prompting with prompt-based steering and implements layered input/output validation, role-aware retrieval, and post-generation fact-checking to enforce RBAC constraints.\"},{\"question\":\"What were the experimental results of ARBITER on query filtering?\",\"answer\":\"On a synthetic dataset with 389 queries, ARBITER achieved 85% accuracy and 89% F1-score for query filtering, close to static RBAC on the same dataset.\"}]","ARBITER - AI-Driven Filtering for Role-Based Access Control | PDF",1791285159,28,[147,151,155,159],{"page_no":8,"url":148,"width":149,"height":150},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-17/bac358853f6050345b360eb0d41349a9.webp",983,1272,{"page_no":14,"url":152,"width":153,"height":154},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-17/1804655a2674e969928738f85b824aef.webp",982,1270,{"page_no":81,"url":156,"width":157,"height":158},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-17/890be3db6ac64ca92cdda885d798020a.webp",981,1269,{"page_no":19,"url":160,"width":161,"height":162},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-17/9e4e93968658c3a6b8de26161dca1830.webp",980,1268]