[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-119812-en":3,"doc-seo-119812-105":29,"detail-sidebar-cat-0-en-105":90},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":11,"language":21,"language_code":22,"site_id":23,"html_lang":22,"table_of_contents":24,"faqs":25,"seo_title":26,"seo_description":14,"update_tm":27,"read_time":28},119812,34359740700684,"Finn","https://ap-avatar.wpscdn.com/avatar/1f400023980c374ae676?_k=1777273430885731487",8,"Research & Report","Lifecycle Management and Security of On-device Machine Learning Models - Defensive Publications Series","Endpoint management solutions cannot be used to manage on-device machine learning models. This disclosure presents techniques that combine cloud ML model management with endpoint management to support lifecycle management and security for on-device ML deployments. The approach uses a model catalog to track device deployments, notify administrators about model upgrades, and allow upgrades and deletion independent of app deployment. With user permission, endpoint-based observability helps detect misuse and supports compliance via deployment views including versions.","Technical Disclosure Commons  \nDefensive Publications Series  \nJuly 2023  \nLifecycle Management and Security of On-device Machine Learning Models  \nHari Bhaskar S  \nFollow this and additional works at: [https://www.tdcommons.org/dpubs_series](https://www.tdcommons.org/dpubs_series)  \nRecommended Citation  \nBhaskar S, Hari, \"Lifecycle Management and Security of On-device Machine Learning Models\", Technical Disclosure Commons,(July 11, 2023)  \n[https://www.tdcommons.org/dpubs_series/6039](https://www.tdcommons.org/dpubs_series/6039)  \nThis work is licensed under a Creative Commons Attribution 4.0 License.  \nThis Article is brought to you for free and open access by Technical Disclosure Commons. It has been accepted for inclusion in Defensive Publications Series by an authorized administrator of Technical Disclosure Commons.  \nLifecycle Management and Security of On-device Machine Learning Models  \nABSTRACT  \nEndpoint management solutions cannot be utilized to manage on-device machine learning models. This disclosure describes techniques to integrate the best-in-class capabilities of cloud ML model management and endpoint management to enable lifecycle management and security of on-device ML model deployments. Endpoint management solutions as described herein include the capability to manage on-device models, e.g., to perform tasks such as model tracking, upgrade, and wipe out compliance. The described techniques, which can be implemented as part ofan endpoint management solution, use a model catalog to track device deployments for a given machine learning model. When a model upgrade is available, a  \nnotification is provided to administrators, app developers, etc. On-device models can be  \nupgraded, deleted, and tracked independent of app deployment. Additionally, with user  \npermission, observability of on-device models is enabled through endpoint management to  \ndetect misuse. The endpoint management solution and model catalog also provide a deployment  \nview of on-device models, including model versions and can be used to ensure compliance.  \nKEYWORDS  \n● Model lifecycle  \n● Machine learning ops  \n● MLOps  \n● Model security  \n● On-device machine learning  \n● Endpoint management  \n● Model management  \nPublished by Technical Disclosure Commons, 2023 2  \nBACKGROUND  \nManagement of client apps through endpoint management solutions is common.  \nHowever, such solutions only cover an app itself and not app components such as on-device  \nML models distributed with the app. While server-side models are managed, edge or mobile  \ndevices currently lack capabilities such as model versioning, tracking model deployments, etc.  \nThere is no solution that combines endpoint management and model lifecycle management  \ncapabilities to provide holistic management capabilities such as model refresh, download on  \ndemand, tracking etc. Also, current solutions do not support model tracking at deployment level and lack the ability to address security exploits on device level by wiping out models only as  \nnecessary rather than the entire app.  \nThe advent ofon-device machine learning (ML) has enabled enterprises to deploy machine learning models on mobile devices for use cases such as object/image recognition and other ML tasks that leverage device capabilities such as camera, storage, and processing. Ondevice ML models are often built as student models from a larger server-side teacher model and deployed as distilled miniature lightweight models. Such ML models are usually packaged with mobile applications and are sent as part of the application release to the mobile device. Lack of visibility into deployment of ML models on mobile devices that belong to the enterprise, or are used by enterprise employees or contractors pose a number of challenges for enterprises  \n● Endpoint management solutions do not have specific capability to track ML models deployed on a device, e.g., removal of a model, monitoring the use of a model, etc. since ML models are","cbCaivJlcwdPkSZM","https://ap.wps.com/l/cbCaivJlcwdPkSZM","pdf",177393,1,"English","en",105,"# Background\n## Limitations of current endpoint management\n## Need for holistic on-device ML model lifecycle control\n# Description\n## Integrated lifecycle management approach\n## Model catalog and deployment tracking\n## Upgrade, deletion, and compliance\n## Observability and misuse detection","[{\"question\":\"Why can’t endpoint management solutions manage on-device ML models directly?\",\"answer\":\"They mainly cover client apps rather than the ML model components distributed with apps, so they lack capabilities like model versioning and deployment tracking for edge or mobile devices.\"},{\"question\":\"How does the proposed approach manage the lifecycle of on-device ML model deployments?\",\"answer\":\"It integrates cloud ML model management with endpoint management and uses a model catalog to track device deployments, handle model upgrades, and enable deletion and tracking independent of app deployment.\"},{\"question\":\"How is security and misuse detection achieved for on-device models?\",\"answer\":\"With user permission, endpoint management enables observability of on-device models to detect misuse, and the deployment view supports compliance using model versions.\"}]","Lifecycle Management and Security of On-device Machine Learning Models - Defensive Publications Series | PDF",1785726429,20,{"code":4,"msg":30,"data":31},"ok",{"site_id":23,"language":22,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":85,"head_meta":87,"extra_data":89,"updated_unix":27},"lifecycle-management-and-security-of-on-device-machine-learning-models-defensive-publications-series","",{"@graph":35,"@context":84},[36,53,67],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/lifecycle-management-and-security-of-on-device-machine-learning-models-defensive-publications-series/119812/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":22,"description":14,"dateModified":61,"datePublished":61,"encodingFormat":60,"isAccessibleForFree":62,"interactionStatistic":63},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-08-03",true,{"@type":64,"interactionType":65,"userInteractionCount":20},"InteractionCounter",{"@type":66},"ViewAction",{"@type":68,"mainEntity":69},"FAQPage",[70,76,80],{"name":71,"@type":72,"acceptedAnswer":73},"Why can’t endpoint management solutions manage on-device ML models directly?","Question",{"text":74,"@type":75},"They mainly cover client apps rather than the ML model components distributed with apps, so they lack capabilities like model versioning and deployment tracking for edge or mobile devices.","Answer",{"name":77,"@type":72,"acceptedAnswer":78},"How does the proposed approach manage the lifecycle of on-device ML model deployments?",{"text":79,"@type":75},"It integrates cloud ML model management with endpoint management and uses a model catalog to track device deployments, handle model upgrades, and enable deletion and tracking independent of app deployment.",{"name":81,"@type":72,"acceptedAnswer":82},"How is security and misuse detection achieved for on-device models?",{"text":83,"@type":75},"With user permission, endpoint management enables observability of on-device models to detect misuse, and the deployment view supports compliance using model versions.","https://schema.org",{"og:url":51,"og:type":86,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":88,"canonical":51},"index,follow",{"doc_id":7,"site_id":23},{"code":4,"msg":5,"data":91},[92,96,100,104,109,114,119,122,126,129,133],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":93,"show_sort_weight":94,"slug":95},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":97,"show_sort_weight":98,"slug":99},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":101,"show_sort_weight":102,"slug":103},"Exam",70,"exam",{"id":105,"doc_module":4,"doc_module_name":45,"category_name":106,"show_sort_weight":107,"slug":108},5,"Comic",60,"comic",{"id":110,"doc_module":4,"doc_module_name":45,"category_name":111,"show_sort_weight":112,"slug":113},6,"Technology",50,"technology",{"id":115,"doc_module":4,"doc_module_name":45,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":45,"category_name":124,"show_sort_weight":28,"slug":125},9,"Religion & Spirituality","religion-spirituality",{"id":28,"doc_module":4,"doc_module_name":45,"category_name":127,"show_sort_weight":28,"slug":128},"World Cup","world-cup",{"id":130,"doc_module":4,"doc_module_name":45,"category_name":131,"show_sort_weight":130,"slug":132},10,"Lifestyle","lifestyle",{"id":134,"doc_module":4,"doc_module_name":45,"category_name":135,"show_sort_weight":105,"slug":136},19,"General","general"]