[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-gradient-theft-attacks-on-federated-learning-mostly-dont-work":10,"sections":35},{"siteName":4,"siteTagline":5,"publisherName":4,"contactEmail":6},"The Revision","Tech news, decoded.","editor@therevision.news",{"gaMeasurementId":8,"adsenseClientId":9},"G-ZW2MV82GYR","ca-pub-8533917693782264",{"article":11},{"id":12,"slug":13,"title":14,"dek":15,"body_md":16,"tags_json":17,"published_at":18,"created_at":19,"updated_at":20,"status":21,"review_note":22,"review_notes":23,"image_url":22,"persona_id":22,"persona_name":22,"section":24,"tags":25,"sources":30,"feedback":34,"feedback_at":22,"cost_usd":34,"total_tokens":34},10695,"gradient-theft-attacks-on-federated-learning-mostly-dont-work","Gradient Theft Attacks on Federated Learning Mostly Don't Work","A new study finds that gradient inversion attacks, long cited as a federated learning privacy risk, fail against modern production-grade models.","Turns out stealing someone's photos from a shared AI model is harder than the scarier papers suggest.\n\nResearchers tested gradient inversion - a technique that supposedly reconstructs private training images in federated learning by examining shared gradient updates rather than the raw data - across multiple datasets and tasks, including image classification and object detection. They ran the attacks against modern, performance-optimized vision architectures at realistic image resolutions, not the stripped-down setups used in most earlier demonstrations. The result: those modern models consistently resisted meaningful image reconstruction. Older or transitional architectures stayed vulnerable, but only under narrow assumptions that do not match how systems are actually trained and deployed.\n\nA lot of the alarm around gradient inversion comes from papers that test attacks in inference mode or on simplified architectures, conditions that flatter the attacker and rarely show up in production. This study argues that, under the common honest-but-curious server assumption, the realistic privacy risk from gradient inversion in federated learning is much smaller than the literature implies, at least for the systems companies actually ship.\n\nIt is a useful reminder that an attack working in a lab and an attack working on real infrastructure are two different claims, and privacy threat models should be built on the second one.","[\"federated-learning\",\"privacy\",\"ai-security\",\"machine-learning\"]","2026-10-07T04:00:00.000Z","2026-10-09T04:42:02.790Z","2026-10-09T04:42:07.011Z","published",null,[],"security",[26,27,28,29],"federated-learning","privacy","ai-security","machine-learning",[31],{"name":32,"url":33},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.19819",0,{"sections":36},[37,42,46,51,56,61,65,70,75,79,84,89,94,99],{"name":38,"slug":39,"count":40,"latest_published_at":41},"AI","ai",6517,"2026-10-08T04:00:00.000Z",{"name":43,"slug":24,"count":44,"latest_published_at":45},"Security",911,"2026-10-07T19:53:42.000Z",{"name":47,"slug":48,"count":49,"latest_published_at":50},"Policy","policy",474,"2026-10-07T18:23:21.000Z",{"name":52,"slug":53,"count":54,"latest_published_at":55},"Deals","deals",458,"2026-10-08T10:03:39.000Z",{"name":57,"slug":58,"count":59,"latest_published_at":60},"Hardware","hardware",223,"2026-10-08T08:30:04.000Z",{"name":62,"slug":63,"count":64,"latest_published_at":18},"Science","science",187,{"name":66,"slug":67,"count":68,"latest_published_at":69},"Consumer Tech","consumer-tech",175,"2026-10-08T09:16:05.000Z",{"name":71,"slug":72,"count":73,"latest_published_at":74},"Software","software",113,"2026-10-07T18:10:00.000Z",{"name":76,"slug":77,"count":73,"latest_published_at":78},"Startups","startups","2026-10-07T23:36:57.000Z",{"name":80,"slug":81,"count":82,"latest_published_at":83},"Dev Tools","dev-tools",105,"2026-10-07T16:59:11.000Z",{"name":85,"slug":86,"count":87,"latest_published_at":88},"General","general",61,"2026-10-07T22:00:24.000Z",{"name":90,"slug":91,"count":92,"latest_published_at":93},"Gaming","gaming",56,"2026-10-07T12:00:00.000Z",{"name":95,"slug":96,"count":97,"latest_published_at":98},"Reviews","reviews",33,"2026-10-05T11:57:17.000Z",{"name":100,"slug":101,"count":102,"latest_published_at":103},"How-To","how-to",8,"2026-10-05T09:00:00.000Z"]