{"id":4431,"date":"2024-03-01T04:39:00","date_gmt":"2024-02-29T19:39:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=4431"},"modified":"2024-03-01T04:39:00","modified_gmt":"2024-02-29T19:39:00","slug":"do-membership-inference-attacks-work-on-large-language-models","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=4431","title":{"rendered":"Do Membership Inference Attacks Work on Large Language Models?"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>Do Membership Inference Attacks Work on Large Language Models?\u00a0<\/strong>[145.9]<br>\u30e1\u30f3\u30d0\u30fc\u30b7\u30c3\u30d7\u63a8\u8ad6\u653b\u6483(MIA)\u306f\u3001\u7279\u5b9a\u306e\u30c7\u30fc\u30bf\u30dd\u30a4\u30f3\u30c8\u304c\u30bf\u30fc\u30b2\u30c3\u30c8\u30e2\u30c7\u30eb\u306e\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u30c7\u30fc\u30bf\u306e\u30e1\u30f3\u30d0\u30fc\u3067\u3042\u308b\u304b\u3069\u3046\u304b\u3092\u4e88\u6e2c\u3057\u3088\u3046\u3068\u3059\u308b\u3002 \u6211\u3005\u306f\u3001Pile\u3067\u8a13\u7df4\u3055\u308c\u305f\u8a00\u8a9e\u30e2\u30c7\u30eb\u306b\u5bfe\u3057\u3066\u3001MIA\u306e\u5927\u898f\u6a21\u8a55\u4fa1\u3092\u884c\u3044\u3001\u305d\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u306f160M\u304b\u308912B\u307e\u3067\u3067\u3042\u308b\u3002 \u69d8\u3005\u306a LLM \u30b5\u30a4\u30ba\u3084\u9818\u57df\u306b\u307e\u305f\u304c\u308b\u307b\u3068\u3093\u3069\u306e\u8a2d\u5b9a\u306b\u304a\u3044\u3066,MIA \u306f\u30e9\u30f3\u30c0\u30e0\u306a\u63a8\u6e2c\u3088\u308a\u3082\u307b\u3068\u3093\u3069\u512a\u308c\u3066\u3044\u306a\u3044\u3053\u3068\u304c\u308f\u304b\u3063\u305f\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2402.07841v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2402.07841v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Mon, 12 Feb 2024 17:52:05 GMT)<\/li>\n\n\n\n<li>LLM\u3078\u306e\u30e1\u30f3\u30d0\u30b7\u30c3\u30d7\u63a8\u8ad6\u653b\u6483\u306f\u6709\u52b9\u3067\u306f\u306a\u3044\u306e\u3067\u306f\uff1f\u3068\u3044\u3046\u5831\u544a\u3002\u300cWe identify specific settings where LLMs have been shown to be vulnerable to membership inference and show that the apparent success in such settings can be attributed to a distribution shift, such as when members and non-members are drawn from the seemingly identical domain but with different temporal ranges.\u300d\u3068\u624b\u53b3\u3057\u3044\u3002\u7d50\u8ad6\u306b\u3082\u66f8\u3044\u3066\u3042\u3063\u305f\u304c\u3001\u3053\u3046\u3044\u3063\u305f\u7279\u6027\u3092\u7406\u89e3\u305b\u305a\u306b\u4f55\u304b\u306b\u5fdc\u7528\u3059\u308b\u306e\u306f\u5371\u967a\u3067\u3042\u308b\u3068\u601d\u3046\u3002<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/iamgroot42\/mimir\">iamgroot42\/mimir: Python package for measuring memorization in LLMs (github.com)<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[32,223,242],"class_list":["post-4431","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-attack","tag-llm","tag-membership"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/4431","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=4431"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/4431\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=4431"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=4431"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=4431"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}