Large language model systems face important security risks from maliciously crafted messages that aim to overwrite the system’s original instructions or leak private data. To study this problem, we organized a capture-the-flag competition at IEEE SaTML 2024, where the flag is a secret string in the LLM system prompt. The competition was organized in two phases. In the first phase, teams developed defenses to prevent the model from leaking the secret. During the second phase, teams were challenged to extract the secrets hidden for defenses proposed by the other teams. This report summarizes the main insights from the competition. Notably, we found that all defenses were bypassed at least once, highlighting the difficulty of designing a successful defense and the necessity for additional research to protect LLM systems. To foster future research in this direction, we compiled a dataset with over 137k multi-turn attack chats and open-sourced the platform.<\/p>\n","protected":false},"excerpt":{"rendered":"
Large language model systems face important security risks from maliciously crafted messages that aim to overwrite the system’s original instructions or leak private data. To study this problem, we organized a capture-the-flag competition at IEEE SaTML 2024, where the flag is a secret string in the LLM system prompt. The competition was organized in two […]<\/p>\n","protected":false},"featured_media":0,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","msr-author-ordering":null,"msr_publishername":"","msr_publisher_other":"","msr_booktitle":"","msr_chapter":"","msr_edition":"","msr_editors":"","msr_how_published":"","msr_isbn":"","msr_issue":"","msr_journal":"","msr_number":"","msr_organization":"","msr_pages_string":"","msr_page_range_start":"","msr_page_range_end":"","msr_series":"","msr_volume":"","msr_copyright":"","msr_conference_name":"NeurIPS 2024","msr_doi":"","msr_arxiv_id":"","msr_s2_paper_id":"","msr_mag_id":"","msr_pubmed_id":"","msr_other_authors":"","msr_other_contributors":"","msr_speaker":"","msr_award":"","msr_affiliation":"","msr_institution":"","msr_host":"","msr_version":"","msr_duration":"","msr_original_fields_of_study":"","msr_release_tracker_id":"","msr_s2_match_type":"","msr_citation_count_updated":"","msr_published_date":"2024-6-12","msr_highlight_text":"","msr_notes":"","msr_longbiography":"","msr_publicationurl":"","msr_external_url":"","msr_secondary_video_url":"","msr_conference_url":"","msr_journal_url":"","msr_s2_pdf_url":"","msr_year":0,"msr_citation_count":0,"msr_influential_citations":0,"msr_reference_count":0,"msr_s2_match_confidence":0,"msr_microsoftintellectualproperty":true,"msr_s2_open_access":false,"msr_s2_author_ids":[],"msr_pub_ids":[],"msr_hide_image_in_river":null,"footnotes":""},"msr-research-highlight":[],"research-area":[13556,13558],"msr-publication-type":[193716],"msr-publisher":[],"msr-focus-area":[],"msr-locale":[268875],"msr-post-option":[],"msr-field-of-study":[246685,257080],"msr-conference":[259048],"msr-journal":[],"msr-impact-theme":[],"msr-pillar":[],"class_list":["post-1047069","msr-research-item","type-msr-research-item","status-publish","hentry","msr-research-area-artificial-intelligence","msr-research-area-security-privacy-cryptography","msr-locale-en_us","msr-field-of-study-machine-learning","msr-field-of-study-security-and-privacy"],"msr_publishername":"","msr_edition":"","msr_affiliation":"","msr_published_date":"2024-6-12","msr_host":"","msr_duration":"","msr_version":"","msr_speaker":"","msr_other_contributors":"","msr_booktitle":"","msr_pages_string":"","msr_chapter":"","msr_isbn":"","msr_journal":"","msr_volume":"","msr_number":"","msr_editors":"","msr_series":"","msr_issue":"","msr_organization":"","msr_how_published":"","msr_notes":"","msr_highlight_text":"","msr_release_tracker_id":"","msr_original_fields_of_study":"","msr_download_urls":"","msr_external_url":"","msr_secondary_video_url":"","msr_longbiography":"","msr_microsoftintellectualproperty":1,"msr_main_download":"","msr_publicationurl":"","msr_doi":"","msr_publication_uploader":[{"type":"url","viewUrl":"false","id":"false","title":"https:\/\/arxiv.org\/abs\/2406.07954","label_id":"252679","label":0}],"msr_related_uploader":[{"type":"url","viewUrl":"false","id":"false","title":"https:\/\/github.com\/ethz-spylab\/satml-llm-ctf","label_id":"264520","label":0}],"msr_citation_count":0,"msr_citation_count_updated":"","msr_s2_paper_id":"","msr_influential_citations":0,"msr_reference_count":0,"msr_arxiv_id":"","msr_s2_author_ids":[],"msr_s2_open_access":false,"msr_s2_pdf_url":null,"msr_attachments":[],"msr-author-ordering":[{"type":"text","value":"Edoardo Debenedetti","user_id":0,"rest_url":false},{"type":"text","value":"Javier Rando","user_id":0,"rest_url":false},{"type":"text","value":"Daniel Paleka","user_id":0,"rest_url":false},{"type":"text","value":"Fineas Silaghi","user_id":0,"rest_url":false},{"type":"text","value":"Dragos Albastroiu","user_id":0,"rest_url":false},{"type":"text","value":"Niv Cohen","user_id":0,"rest_url":false},{"type":"text","value":"Yuval Lemberg","user_id":0,"rest_url":false},{"type":"user_nicename","value":"Reshmi Ghosh","user_id":42471,"rest_url":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Reshmi Ghosh"},{"type":"user_nicename","value":"Ahmed Salem","user_id":43959,"rest_url":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Ahmed Salem"},{"type":"text","value":"Rui Wen","user_id":0,"rest_url":false},{"type":"user_nicename","value":"Giovanni Cherubin","user_id":41410,"rest_url":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Giovanni Cherubin"},{"type":"user_nicename","value":"Santiago Zanella-B\u00e9guelin","user_id":33518,"rest_url":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Santiago Zanella-B\u00e9guelin"},{"type":"text","value":"Robin Schmid","user_id":0,"rest_url":false},{"type":"text","value":"Victor Klemm","user_id":0,"rest_url":false},{"type":"text","value":"Takahiro Miki","user_id":0,"rest_url":false},{"type":"text","value":"Chenhao Li","user_id":0,"rest_url":false},{"type":"text","value":"Stefan Kraft","user_id":0,"rest_url":false},{"type":"text","value":"Mario Fritz","user_id":0,"rest_url":false},{"type":"text","value":"Florian Tramer","user_id":0,"rest_url":false},{"type":"text","value":"Sahar Abdelnabi","user_id":0,"rest_url":false},{"type":"text","value":"Lea Schönherr","user_id":0,"rest_url":false}],"msr_impact_theme":[],"msr_research_lab":[199561],"msr_event":[1088157],"msr_group":[559983,761911],"msr_project":[648207],"publication":[],"video":[],"msr-tool":[],"msr_publication_type":"inproceedings","related_content":{"projects":[{"ID":648207,"post_title":"Confidential AI","post_name":"confidential-ai","post_type":"msr-project","post_date":"2020-05-15 05:46:38","post_modified":"2023-02-15 01:10:13","post_status":"publish","permalink":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/project\/confidential-ai\/","post_excerpt":"Our goal is to make Azure the most trustworthy cloud platform for AI. The platform we envisage offers confidentiality and integrity against privileged attackers including attacks on the code, data and hardware supply chains, performance close to that offered by GPUs, and programmability of state-of-the-art ML frameworks. The confidential AI platform will enable multiple entities to collaborate and train accurate models using sensitive data, and serve these models with assurance that their data and models…","_links":{"self":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/648207"}]}}]},"_links":{"self":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/1047069","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item"}],"about":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-research-item"}],"version-history":[{"count":2,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/1047069\/revisions"}],"predecessor-version":[{"id":1148174,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/1047069\/revisions\/1148174"}],"wp:attachment":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/media?parent=1047069"}],"wp:term":[{"taxonomy":"msr-research-highlight","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-highlight?post=1047069"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=1047069"},{"taxonomy":"msr-publication-type","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-publication-type?post=1047069"},{"taxonomy":"msr-publisher","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-publisher?post=1047069"},{"taxonomy":"msr-focus-area","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-focus-area?post=1047069"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=1047069"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=1047069"},{"taxonomy":"msr-field-of-study","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-field-of-study?post=1047069"},{"taxonomy":"msr-conference","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-conference?post=1047069"},{"taxonomy":"msr-journal","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-journal?post=1047069"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=1047069"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=1047069"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}