{"692375":{"#nid":"692375","#data":{"type":"event","title":"PhD Proposal by Yu Wei","body":[{"value":"\u003Cp\u003E\u003Cstrong\u003ETitle:\u003C\/strong\u003E\r\nCharacterizing Differential Privacy: Analytical and Black-Box Approaches\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003EDate\u003C\/strong\u003E:\r\nSeptember\r\n11 (Friday)\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003ETime:\u003C\/strong\u003E10am - 12 pm EST\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003ELocation\u003C\/strong\u003E:\r\n(In-person)\r\nCoda C1008 Bolton\u003C\/p\u003E\u003Cp\u003E(Virtual)\r\n\u003Ca href=\u0022https:\/\/teams.microsoft.com\/meet\/263011129320082?p=eWNP2XKBitnUg9s34M\u0022\u003Ehttps:\/\/teams.microsoft.com\/meet\/263011129320082?p=eWNP2XKBitnUg9s34M\u003C\/a\u003E\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003EYu\r\nWei\u003C\/strong\u003E\u003Cbr\u003EPh.D. Student -\u0026nbsp;School of Cybersecurity and Privacy\u003Cbr\u003EGeorgia Institute of Technology\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003ECommittee\r\nMembers\u003C\/strong\u003E\u003Cbr\u003EDr. Vassilis Zikas\r\n(Advisor) - School of Cybersecurity and Privacy, Georgia Institute of Technology\u003Cbr\u003EDr. Vladimir Kolesnikov - School of Cybersecurity and\r\nPrivacy, Georgia Institute of Technology\u003Cbr\u003EDr. Teodora Baluta - School of Cybersecurity and Privacy,\r\nGeorgia Institute of Technology\u003Cbr\u003EDr. Alex Ozdemir - School of Cybersecurity and Privacy,\r\nGeorgia Institute of Technology\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003EAbstract\u003C\/strong\u003E\u003C\/p\u003E\u003Cp\u003EDifferential\r\nprivacy provides a rigorous framework for controlling how much the behavior of\r\na randomized computation can change when an individual\u2019s data changes. Yet\r\nunderstanding and deploying differentially private computations raises several\r\nfundamental questions: How private is a given computation? Can we design\r\ncomputations with better privacy\u2013utility tradeoffs? And can we verify that a\r\nrealized computation actually satisfies its claimed privacy guarantee?\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003EThis\r\ndissertation approaches these questions by interpreting differential privacy\r\nthrough the lens of indistinguishability.\u003C\/strong\u003E\r\nRather than working with a single representation, I characterize this\r\nindistinguishability through different views and reductions that make the\r\nrelated privacy questions tractable. These characterizations lead to two\r\ncomplementary approaches. In the\u0026nbsp;\u003Cstrong\u003Eblack-box approach\u003C\/strong\u003E, I reason from samples of a computation\u2019s observed\r\noutputs, using classification and hypothesis testing to estimate and audit its\r\nindistinguishability. This line of work develops from black-box estimation of\r\n(epsilon, delta)-privacy, to estimation\/auditing of the full f-DP curve, to\r\nsequential and one-run privacy auditing.\u003C\/p\u003E\u003Cp\u003EThe\r\nsecond is an\u0026nbsp;\u003Cstrong\u003Eanalytical approach\u003C\/strong\u003E, which exploits\r\ndistributional structure in the observer\u2019s view to derive tractable\r\ncharacterizations of indistinguishability. I study computations whose\r\nobservable outputs are Gaussian and develop tools for characterizing their\r\ndifferential privacy guarantees, enabling both privacy analysis and mechanism\r\ndesign. Building on this perspective, I also study additive-noise mechanisms,\r\nestablishing the asymptotic optimality of Gaussian noise among additive-noise\r\nmechanisms in high dimensions and designing improved mechanisms in low\r\ndimensions.\u003C\/p\u003E\u003Cp\u003ETogether,\r\nthis dissertation develops a methodology for studying differential privacy as\r\nan indistinguishability notion: characterize indistinguishability through\r\ndifferent views, and use the resulting characterizations to address the three\r\nfundamental problems in differentially private computations.\u0026nbsp;\u0026nbsp;\u003C\/p\u003E\u003Cp\u003E\u0026nbsp;\u003C\/p\u003E","summary":"","format":"limited_html"}],"field_subtitle":"","field_summary":[{"value":"\u003Cp\u003ECharacterizing Differential Privacy: Analytical and Black-Box Approaches\u003C\/p\u003E","format":"limited_html"}],"field_summary_sentence":[{"value":"Characterizing Differential Privacy: Analytical and Black-Box Approaches"}],"uid":"27707","created_gmt":"2026-09-08 14:27:26","changed_gmt":"2026-09-08 14:27:57","author":"Tatianna Richardson","boilerplate_text":"","field_publication":"","field_article_url":"","field_event_time":{"event_time_start":"2026-09-11T10:00:00-04:00","event_time_end":"2026-09-11T12:00:00-04:00","event_time_end_last":"2026-09-11T12:00:00-04:00","gmt_time_start":"2026-09-11 14:00:00","gmt_time_end":"2026-09-11 16:00:00","gmt_time_end_last":"2026-09-11 16:00:00","rrule":null,"timezone":"America\/New_York"},"location":"Coda C1008 Bolton","extras":[],"groups":[{"id":"221981","name":"Graduate Studies"}],"categories":[],"keywords":[{"id":"102851","name":"Phd proposal"}],"core_research_areas":[],"news_room_topics":[],"event_categories":[{"id":"1788","name":"Other\/Miscellaneous"}],"invited_audience":[{"id":"78771","name":"Public"}],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[],"email":[],"slides":[],"orientation":[],"userdata":""}}}