{"676198":{"#nid":"676198","#data":{"type":"event","title":"PhD Defense by Jaewon Lee","body":[{"value":"\u003Cp\u003ETitle: HW-SW Co-design method to mitigate GPU Memory Safety Vulnerability\u003C\/p\u003E\u003Cp\u003EDate: Tuesday, September 3, 2024\u003Cbr\u003ETime: 1:00 PM - 3:00 PM EDT\u003Cbr\u003ELocation:\u003Cbr\u003E\u0026nbsp; \u0026nbsp; \u0026nbsp;In-Person: KACB 2100\u003Cbr\u003E\u0026nbsp; \u0026nbsp; \u0026nbsp;Online: Click here to join the meeting\u003C\/p\u003E\u003Cp\u003EJaewon Lee\u003Cbr\u003EPh.D. Candidate\u003Cbr\u003ESchool of Computer Science\u003Cbr\u003ECollege of Computing\u003Cbr\u003EGeorgia Institute of Technology\u003Cbr\u003E\u003Cbr\u003ECommittee:\u003Cbr\u003EDr. Hyesoon Kim (advisor), School of Computer Science, Georgia Institute of Technology\u003Cbr\u003EDr. Moinuddin Qureshi, School of Computer Science, Georgia Institute of Technology\u003Cbr\u003EDr. Tushar Krishna, School of Electrical and Computer Engineering \u0026amp; School of Computer Science, Georgia Institute of Technology\u003Cbr\u003EDr. Saibal Mukhopadhyay, School of Electrical and Computer Engineering, Georgia Institute of Technology\u003Cbr\u003EDr. Jaekyu Lee, Arm\u003Cbr\u003E\u003Cbr\u003EAbstract\u003Cbr\u003EGraphics Processing Units (GPUs) were once considered acceptable even if they were insecure; however, the surge in Artificial Intelligence (AI) applications has now brought GPUs into critical roles in life and financial decision-making. Recent studies have successfully demonstrated that attackers can induce failures in AI models by exploiting vulnerabilities in GPU memory.\u003Cbr\u003EIn this dissertation, we first propose GPUShield, an efficient hardware\/software co-designed GPU memory bounds-checking mechanism that fully leverages the characteristics of GPU programs. GPUShield introduces efficient hardware bounds-checking logic by utilizing pointer tagging methods. It minimizes metadata access by taking advantage of the GPU\u0027s region-based memory access, metadata caching, and memory coalescing.\u003Cbr\u003ESecondly, we propose a more lightweight and practical mechanism, LMI. LMI is a fine-grained memory safety mechanism that leverages these concepts. Even when thousands of threads simultaneously allocate and access memory buffers in both the stack and heap, LMI ensures negligible impact on performance and hardware costs,. This is achieved by implementing power-of-two-sized and aligned pointers and performing static analysis to identify pointer operations. This approach also enables storing metadata inside the unused upper bits of pointers, which are shrinking due to the expansion of the virtual memory address space. The unique characteristics of GPU programs make this approach feasible, unlike in CPUs, where the inherent complexity of programs poses challenges. Our evaluation demonstrates that LMI incurs extremely low hardware costs and near-zero performance overhead, making it a practical and efficient solution for enhancing GPU memory safety.\u003C\/p\u003E\u003Cp\u003E\u0026nbsp;\u003C\/p\u003E","summary":"","format":"limited_html"}],"field_subtitle":"","field_summary":[{"value":"\u003Cp\u003EHW-SW Co-design method to mitigate GPU Memory Safety Vulnerability\u003C\/p\u003E","format":"limited_html"}],"field_summary_sentence":[{"value":"HW-SW Co-design method to mitigate GPU Memory Safety Vulnerability"}],"uid":"27707","created_gmt":"2024-08-23 18:47:00","changed_gmt":"2024-08-23 18:47:00","author":"Tatianna Richardson","boilerplate_text":"","field_publication":"","field_article_url":"","field_event_time":{"event_time_start":"2024-09-03T13:00:00-04:00","event_time_end":"2024-09-03T15:00:05-04:00","event_time_end_last":"2024-09-03T15:00:05-04:00","gmt_time_start":"2024-09-03 17:00:00","gmt_time_end":"2024-09-03 19:00:05","gmt_time_end_last":"2024-09-03 19:00:05","rrule":null,"timezone":"America\/New_York"},"location":"KACB 2100","extras":[],"groups":[{"id":"221981","name":"Graduate Studies"}],"categories":[],"keywords":[{"id":"100811","name":"Phd Defense"}],"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":""}}}