BEGIN:VCALENDAR VERSION:2.0 X-WR-CALNAME:EventsCalendar PRODID:-//hacksw/handcal//NONSGML v1.0//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/New_York LAST-MODIFIED:20240422T053451Z TZURL:https://www.tzurl.org/zoneinfo-outlook/America/New_York X-LIC-LOCATION:America/New_York BEGIN:DAYLIGHT TZNAME:EDT TZOFFSETFROM:-0500 TZOFFSETTO:-0400 DTSTART:19700308T020000 RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU END:DAYLIGHT BEGIN:STANDARD TZNAME:EST TZOFFSETFROM:-0400 TZOFFSETTO:-0500 DTSTART:19701101T020000 RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU END:STANDARD END:VTIMEZONE BEGIN:VEVENT CATEGORIES:College of Engineering,Thesis/Dissertations DESCRIPTION:Abstract:      Wireless sensing enables device-free and non- intrusive perception by using changes in wireless signals caused by human activities and the surrounding environment. Recent advances in deep learni ng and large language models (LLMs) have expanded wireless sensing capabil ities for applications such as localization, human activity understanding, and pose estimation. However, existing learning models face challenges fr om privacy leakage, corrupted observations, modality mismatch, and adversa rial manipulation. These challenges can limit their performance and reliab ility in practical sensing environments. This dissertation develops novel deep learning frameworks for trustworthy wireless sensing. It focuses on p rivacy-preserving learning, corruption-resilient representations, language -driven wireless sensing, and adversarial security. The first component de velops a hybrid DNN and genetic algorithm framework for fine-grained locat ion privacy in 6G Integrated Sensing and Communication (ISAC) systems. The framework generates high-dimensional, physically constrained subcarrier o bfuscation vectors that control location detection resolution by making th e true channel indistinguishable from channels associated with selected du mmy locations. While protecting sensitive information is essential, trustw orthy wireless sensing also requires learning models that remain effective when the underlying channel observations are unreliable. Therefore, the s econd component introduces a two-stage corruption-resilient learning archi tecture for human pose and shape estimation under corrupted Channel State Information (CSI). The framework first mitigates unreliable observations a t the channel level and then addresses residual corruption at the represen tation level through a dual-stream architecture that uses reliable represe ntations as reference targets to progressively constrain corrupted represe ntations. This approach improves the reliability of learning-based sensing under corrupted observations, but conventional wireless sensing models st ill rely heavily on predefined activity labels and temporal segmentation, which limits their ability to understand continuous and unseen human activ ities. To overcome this limitation, this research further develops Wireles sSenseLLM, a language-driven framework for wireless human sensing. A speci alized wireless encoder, CSI-to-Language Adapter, and cross-domain project ion mechanism bridge continuous CSI and human language, enabling segmentat ion-free and zero-shot understanding of sequential and overlapping human m otions. Building on these learning frameworks, the proposed research will investigate backdoor vulnerabilities and robust representation learning in wireless sensing models. This work will study how adversarial triggers al ter learned representations and induce attacker controlled model behavior, and will develop mechanisms to detect and mitigate such manipulation whil e preserving benign representations and downstream sensing performance. In summary, this dissertation advances both deep learning methodologies and wireless sensing systems. It develops new learning approaches for privacy protection, corruption-resilient representations, cross-modal signal-langu age alignment, and adversarial robustness. Together, these contributions p rovide a foundation for developing trustworthy deep learning models for wi reless sensing applications. ADVISOR(s):  Dr. Long Jiao, Department of Co mputer and Information Science (ljiao@umassd.edu)                               COMMITTEE MEMBERS: Dr. Jiawei Yuan, Department of Computer and Information Science Dr. Yuchou Chang, Department of Computer and Information Science Dr. Roulin Zhou, Department of Electrical & Comput er Engineering NOTE: All EAS Students are ENCOURAGED to attend.\nEvent pa ge: https://www.umassd.edu/events/cms/9-9-26-eas-doctoral-proposal-defense --by-mahmuda-akter-keya.php\nEvent link: https://umassd.zoom.us/j/40619990 32?pwd=bUw0WGpDbTQ4UzJneFd5TTBFeUw1dz09 X-ALT-DESC;FMTTYPE=text/html:

Abstract:     

\n

Wirel ess sensing enables device-free and non-intrusive perception by using chan ges in wireless signals caused by human activities and the surrounding env ironment. Recent advances in deep learning and large language models (LLMs ) have expanded wireless sensing capabilities for applications such as loc alization\, human activity understanding\, and pose estimation. However\, existing learning models face challenges from privacy leakage\, corrupted observations\, modality mismatch\, and adversarial manipulation. These cha llenges can limit their performance and reliability in practical sensing e nvironments. This dissertation develops novel deep learning frameworks for trustworthy wireless sensing. It focuses on privacy-preserving learning\, corruption-resilient representations\, language-driven wireless sensing\, and adversarial security.

\n

The first component develops a hybrid D NN and genetic algorithm framework for fine-grained location privacy in 6G Integrated Sensing and Communication (ISAC) systems. The framework genera tes high-dimensional\, physically constrained subcarrier obfuscation vecto rs that control location detection resolution by making the true channel i ndistinguishable from channels associated with selected dummy locations. W hile protecting sensitive information is essential\, trustworthy wireless sensing also requires learning models that remain effective when the under lying channel observations are unreliable. Therefore\, the second componen t introduces a two-stage corruption-resilient learning architecture for hu man pose and shape estimation under corrupted Channel State Information (C SI). The framework first mitigates unreliable observations at the channel level and then addresses residual corruption at the representation level t hrough a dual-stream architecture that uses reliable representations as re ference targets to progressively constrain corrupted representations. This approach improves the reliability of learning-based sensing under corrupt ed observations\, but conventional wireless sensing models still rely heav ily on predefined activity labels and temporal segmentation\, which limits their ability to understand continuous and unseen human activities. To ov ercome this limitation\, this research further develops WirelessSenseLLM\, a language-driven framework for wireless human sensing. A specialized wir eless encoder\, CSI-to-Language Adapter\, and cross-domain projection mech anism bridge continuous CSI and human language\, enabling segmentation-fre e and zero-shot understanding of sequential and overlapping human motions.

\n

Building on these learning frameworks\, the proposed research wil l investigate backdoor vulnerabilities and robust representation learning in wireless sensing models. This work will study how adversarial triggers alter learned representations and induce attacker controlled model behavio r\, and will develop mechanisms to detect and mitigate such manipulation w hile preserving benign representations and downstream sensing performance.

\n

In summary\, this dissertation advances both deep learning method ologies and wireless sensing systems. It develops new learning approaches for privacy protection\, corruption-resilient representations\, cross-moda l signal-language alignment\, and adversarial robustness. Together\, these contributions provide a foundation for developing trustworthy deep learni ng models for wireless sensing applications.

\n

ADVISOR(s):  Dr. Lon g Jiao\, Department of Computer and Information Science (ljiao@umassd.edu)                               

\n

COMMITTEE MEMBERS:< /p>\n

\n

NOTE: All EAS Students are ENCOURAGED to attend.

Event page: https://www.umassd.ed u/events/cms/9-9-26-eas-doctoral-proposal-defense--by-mahmuda-akter-keya.p hp
Event link: https://umassd.zoom.us/j/4061999032?pwd=bUw 0WGpDbTQ4UzJneFd5TTBFeUw1dz09

DTSTAMP:20260818T224141 DTSTART;TZID=America/New_York:20260909T110000 DTEND;TZID=America/New_York:20260909T130000 LOCATION:DION 311 SUMMARY;LANGUAGE=en-us:EAS Doctoral Proposal Defense by Mahmuda Akter Keya UID:80ee7841a8399992c17a09f850b78d98@www.umassd.edu END:VEVENT END:VCALENDAR