{"id":{"repo_id":"msu","oai_identifier":"oai:d.lib.msu.edu:etd_53018"},"canonical_url":"https://search.ndltd.org/etd/msu/oai:d.lib.msu.edu:etd_53018","repository":{"repo_id":"msu","name":"Michigan State University","base_url":"https://d.lib.msu.edu/oai"},"display":{"title":"Securing and Improving Earbud Voice Interfaces : Attacks, Authentication, and Speech Enhancement","abstract":"Thesis (Ph.D.)--Michigan State University. Computer Science - Doctor of Philosophy, 2026","abstract_html":"Thesis (Ph.D.)--Michigan State University. Computer Science - Doctor of Philosophy, 2026","abstract_has_math":false,"creators":["Li, Gen"],"institution":null,"degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":["Cao, Zhichao","Li, Tianxing","Ren, Jian","Yan, Qiben"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2026,"date_issued":"2026","date_published":"2026","updated_at":"2026-11-18T18:59:34Z","subjects":["Computer science"],"languages":["eng"],"rights":["Attribution-NonCommercial 4.0 International"],"rights_urls":[],"identifier_entries":[{"key":"dc:identifier","label":"Identifier","values":["etd:53018","local:Li_grad.msu_0128D_22022"],"render_values":[{"text":"etd:53018","href":null,"code":true},{"text":"local:Li_grad.msu_0128D_22022","href":null,"code":true}]}],"person_identifier_entries":[]},"links":{"outbound_url":"https://doi.org/doi:10.25335/e6tf-y242","outbound_label":"DOI","outbound_source":"dc:identifier","record_url":"https://doi.org/doi:10.25335/e6tf-y242","record_label":"Repository record","record_kind":"doi","record_source":"dc:identifier","download_url":null,"download_label":null,"download_source":null,"access_status":null,"embargo_until":null,"access_note":null},"source_record":{"url":"https://d.lib.msu.edu/oai?verb=GetRecord&metadataPrefix=oai_dc&identifier=oai%3Ad.lib.msu.edu%3Aetd_53018","prefix":"oai_dc"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Cao, Zhichao","Li, Tianxing","Ren, Jian","Yan, Qiben"]},{"key":"dc:creator","label":"Author","values":["Li, Gen"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date","label":"Dc Date","values":["2026"]},{"key":"dc:relation","label":"Dc Relation","values":["Electronic Theses & Dissertations"]},{"key":"dc:type","label":"Dc Type","values":["text","Electronic dissertations"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Computer science"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["eng"]},{"key":"dc:rights","label":"Dc Rights","values":["Attribution-NonCommercial 4.0 International"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["etd:53018","local:Li_grad.msu_0128D_22022","https://doi.org/doi:10.25335/e6tf-y242"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["Thesis (Ph.D.)--Michigan State University. Computer Science - Doctor of Philosophy, 2026","Earable devices such as wireless earbuds are increasingly used as the front-end for voice assistants, calls, and dictation. Unlike smartphones and smart speakers, earbuds are worn close to the user and are expected to work in highly variable acoustic environments while operating under strict size, power, and compute budgets. These properties reshape the end-to-end voice pipeline: they affect what attacks are feasible, what defenses are deployable, and how reliably speech can be captured for downstream tasks. This dissertation studies earbud voice interfaces as a systems problem and contributes methods that span vulnerability characterization, on-device authentication, and real-time speech enhancement.We first examine the acoustic attack surface of earbud-mediated voice interaction and introduce EchoAttack, a practical inaudible command injection attack tailored to smart earbuds. Earbuds differ from prior ultrasound attack targets in microphone configuration, signal paths, and user proximity. EchoAttack exploits both direct and indirect acoustic paths and uses a model-guided path-search strategy to identify effective injection routes in real environments. To improve reliability, it applies harmonic-noise suppression to increase the signal-to-noise ratio of injected commands. We further demonstrate a hidden feedback channel that enables stealthy adaptation by monitoring Bluetooth activity with commodity Zigbee radios. Across representative indoor and outdoor settings and multiple commercial earbuds, EchoAttack achieves substantially higher success than a direct-path baseline, showing that earbud voice interfaces can be manipulated in realistic conditions.Motivated by these observations and by deployment constraints, we then develop piezoelectric-enhanced multimodal sensing as an earable-friendly approach to improve both access control and speech quality. For authentication, we present PiezoBud, a lightweight speaker authentication system that combines microphone audio with a miniature piezoelectric sensor capturing body-coupled vibrations. This auxiliary channel is compact, energy-efficient, and difficult to reproduce using replay or deepfake audio alone, providing a practical robustness gain under on-device latency constraints. In a study with 85 participants, PiezoBud achieves low error rates with short enrollment time and maintains real-time responsiveness on commodity smartphones under everyday conditions such as ambient noise, user motion, and in-ear media playback. For speech quality, we introduce PiezoSE, a real-time multimodal speech enhancement system for earbuds. PiezoSE fuses the noise-resilient vibration channel with microphone audio using a compact neural architecture designed for on-device inference. Using a dataset collected from 70 participants across diverse acoustic conditions, PiezoSE improves perceptual speech quality over prior enhancement baselines while keeping the model footprint small and meeting strict latency requirements on commercial mobile hardware.Overall, this dissertation links earbud-specific vulnerability analysis with deployable multimodal designs, and it provides an end-to-end view of how to secure and improve earbud voice interfaces under real-world sensing and compute constraints.","Description based on online resource. Title from PDF t.p. (Michigan State University Fedora Repository, viewed ).","Includes bibliographical references."]},{"key":"dc:format","label":"Dc Format","values":["application/pdf","1 online resource (132 p.) : ill."]},{"key":"dc:title","label":"Title","values":["Securing and Improving Earbud Voice Interfaces : Attacks, Authentication, and Speech Enhancement"]}]}],"canonical_facts":{"dc:contributor":["Cao, Zhichao","Li, Tianxing","Ren, Jian","Yan, Qiben"],"dc:creator":["Li, Gen"],"dc:date":["2026"],"dc:description":["Thesis (Ph.D.)--Michigan State University. Computer Science - Doctor of Philosophy, 2026","Earable devices such as wireless earbuds are increasingly used as the front-end for voice assistants, calls, and dictation. Unlike smartphones and smart speakers, earbuds are worn close to the user and are expected to work in highly variable acoustic environments while operating under strict size, power, and compute budgets. These properties reshape the end-to-end voice pipeline: they affect what attacks are feasible, what defenses are deployable, and how reliably speech can be captured for downstream tasks. This dissertation studies earbud voice interfaces as a systems problem and contributes methods that span vulnerability characterization, on-device authentication, and real-time speech enhancement.We first examine the acoustic attack surface of earbud-mediated voice interaction and introduce EchoAttack, a practical inaudible command injection attack tailored to smart earbuds. Earbuds differ from prior ultrasound attack targets in microphone configuration, signal paths, and user proximity. EchoAttack exploits both direct and indirect acoustic paths and uses a model-guided path-search strategy to identify effective injection routes in real environments. To improve reliability, it applies harmonic-noise suppression to increase the signal-to-noise ratio of injected commands. We further demonstrate a hidden feedback channel that enables stealthy adaptation by monitoring Bluetooth activity with commodity Zigbee radios. Across representative indoor and outdoor settings and multiple commercial earbuds, EchoAttack achieves substantially higher success than a direct-path baseline, showing that earbud voice interfaces can be manipulated in realistic conditions.Motivated by these observations and by deployment constraints, we then develop piezoelectric-enhanced multimodal sensing as an earable-friendly approach to improve both access control and speech quality. For authentication, we present PiezoBud, a lightweight speaker authentication system that combines microphone audio with a miniature piezoelectric sensor capturing body-coupled vibrations. This auxiliary channel is compact, energy-efficient, and difficult to reproduce using replay or deepfake audio alone, providing a practical robustness gain under on-device latency constraints. In a study with 85 participants, PiezoBud achieves low error rates with short enrollment time and maintains real-time responsiveness on commodity smartphones under everyday conditions such as ambient noise, user motion, and in-ear media playback. For speech quality, we introduce PiezoSE, a real-time multimodal speech enhancement system for earbuds. PiezoSE fuses the noise-resilient vibration channel with microphone audio using a compact neural architecture designed for on-device inference. Using a dataset collected from 70 participants across diverse acoustic conditions, PiezoSE improves perceptual speech quality over prior enhancement baselines while keeping the model footprint small and meeting strict latency requirements on commercial mobile hardware.Overall, this dissertation links earbud-specific vulnerability analysis with deployable multimodal designs, and it provides an end-to-end view of how to secure and improve earbud voice interfaces under real-world sensing and compute constraints.","Description based on online resource. Title from PDF t.p. (Michigan State University Fedora Repository, viewed ).","Includes bibliographical references."],"dc:format":["application/pdf","1 online resource (132 p.) : ill."],"dc:identifier":["etd:53018","local:Li_grad.msu_0128D_22022","https://doi.org/doi:10.25335/e6tf-y242"],"dc:language":["eng"],"dc:relation":["Electronic Theses & Dissertations"],"dc:rights":["Attribution-NonCommercial 4.0 International"],"dc:subject":["Computer science"],"dc:title":["Securing and Improving Earbud Voice Interfaces : Attacks, Authentication, and Speech Enhancement"],"dc:type":["text","Electronic dissertations"]},"updated_at":"2026-11-18T18:59:34Z"}