BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/Los_Angeles BEGIN:DAYLIGHT DTSTART:20260308T030000 TZOFFSETFROM:-0800 TZOFFSETTO:-0700 RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3 TZNAME:PDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20261101T010000 TZOFFSETFROM:-0700 TZOFFSETTO:-0800 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11 TZNAME:PST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20260916T040817Z UID:79776A60-95BB-44D6-B632-4516B1C8121C DTSTART;TZID=America/Los_Angeles:20260922T190000 DTEND;TZID=America/Los_Angeles:20260922T200000 DESCRIPTION:The San Francisco Bay Area chapter of the IEEE Computer Society invites to our free and open Virtual Tech Talks (no IEEE membership requi red):\n\nSpeaker: Shravya Kanchi\n\nTitle: Data Challenges in ML-based Sec urity Tasks\n\nAbstract: Machine learning classifiers are widely used for security tasks\, and efforts to improve them have largely focused on algor ithmic advances\, while the data challenges that limit their performance h ave received far less attention. This talk asks a simple question: can Gen erative AI address these data challenges? We augment training datasets wit h synthetic data produced by GenAI techniques to improve classifier genera lization\, evaluating the approach across 7 diverse security tasks using 6 state-of-the-art GenAI methods\, and introduce Nimai\, a novel scheme tha t enables highly controlled data synthesis. We find that GenAI can signifi cantly improve security classifiers\, with gains of up to 32.6% even in se verely data-constrained settings of roughly 180 training samples\, and tha t it enables rapid adaptation to concept drift after deployment with minim al labeling. We also share where it breaks down: some GenAI schemes fail t o initialize on certain security tasks\, and task characteristics such as noisy labels\, overlapping class distributions\, and sparse feature vector s hinder the benefit. The talk closes with what these findings mean for bu ilding the next generation of GenAI tools for security. This work was acce pted at ACM Asia CCS 2026 and is available at [arxiv.org/abs/2507.06092](h ttps://www.google.com/url?q=http://arxiv.org/abs/2507.06092&sa=D&source=ca lendar&ust=1789963223372342&usg=AOvVaw3HOPUVBQUJ6HQ8xxfIrVSC).\n\nBio: Shr avya Kanchi is an Applied AI Scientist at Kai Cyber Inc. She completed her Ph.D. in Computer Science at Virginia Tech\, USA\, advised by Dr. Daphne Yao. Her research sits at the intersection of security and generative AI\, spanning LLM safety\, synthetic data generation for security applications \, adversarial machine learning\, and automated security testing framework s. Her work has been published at IEEE S&P\, ACM Asia CCS\, and ACSAC. She previously earned a Masters by Research in Computer Science from IIIT Hyd erabad\, India and a B.Tech. in Computer Science from IIIT Sri City\, Indi a. She works hands-on with LLMs\, GANs\, and diffusion models applied to s ecurity-specific challenges\, with an emphasis on practical defenses and e valuation-driven systems.\n\nSpeaker(s): Shravya Kanchi\n\nVirtual: https: //events.vtools.ieee.org/m/577937 LOCATION:Virtual: https://events.vtools.ieee.org/m/577937 ORGANIZER:ruben.glatt@ieee.org SEQUENCE:27 SUMMARY:Tech Talk: Data Challenges in ML-based Security Tasks URL;VALUE=URI:https://events.vtools.ieee.org/m/577937 X-ALT-DESC:Description: <br /><p>The San Francisco Bay Area chapter of the IEEE Computer Society invites to our free and open Virtual Tech Talks (no IEEE membership required):</p>\n<p><strong>Speaker:</strong><strong>&nbsp\ ;</strong>Shravya Kanchi</p>\n<p><strong>Title: </strong>Data Challenges i n ML-based Security Tasks</p>\n<p><strong>Abstract:</strong>&nbsp\;Machine learning classifiers are widely used for security tasks\, and efforts to improve them have largely focused on algorithmic advances\, while the data challenges that limit their performance have received far less attention. This talk asks a simple question: can Generative AI address these data ch allenges? We augment training datasets with synthetic data produced by Gen AI techniques to improve classifier generalization\, evaluating the approa ch across 7 diverse security tasks using 6 state-of-the-art GenAI methods\ , and introduce Nimai\, a novel scheme that enables highly controlled data synthesis. We find that GenAI can significantly improve security classifi ers\, with gains of up to 32.6% even in severely data-constrained settings of roughly 180 training samples\, and that it enables rapid adaptation to concept drift after deployment with minimal labeling. We also share where it breaks down: some GenAI schemes fail to initialize on certain security tasks\, and task characteristics such as noisy labels\, overlapping class distributions\, and sparse feature vectors hinder the benefit. The talk c loses with what these findings mean for building the next generation of Ge nAI tools for security. This work was accepted at ACM Asia CCS 2026 and is available at <a href="https://www.google.com/url?q=http://arxiv.org/abs/2 507.06092&amp\;sa=D&amp\;source=calendar&amp\;ust=1789963223372342&amp\;us g=AOvVaw3HOPUVBQUJ6HQ8xxfIrVSC" target="_blank" rel="noopener">arxiv.org/a bs/2507.06092</a>.</p>\n<p><strong>Bio:</strong> Shravya Kanchi is an Appl ied AI Scientist at Kai Cyber Inc. She completed her Ph.D. in Computer Sci ence at Virginia Tech\, USA\, advised by Dr. Daphne Yao. Her research sits at the intersection of security and generative AI\, spanning LLM safety\, synthetic data generation for security applications\, adversarial machine learning\, and automated security testing frameworks. Her work has been p ublished at IEEE S&amp\;P\, ACM Asia CCS\, and ACSAC. She previously earne d a Masters by Research in Computer Science from IIIT Hyderabad\, India an d a B.Tech. in Computer Science from IIIT Sri City\, India. She works hand s-on with LLMs\, GANs\, and diffusion models applied to security-specific challenges\, with an emphasis on practical defenses and evaluation-driven systems.</p> END:VEVENT END:VCALENDAR