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:20260803T022355Z UID:CE02D2E4-AFE4-4E6F-81EC-EBD2A503F689 DTSTART;TZID=America/Los_Angeles:20260812T180000 DTEND;TZID=America/Los_Angeles:20260812T200200 DESCRIPTION:Cyber Security of Sensor Systems: an AI Approach with Explainab le Anomaly Root Cause & Type Classification Abstract: Due to possible deva stating consequences\, counteracting sensor data attacks is an extremely i mportant topic\, which has not seen sufficient study. This presentation pr esents the first methods that accurately identify/eliminate only the probl ematic attacked sensor data presented to a sequence estimation/regression algorithm under a powerful attack model. The approach does not assume a kn own form for the statistical model of the sensor data\, allowing data-driv en and machine learning sequence estimation/regression algorithms to be pr otected. A simple protection approach for attackers not endowed with knowl edge of the details of our protection approach is first developed\, follow ed by additional processing for attacks based on protection system knowled ge. Experimental results show that the simple approach achieves performanc e indistinguishable from that for an approach which knows which sensors ar e attacked. For cases where the attacker has knowledge of the protection a pproach\, experimental results indicate the additional processing can be c onfigured so that the worst-case degradation under the additional processi ng and a large number of sensors attacked can be made significantly smalle r than the worst-case degradation of the simple approach\, and close to an approach which knows which sensors are attacked\, with just a slight degr adation under no attacks. Mathematical descriptions of the worst-case atta cks are used to demonstrate the additional processing will provide similar advantages for cases for which we do not have numerical results. All the data-driven processing used in our approaches employs only unattacked trai ning data. Explainable anomaly root cause analysis and type classification are discussed.\n\nCo-sponsored by: Product Safety Chapter \n\nSpeaker(s): Dr. Rick S. Blum\n\nAgenda: \nDinner & Networking 6PM PRESENTATION 6:45-8 PM\n\nBldg: Building C training room\, 10401 Roselle Street\, Advanced Tes t Equipment Corp.\, San Diego\, California\, United States\, 92121 LOCATION:Bldg: Building C training room\, 10401 Roselle Street\, Advanced T est Equipment Corp.\, San Diego\, California\, United States\, 92121 ORGANIZER:devans0179@sdsu.edu SEQUENCE:2 SUMMARY:Distinguished Lecturer-Cyber Security of Sensor Systems: an AI Appr oach with Explainable Anomaly Root Cause & Type Classification URL;VALUE=URI:https://events.vtools.ieee.org/m/570878 X-ALT-DESC:Description: <br /><p>Cyber Security of Sensor Systems: an AI Ap proach with Explainable Anomaly Root Cause &amp\; Type Classification Abst ract: Due to possible devastating consequences\, counteracting sensor data attacks is an extremely important topic\, which has not seen sufficient s tudy. This presentation presents the first methods that accurately identif y/eliminate only the problematic attacked sensor data presented to a seque nce estimation/regression algorithm under a powerful attack model. The app roach does not assume a known form for the statistical model of the sensor data\, allowing data-driven and machine learning sequence estimation/regr ession algorithms to be protected. A simple protection approach for attack ers not endowed with knowledge of the details of our protection approach i s first developed\, followed by additional processing for attacks based on protection system knowledge. Experimental results show that the simple ap proach achieves performance indistinguishable from that for an approach wh ich knows which sensors are attacked. For cases where the attacker has kno wledge of the protection approach\, experimental results indicate the addi tional processing can be configured so that the worst-case degradation und er the additional processing and a large number of sensors attacked can be made significantly smaller than the worst-case degradation of the simple approach\, and close to an approach which knows which sensors are attacked \, with just a slight degradation under no attacks. Mathematical descripti ons of the worst-case attacks are used to demonstrate the additional proce ssing will provide similar advantages for cases for which we do not have n umerical results. All the data-driven processing used in our approaches em ploys only unattacked training data. Explainable anomaly root cause analys is and type classification are discussed.</p><br /><br />Agenda: <br /><p> Dinner &amp\; Networking 6PM PRESENTATION 6:45-8PM</p> END:VEVENT END:VCALENDAR