BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/New_York BEGIN:DAYLIGHT DTSTART:20260308T030000 TZOFFSETFROM:-0500 TZOFFSETTO:-0400 RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3 TZNAME:EDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20261101T010000 TZOFFSETFROM:-0400 TZOFFSETTO:-0500 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11 TZNAME:EST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20260902T173213Z UID:33E1F326-DB26-4494-A1A0-5D5848BF440F DTSTART;TZID=America/New_York:20260925T170000 DTEND;TZID=America/New_York:20260925T175000 DESCRIPTION:This session explores why traditional identity and access contr ols fall short when autonomous AI agents can invoke tools\, access resourc es\, and delegate work across systems. It introduces context graphs—a pr actical way to model agent identities\, permissions\, tools\, MCP servers\ , resources\, and delegation paths—so security teams can evaluate the co mplete chain of activity instead of reviewing each API call in isolation. The talk covers excessive agency\, privilege chaining\, MCP tool poisoning \, lateral movement\, and graph-aware policy enforcement. A live Java demo nstration will walk through three scenarios: a legitimate request\, a bloc ked privilege-escalation attempt\, and a risky multi-hop path that traditi onal access controls would allow.\n\n💡 Practical security patterns for agentic AI\n🖥️ Live policy-engine demonstration\n💬 Interactive aud ience Q&A\n\nOpen to engineers\, architects\, security professionals\, stu dents\, and anyone interested in building and governing secure AI agents. Register soon!\n\nCo-sponsored by: Central Indiana IEEE Computer Society\n \nVirtual: https://events.vtools.ieee.org/m/572100 LOCATION:Virtual: https://events.vtools.ieee.org/m/572100 ORGANIZER:siva.bayyavarapu@ieee.org SEQUENCE:106 SUMMARY:Securing Agentic AI with Context Graphs URL;VALUE=URI:https://events.vtools.ieee.org/m/572100 X-ALT-DESC:Description: <br /><p>This session explores why traditional iden tity and access controls fall short when autonomous AI agents can invoke t ools\, access resources\, and delegate work across systems. It introduces context graphs&mdash\;a practical way to model agent identities\, permissi ons\, tools\, MCP servers\, resources\, and delegation paths&mdash\;so sec urity teams can evaluate the complete chain of activity instead of reviewi ng each API call in isolation. The talk covers excessive agency\, privileg e chaining\, MCP tool poisoning\, lateral movement\, and graph-aware polic y enforcement. A live Java demonstration will walk through three scenarios : a legitimate request\, a blocked privilege-escalation attempt\, and a ri sky multi-hop path that traditional access controls would allow.</p>\n<p> 💡 Practical security patterns for agentic AI<br>🖥️ Live policy-eng ine demonstration<br>💬 Interactive audience Q&amp\;A</p>\n<p>Open to en gineers\, architects\, security professionals\, students\, and anyone inte rested in building and governing secure AI agents. <strong>Register soon!< /strong></p> END:VEVENT END:VCALENDAR