BEGIN:VCALENDAR
VERSION:2.0
PRODID:Data::ICal 0.24
BEGIN:VEVENT
DESCRIPTION:   'Title: Trust No History: Why Every "Remembered" Interaction
  is a\n   Potential Backdoor\n   Tags: Intermediate | AppSec Village | Cre
 ator Talk/Panel\n   When: Friday\, Aug 7\, 17:30 - 17:59 PDT\n   Where: LV
 CCW Level 1 Hall 2 604 (Appsec Village) Main Stage - [1]Map\n\n   Descript
 ion:\n\n   As AI transitions from stateless tools to autonomous agents\, t
 he\n   context window has become the primary attack surface. By giving age
 nts\n   the ability to remember\, summarize\, and collaborate\, we have cr
 eated a\n   machine that can be gaslit. This session moves beyond transien
 t prompt\n   injections into the realm of persistent memory corruption. We
  explore\n   how an adversary can rewrite an agent’s history\, bias its 
 knowledge\n   base\, and plant sleeper instructions that trigger long afte
 r the\n   initial interaction. We will dissect the systematic subversion o
 f the\n   agentic memory stack and demonstrate why developers must stop tr
 eating\n   agent memory as a passive data store and start defending it as 
 the\n   engine of the agent’s survival\n\n   Speakers:Barno Kaharova\,Ri
 co Komenda\n\n   SpeakerBio:  Barno Kaharova\n\n   Barno is a consultant a
 nd researcher specializing in AI security\, data\n   engineering\, and ML 
 security. She works at the intersection of\n   offensive and defensive AI 
 security\, developing methodologies to\n   protect AI systems from adversa
 rial threats across the full stack from\n   data pipelines and ML models t
 o LLM-powered applications and\n   autonomous agents. As an AI security tr
 ainer\, Barno designs and\n   delivers hands-on programs covering AI red t
 eaming\, prompt injection\,\n   adversarial ML\, RAG and vector database s
 ecurity\, and the OWASP Top 10\n   for LLM and Agentic AI. Her approach is
  practitioner-first:\n   participants attack and defend real-world AI syst
 ems\, leaving with\n   techniques they can apply immediately. She is activ
 ely involved in AI\n   governance and evaluation efforts at the national l
 evel and regularly\n   contributes to the AI security community through sp
 eaking engagements\,\n   publications\, and conference submissions\, bridg
 ing rapid AI adoption\n   with the need for robust\, field-tested defenses
 .\n\n   SpeakerBio:  Rico Komenda\n\n   Rico Komenda is a security enginee
 r and researcher specializing in\n   application security\, cloud security
 \, and AI security. He started as a\n   software developer\, moved into se
 curity engineering\, and now focuses\n   on the attack surface created by 
 AI-integrated systems and secure\n   agentic development.\n\n   He is Co-L
 ead of the OWASP AI Security Verification Standard (AISVS)\n   and a core 
 contributor to multiple OWASP GenAI security projects. Rico\n   has spoken
  at NDC Security\, OWASP Global AppSec EU\, OWASP LASCON\,\n   WeAreDevelo
 pers World Congress\, DefCamp\, and others.\n\n   He got into this field t
 he way a lot of hackers do: writing scripts\n   for video games as a teena
 ger\, getting curious about how systems\n   break\, and never stopping.\n\
 n   '\n\n   1. #LVCCW_Level1_Hall2\n\n\n
DTEND:20260808T005900Z
DTSTART:20260808T003000Z
LOCATION:AppSec Village - LVCCW Level 1 Hall 2 604 (Appsec Village) Main St
 age
SUMMARY:Trust No History: Why Every "Remembered" Interaction is a Potential
  Backdoor
END:VEVENT
END:VCALENDAR