BEGIN:VCALENDAR
VERSION:2.0
PRODID:Data::ICal 0.24
BEGIN:VEVENT
DESCRIPTION:   'Title: Living Off Someone Else's Inference\n   Tags: Advers
 ary Village | Creator Talk/Panel\n   When: Saturday\, Aug 8\, 12:00 - 12:2
 5 PDT\n   Where: LVCCW Level 1 Hall 2 602 (Adversary Village) Hands-on Act
 ivity\n   Area - [1]Map\n\n   Description:\n\n   LLM-powered tooling is be
 coming a force multiplier for attackers\, from\n   reconnaissance automati
 on to post-exploitation enumeration. Using\n   their own API keys creates 
 attribution and cost\, so they look for\n   alternatives.\n\n   Thousands 
 of inference endpoints sit exposed on the internet:\n   misconfigured Olla
 ma instances\, open vLLM deployments\, leaked API keys\n   in directory li
 stings\, and expired OAuth tokens from AI coding\n   assistants that can b
 e silently refreshed. Attackers can discover\n   these resources and use t
 hem as free compute for their operations\,\n   without spending a dollar o
 r registering an account.\n\n   Attendees will learn how to:\n\n   •�  
 � Discover exposed inference endpoints and leaked API keys\n   using Infr
 eerence\n\n   •�  � Validate that discovered resources provide usable 
 inference\n   access\n\n   •�  � Operate Echidna\, powered entirely by
  discovered inference\n\n   •�  � Chain LLM skill agents together to e
 xecute a guided campaign\n   against a target network\n\n   Current LLMs a
 re effective force multipliers when directed\, and\n   significantly more 
 so when the compute bill goes to someone else. This\n   is resource hijack
 ing applied to the AI era: living off someone else's\n   inference. Unders
 tanding this threat is essential for any organization\n   deploying or exp
 osing AI infrastructure.\n\n   Two tools will be released as open source d
 uring the session:\n\n   •�  � Infreerence: A multi-phase scanner and 
 dashboard that\n   discovers exposed inference endpoints and leaked API ke
 ys through\n   Shodan and Censys\, validates them against live provider AP
 Is\, and\n   catalogs usable inference resources across 10+ providers.\n\n
    •�  � Echidna: A Mythic C2 agent type that consumes discovered\n   
 inference endpoints and turns them into operator-directed skill agents\n  
  for reconnaissance\, exploitation planning\, post-exploitation\, and\n   
 lateral movement.\n\n   Speakers:Redon Gashi\,Armend Gashi\n\n   SpeakerBi
 o:  Redon Gashi\, Managing Security Consultant at Sentry\n   Cybersecurity
 \n\n   Redon Gashi is a Managing Security Consultant and offensive team le
 ad\n   at Sentry\, where he has spent close to a decade leading and execut
 ing\n   penetration tests and red team operations for Fortune 500 clients\
 n   across banking\, telecom\, insurance\, and fintech. He holds the OSEP\
 ,\n   CRTL\, and CRTO certifications\, and has personally performed over 2
 00\n   engagements spanning web applications\, internal infrastructure\, a
 nd\n   mobile platforms\, while leading many more across his team. A forme
 r\n   lecturer at Cyber Academy\, speaker at multiple BSides conferences\,
  and\n   multiple-time CTF champion\, Redon combines deep hands-on technic
 al\n   expertise with a proven ability to build and scale offensive securi
 ty\n   teams.\n\n   SpeakerBio:  Armend Gashi\, Managing Security Consulta
 nt at Sentry\n   Cybersecurity\n\n   Armend Gashi is Managing Security Con
 sultant at Sentry. With over 5\n   years in the industry\, he specializes 
 in application security and AWS\n   cloud assessments. Armend has conducte
 d AI red teaming engagements\,\n   developed multi-agent systems to perfor
 m security-focused tasks such\n   as code auditing and exploit development
 \, and was part of\n   Anthropic’s external AI red team capability throu
 gh HackerOne.\n\n   Armend has led security research initiatives resulting
  in the\n   discovery of multiple vulnerabilities\, including:\n\n   CVE-2
 023-26482 - Critical-risk vulnerability enabling remote code\n   execution
  CVE-2023-48239 - High-risk vulnerability allowing arbitrary\n   user stor
 age updates CVE-2023-35928 - High-risk vulnerability enabling\n   user acc
 ount hijacking CVE-2023-45660 - Medium-risk application-level\n   denial o
 f service vulnerability CVE-2023-48301 - Medium-risk arbitrary\n   browser
  code injection vulnerability CVE-2023-48307 - Low-risk\n   server-side re
 quest forgery vulnerability\n\n   Links:\n       adversaryvillage.org/adve
 rsary-events/DEFCON-34/ - [2]https://adversaryvillage.org/adversary-events
 /DEFCON-34/\n   '\n\n   1. #LVCCW_Level1_Hall2\n   2. https://adversaryvil
 lage.org/adversary-events/DEFCON-34/\n\n\n
DTEND:20260808T192500Z
DTSTART:20260808T190000Z
LOCATION:Adversary Village - LVCCW Level 1 Hall 2 602 (Adversary Village) H
 ands-on Activity Area
SUMMARY:Living Off Someone Else's Inference
END:VEVENT
END:VCALENDAR