BEGIN:VCALENDAR
VERSION:2.0
PRODID:Data::ICal 0.24
BEGIN:VEVENT
DESCRIPTION:   'Title: Assessing the Vulnerabilities of the Open-Source Art
 ificial\n   Intelligence (AI) Landscape: A Large-Scale Analysis of the Hug
 ging\n   Face Platform\n   When: Friday\, Aug 11\, 12:00 - 12:25 PDT\n   W
 here: Caesars Forum - Academy - 401-406 - AI Village - [1]Map\n   Speakers
 :Adhishree Kathikar\,Aishwarya Nair\n\n   SpeakerBio:Adhishree Kathikar\n 
   No BIO available\n\n   SpeakerBio:Aishwarya Nair\n   No BIO available\n\
 n   Description:\n   Artificial Intelligence (AI) has earned its title as 
 one of the most\n   critical disruptive technologies in the 21st century. 
 As AI develops\n   at a rapid rate\, open-source software (OSS) platforms 
 develop\n   alongside it. Hugging Face is one of these prevailing OSS plat
 forms as\n   it hosts pre-trained AI models\, facilitating the accessibili
 ty of AI\n   models. Hugging Face is used by over 22\,000 organizations\, 
 including\n   Intel and Microsoft\, has supported more than 2.6 billion mo
 del\n   downloads\, and is rapidly growing. Just in the past year\, Huggin
 g Face\n   has more than doubled its model library from 80\,000 models to 
 203\,000\n   models. However\, while Hugging Face democratizes access to A
 I models\,\n   these models may contain unknown security vulnerabilities. 
 Our\n   research focuses on automating our collection process of Hugging F
 ace\n   models\, linking them to their primary codebases on GitHub\, and\n
    executing a large-scale vulnerability assessment of these GitHub\n   re
 positories using static scanners. We collected more than 110\,000\n   Hugg
 ing Face models and over 29\,000 GitHub repositories. Our\n   vulnerabilit
 y assessment of these GitHub models depicted that 35.98%\n   of the severi
 ties detected from the root GitHub repositories\n   (developed by Hugging 
 Face) were high-severity vulnerabilities while\n   only 6.79% were low-sev
 erity. On the other hand\, 82.89% of\n   vulnerabilities from searched rep
 ositories (determined through the\n   ‘huggingface’ keyword) are low-s
 everity and 7.49% high-severity\,\n   while 82.69% of vulnerabilities from
  the repositories forked from the\n   root repositories were low-severity 
 and 9.22% were high-severity. The\n   trend in severity levels found in ro
 ot repositories contradicts the\n   results of severities detected in fork
 ed and searched repositories.\n   Given that many of the vulnerabilities r
 eside in fundamental AI\n   repositories such as Transformers\, this vulne
 rability assessment has\n   significant implications for supply chain soft
 ware security and AI\n   risk management more broadly.\n   '\n\n   1. #Cae
 sarsAcademyBR\n\n\n
DTEND:20230811T192500Z
DTSTART:20230811T190000Z
LOCATION:AIV - Caesars Forum - Academy - 401-406 - AI Village
SUMMARY:Assessing the Vulnerabilities of the Open-Source Artificial Intelli
 gence (AI) Landscape: A Large-Scale Analysis of the Hugging Face Platform
END:VEVENT
END:VCALENDAR