BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Los_Angeles
X-LIC-LOCATION:America/Los_Angeles
BEGIN:DAYLIGHT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20240626T180034Z
LOCATION:3002\, 3rd Floor
DTSTART;TZID=America/Los_Angeles:20240626T170000
DTEND;TZID=America/Los_Angeles:20240626T171500
UID:dac_DAC 2024_sess106_RESEARCH374@linklings.com
SUMMARY:EOS: An Energy-Oriented Attack Framework for Spiking Neural Networ
 ks
DESCRIPTION:Research Manuscript\n\nNing Yang (Shanghai Jiao Tong Universit
 y), Fangxin Liu and Zongwu Wang (Shanghai Jiaotong University), Haomin Li 
 and Zhuoran Song (Shanghai Jiao Tong University), and Songwen Pei (Univers
 ity of Shanghai for Science and Technology)\n\nSpiking neural networks (SN
 Ns) are emerging as energy-efficient alternatives to conventional artifici
 al neural networks (ANNs). Their event-driven information processing signi
 ficantly reduces computational demands while maintaining competitive perfo
 rmance. \nHowever, as SNNs are increasingly deployed in edge devices, vari
 ous security concerns have emerged. While significant research efforts hav
 e been dedicated to addressing the security vulnerabilities stemming from 
 malicious input, often referred to as adversarial examples, the security o
 f SNN parameters remains relatively unexplored. \nThis work introduces a n
 ovel attack methodology for SNNs known as Energy-Oriented SNN attack (EOS)
 . EOS is designed to increase the energy consumption of SNNs through the m
 alicious manipulation of binary bits within their memory systems (i.e., DR
 AM), where neuronal information is stored.\nThe key insight of  EOS lies i
 n the observation that energy consumption in SNN implementations is intric
 ately linked to spiking activity. \nThe bit-flip operation, the well-known
  Row Hammer technique, is employed in EOS. It achieves this by identifying
  the most robust neurons in the SNN based on the spiking activity, particu
 larly those related to the firing threshold, which is stored as binary bit
 s in memory. EOS employs a combination of spiking activity analysis and a 
 progressive search strategy to pinpoint the target neurons for bit-flip at
 tacks. The primary objective is to incrementally increase the energy consu
 mption of the SNN while ensuring that accuracy remains intact.\nWith the i
 mplementation of EOS, successful attacks on SNNs can lead to an average of
  $43\%$ energy increase with no drop in accuracy.\n\nTopic: AI, Security\n
 \nKeyword: AI/ML Security/Privacy\n\nSession Chair: Benjamin Tan (Universi
 ty of Calgary)
END:VEVENT
END:VCALENDAR