BEGIN:VCALENDAR
VERSION:2.0
PRODID:IEEE vTools.Events//EN
CALSCALE:GREGORIAN
BEGIN:VTIMEZONE
TZID:Asia/Shanghai
BEGIN:STANDARD
DTSTART:19910915T010000
TZOFFSETFROM:+0900
TZOFFSETTO:+0800
TZNAME:CST
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260705T052631Z
UID:D2967D23-DF3C-4DE2-8B43-236CAB0D2456
DTSTART;TZID=Asia/Shanghai:20260709T103000
DTEND;TZID=Asia/Shanghai:20260709T233000
DESCRIPTION:Over the past few years\, machine learning has been widely util
 ized to enhance cybersecurity. In this talk\, we will present two of our r
 ecent projects in this area: the use of autoencoders to defend against adv
 ersarial attacks and the detection of darknet traffic over Tor. In the fir
 st project\, we observed that machine learning-based intrusion detection s
 ystems (IDSs) have become increasingly popular due to their strong detecti
 on capability and adaptability. However\, these IDSs remain vulnerable to 
 adversarial attacks\, in which small perturbations to input features can c
 ause misclassification. Although autoencoders (AEs) have shown promise in 
 mitigating adversarial attacks in computer vision\, their potential for im
 proving IDS robustness remains largely unexplored. In this project\, we pr
 oposed a novel AE-based scheme for detecting adversarial network flows. In
  the second project\, we found that the rapid growth of encrypted communic
 ation has strengthened user privacy but has also introduced challenges for
  network monitoring\, cybersecurity\, and digital forensics. Among anonymi
 ty technologies\, Tor has gained significant popularity due to its ability
  to conceal user identity and traffic origins. While Tor serves legitimate
  privacy needs\, it can also be exploited for cybercrime\, command-and-con
 trol communication\, and darknet marketplace access. Therefore\, accuratel
 y distinguishing Tor traffic from regular Internet traffic is an important
  task in modern network security systems. In this project\, we developed a
  state-of-the-art dataset for Tor traffic detection and evaluated the perf
 ormance of machine learning-based detection schemes.\n\nSpeaker(s): Qiang 
 Ye\, \n\nRoom: W2-201\, HKUST(GZ)\, Guangzhou\, Guangdong\, China
LOCATION:Room: W2-201\, HKUST(GZ)\, Guangzhou\, Guangdong\, China
ORGANIZER:gongzijun@hkust-gz.edu.cn
SEQUENCE:162
SUMMARY:Machine Learning for Cybersecurity: Defending Against Adversarial A
 ttacks and Detecting Darknet Traffic
URL;VALUE=URI:https://events.vtools.ieee.org/m/564675
X-ALT-DESC:Description: <br /><p>Over the past few years\, machine learning
  has been widely utilized to enhance cybersecurity. In this talk\, we will
  present two of our recent projects in this area: the use of autoencoders 
 to defend against adversarial attacks and the detection of darknet traffic
  over Tor. In the first project\, we observed that machine learning-based 
 intrusion detection systems (IDSs) have become increasingly popular due to
  their strong detection capability and adaptability. However\, these IDSs 
 remain vulnerable to adversarial attacks\, in which small perturbations to
  input features can cause misclassification. Although autoencoders (AEs) h
 ave shown promise in mitigating adversarial attacks in computer vision\, t
 heir potential for improving IDS robustness remains largely unexplored. In
  this project\, we proposed a novel AE-based scheme for detecting adversar
 ial network flows. In the second project\, we found that the rapid growth 
 of encrypted communication has strengthened user privacy but has also intr
 oduced challenges for network monitoring\, cybersecurity\, and digital for
 ensics. Among anonymity technologies\, Tor has gained significant populari
 ty due to its ability to conceal user identity and traffic origins. While 
 Tor serves legitimate privacy needs\, it can also be exploited for cybercr
 ime\, command-and-control communication\, and darknet marketplace access. 
 Therefore\, accurately distinguishing Tor traffic from regular Internet tr
 affic is an important task in modern network security systems. In this pro
 ject\, we developed a state-of-the-art dataset for Tor traffic detection a
 nd evaluated the performance of machine learning-based detection schemes.<
 /p>
END:VEVENT
END:VCALENDAR