Top 3 Numberof Attacking Hosts from JP (2017)
445/tcp
(2,000 IP address)
22/tcp
(14,000 IP address)
23/tcp
(24,000 IP address)
Jan 1 – Dec 31, 2017
- Unique number of source IP addresses from JP / day -
10.
Infected Devices inJP(2017)
Weekly ASCII
http://weekly.ascii.jp/elem/000/000/404/404196/
10
445/tcp (SMB)
May, 2017
WannaCry (Windows)
Logitec
http://www.logitec.co.jp/info/wireless-router.html
Symantec
https://www.symantec.com/security_response/writeup.jsp?docid=2017-051310-3522-99
22/tcp (SSH)
June, 2017
Mobile Router (DoCoMo)
23/tcp (telnet)
Nov, 2017
Home Router (Logitec)
NICT
1st
discovere
d
NICT
1st
discovere
d
!
(1) DBD AttackMonitoring
web browser plug-in sensor
(2) DBD Attack Analysis
macroscopic analysis of
users’ web accesses
(3) DBD Attack Countermeasure
web access stopper
Solution: User as a Sensor
17.
A Tachikoma isa fictional walker with artificial intelligence (AI) from the Ghost in the
Shell universe, appearing in the manga (created by Masamune Shirow) and in the Stand
Alone Complex sub-universe. Nine of them are initially deployed to Section 9. They are
spider-like, multi-legged combat vehicles, and are equipped with adaptive artificial
intelligence. (Wikipedia, Jun 18, 2018)
Web-based Attack Response with Practical and Deployable Research InitiatiVE
WarpDrive project makes Tachikoma as...
1. Sensor in the browser
2. Actuator to block web-based attacks
3. Communicator with users
17
世界中がサイバーセキュリティへのAI活用を検討
欧州
• EPFL
• FrauhoferFKIE
• Max Planck Institute for Informatics
• RWTH Aachen University
• Siemens CERT
• Universidade de Lisboa
米国
• Boston University
• Columbia University
• Florida Institute of
Technology
• Google Inc
• Indiana University
• Iowa State University
• MIT
• UC Santa Barbara
• University of Chicago
• University of Delaware
• University of Illinois
• University of Maryland
• Virginia Tech
イスラエル
• Bar-Ilan
Uniersity
アジア
• Chinese Academy
of Science
• Beijing Jiaotong
University
近年になって、世界中の著名な研究組織がAIの適用可能性を模索
USENIX Security 2018にてAI関連の報告を実施した組織は下記の通り
24.
世界中がサイバーセキュリティへのAI活用を検討
欧州
• Lancaster University
•University College London
米国
• University of Central Florida
• Florida International
University
• Northwest University
• Lehigh University
• The Pennsylvania State
University
• Virginia Tech
• University of Pennsylvania
• Symantec
• UC Riverside
• UC Berkeley
• University of Illinois at
Urbana-Champaign
• University of
Massachusetts
アジア
• Inha University
• Peking University
• Zhejiang University
• The Hong Kong Polytechnic
University
• Chinese Academy of
Sciences
• Hanyang University
• National University of
Singapore
近年になって、世界中の著名な研究組織がAIの適用可能性を模索
CCS 2018にてAI関連の報告を実施した組織は下記の通り
25.
最近報告されている研究内容
25
機械学習の脆弱性
• Poisoning attacks
•Vulnerabilities of transfer
learning
• Attribute inference
attacks
• Model reuse attack
トラフィックの異常検知&マルウェア検知
(long standing area)
• Explainable system
• Performance improvements
/real-time operations
非匿名化 (プライバシーに対する攻撃)
• Code Authorship Identification
• Document author attribute classification
• Identification of account pertaining
review comments
各種コンピューティングシステムへの攻撃
• Solving captcha
• Malfunctioning voice recognition systems
対策・防衛技術
• Program debloating
(minimize vulnerabilities)
• Watermarking DNN
• Event prediction
26.
NICTはAI x cybersecurityの重要性をいち早く認識
26
•11th International Data Mining and Cybersecurity Workshop (DMC),
2018: ICONIP併催
• 9th International Cybersecurity Data Mining Competition (CDMC),
2018: DMC併催