Keystroke dynamics analysis for user authentication and account sharing detection

Affordable Access

Keystroke dynamics analysis for user authentication and account sharing detection

서울대학교 대학원
  • Data Mining
  • Internet
  • Classification
  • Clustering
  • Password
  • Biometrics
  • Data Quality
  • 인터넷
  • Mobile Device
  • 분류
  • Novelty Detection
  • Account Management
  • 데이터마이닝
  • User Authentication
  • 암호
  • Keystroke Dynamics
  • Tempo Cues
  • Artificial Rhythms
  • 계정관리
  • 생체인식
  • 키스트로크 다이나믹스
  • 타이핑 패턴
  • Typing Pattern
  • 사용자 인증
  • 이상탐지
  • 데이터 품질
  • 인공리듬
  • 템포큐
  • 모바일 기기
  • 계정공유
  • Account Sharing
  • 군집화


As the internet environment has been growing rapidly for decades, one of the primary concerns of internet companies is managing customer accounts, the main source of revenue. As an advanced tool to manage customer accounts, we introduce keystroke dynamics analysis, which uses customers¡¯ biometrics information. The term biometrics is defined as the automated use of physiological (such as fingerprint, face, hand geometry and iris) or behavioral (such as signature, keystroke dynamics and voice) characteristics to determine or verify identity. Biometrics is an attractive complement to the historical password scheme for security reinforcement in verifying identity. Keystroke dynamics provide a method of analyzing the way a user types by monitoring the keyboard inputs in an attempt to identify users based on habitual typing rhythm patterns. A password of m characters is transformed into (2m-1)-dimensional timing vector. ¡°Duration¡± denotes a time period during which a key is pressed while an ¡°Interval" is a time period between releasing a key and stroking the next key. Unlike other biometrics-based methods, it can be very efficient in terms of cost and convenience. This dissertation demonstrates that we can practically use keystroke dynamics analysis in user authentication and account sharing detection. For user authentication, we show its viability in keyboard and keypad applications while addressing various practical issues. There are three steps in a keystroke dynamics-based authentication (KDA) framework: enrollment, classifier building, and login. In the enrollment stage, a user enrolls his keystroke patterns. He only needs to type his user ID and password as he always does since a KDA system automatically collects his keystroke patterns. These enrolled patterns form a reference set. Then the system builds a classifier using the reference set. The classifier plays the role of authenticator in the login stage where it accepts or rejects a newly presented keystroke base

There are no comments yet on this publication. Be the first to share your thoughts.