Not Quite Yourself Today: Behaviour-Based Continuous Authentication in IoT Environments
summary
An IoT testbed captures typing, plate force, infrared/Hall movement and PC telemetry and releases 16 hours from 24 sensors as a public dataset. Static models give up to 68% identification (7% baseline) among 20 people, so the system instead trains a continuously updating random forest that reaches 99.3% verification and stops simulated session-takeover attacks in about 12 seconds with ≲1% false detections.
AI-written summary · about
related talks
44mIndustrial IoT Security: Best Practice for Authentication
04mDBTB INT John Whaley r
47mPatterns of Authentication and Self-Announcement in IoT
22mThe high cost of cheap biometrics
25mContinuous Biometrics
20mThe End of User Authentication & the Coming Auto Authentication by the Server
20mInjection Attacks in Biometric
11mLet The Right One In: Attestation as a Usable CAPTCHA Alternative