Content-Independent Avatar Ownership Detection for Preventing Sockpuppet-Enabled Violations in th...
summary
AvatarJudger identifies all avatars controlled by the same person using 3D finger-joint telemetry rather than account metadata, claiming cross-device ownership detection with a multi-stage pipeline that combines a random-forest global model and a deep neural net for short-term habits. Evaluation on 2,217 avatars from 30 participants across HoloLens 2, Quest 2, and Vision Pro yields >94% accuracy, finds pink noise in 90% of signals, highlights left-hand fingertip–wrist distances and 44 stable features, and notes time drift and reliance on hand‑to‑keyboard input.
AI-written summary · about
chapters (7)
- 0:00Introduction to research
- 1:09Challenges in avatar identity detection
- 2:35Proposed solution using finger movement data
- 5:08Feature extraction and methodology
- 7:14Evaluation results and accuracy
- 8:40Robustness and limitations
- 10:06Q&A session
related talks
16mFull-Body Ownership Illusion Can Change Our Emotion
09mAvatar Tracking Control with Generations of Physically Natural Responses on Contact to Reduce ...
06mCoolMoves: User Motion Accentuation in Virtual Reality
26mNew Face, Who Dis? Recent Adversarial Approaches to Facial Recognition
07mNot Quite Yourself Today: Behaviour-Based Continuous Authentication in IoT Environments
06mOptimizing Shoulder-Surfing Resistance in Augmented and Virtual Reality
19mOwnershift: Facilitating Overhead Interaction in Virtual Reality with an Ownership-Preserving ...
03mThe Alt-Self: Investigating the Inclusivity of Self-Avatar Representations in Social VR