Loading
Life sciences · Preprint
arXiv · September 4, 2026
Early or partial results. Treat as a signal, not a conclusion.
This preprint reports that choice of gaze representation (raw, heatmap, or engineered features) materially alters the privacy-utility tradeoff in XR systems. Engineered eye-movement features retain ~85% action-recognition accuracy while reducing re-identification risk by ~10-fold to near chance on a 206-identity closed-set task, but do not eliminate identity leakage and lack formal privacy guarantees.
Single-dataset empirical comparison across three gaze representations. Users in the HoloAssist egocentric dataset; specific eligibility, setting, or demographics not stated.. Intervention: Choice of gaze representation: engineered eye-movement features, spatial attention heatmaps, or raw gaze signals. Compared with: Raw gaze (baseline) and three representations versus each other.
Engineered features retain roughly 85% of raw gaze's action-recognition accuracy Re-identification reduced by about an order of magnitude using engineered features, to roughly four times chance rate across 206 identities Abstraction alone does not guarantee privacy; differences across representations show leakage persists
Safety was not reported in the material analysed. Check the source before drawing any conclusion about harm.
The source did not state who this applies to in practice.
Single-dataset empirical study of a design intervention (gaze representation choice) on privacy-utility tradeoff in XR, with no peer review, no randomization, and no comparison to established baselines or formal privacy guarantees.
As stated by the source record.
Quoted from the source exactly as published.
Graded across the dimensions that decide whether you should act, each from what the source actually supports. There is no single score, and where a dimension was not assessed it says so.
Intelligent extended reality (XR) systems increasingly use eye and head tracking to infer user intent, task, and attention, but the same signals can also reveal biometric identity. We study whether gaze data representation choice can serve as a lightweight privacy control at feature extraction, before adding perturbation or formal privacy mechanisms. Using the egocentric HoloAssist dataset, we compare three gaze representations under matched model capacity: raw gaze, spatial attention heatmaps, and engineered eye-movement features. We evaluate each representation on action recognition as task utility and closed-set user re-identification as privacy leakage. Representation choice substantially changes the privacy-utility tradeoff. Engineered features retain roughly 85% of raw gaze's action-recognition accuracy while reducing re-identification by about an order of magnitude, to roughly four times the chance rate across 206 identities. This reduction attenuates rather than eliminates identity leakage, and the differences across representations show that abstraction alone does not guarantee privacy. Engineered features expose interpretable and auditable structure, giving designers a transparent privacy lever that complements mechanisms such as differential privacy.
Taken from the source record, never inferred. Follow any of these and new work involving them reaches your briefing.