We release an eye-tracking dataset recorded with a Tobii Pro Fusion eye tracker (120 Hz) across six tasks: (1) repeated free viewing of 100 images over three sessions (300 presentations per participant), (2) gaze-pattern authentication, (3) numeric password entry, and (4-6) three 10-minute naturalistic screen activities (shopping, news browsing, and video viewing). Our key design choice is to present the same 100 images three times to obtain stable, participant-level average gaze behavior; the image order is deterministically randomized with a fixed seed such that all participants view the same sequence. Due to copyright restrictions, we provide only the 80 copyright-free natural images collected from the internet as stimuli. The dataset is released under the CC BY 4.0 license. We do not include baseline models or benchmark evaluations, allowing researchers to apply their own post-processing and modeling approaches to the unconstrained, raw tabular data.

Traditional authentication methods, such as biometrics and passwords, cannot guarantee security once the authenticated user changes after login. Periodic re-authentication at short intervals could preserve security but would severely degrade usability. To balance security and usability, we propose an implicit and continuous identification method based on eye-tracking data. Our approach compares a saliency map personalized for the legitimate user with the gaze heatmap of the current user and identifies the current user as legitimate if the two maps are similar. To demonstrate our approach, we first develop and verify that a personalized saliency map well expresses a legitimate user’s gaze behavior. With one example result of the user study, we then illustrate the potential application of our identification as second-factor authentication and as an unobtrusive re-authentication suggestion.

Patent application filed on 2025-10-29 and currently under examination. Application number: 特願2025-182947.

There is active research on estimating emotions by analyzing facial expressions and pupil responses using facial images. Due to fake facial expressions or changes in environmental brightness, it may not be possible to estimate emotions from facial expressions or pupil diameter. Therefore, we propose a method to estimate emotions from gaze alone. This study proposes a method to classify emotions by fine-tuning gazeNet, a gaze classification model into fixation, saccade, etc. to compensate for the lack of data for deep learning. We evaluated the proposed methods in three ways, classifying emotions into four categories (delighted, relaxed, bored, and afraid), two categories (negative and positive valence), and two categories (low and high arousal), which are based on Russell’s circumplex model. The result of the four-category classification shows that emotions associated with higher arousal, i.e., delighted and afraid, exhibited higher accuracy. Classification along the valence axis suggests that valence classification using gaze data alone is difficult. Classification along the arousal axis suggests that high-arousal states achieve higher recall, precision, and F-measure compared to low-arousal states. In this study, we showed that emotion classification from gaze alone using transfer learning with a pre-trained model that classifies gaze was effective on the arousal axis.