Bogazici mouse dynamics dataset

Abstract

Mouse dynamics is a hot topic under study in session-based authentication and intrusion detection in privileged access management systems. We contribute to this topic, which develops with the increase of big data capabilities, with a missing data set in the field. This data article describes an extensive dataset with free usage mouse dynamics data. This dataset has seven variables and 24 users with 2550 h of active usage data. Each user has training data, internal attack test data, and external attack test data. An application in Python programming language that continuously listens to mouse movements and mouse clicks and writes them into a file with a timestamp, foremost window name, and mouse action details, is implemented. Among 24 unique users, we labeled five users with the least amount of data as external users since their data would be weak for training purposes, and we can test external threat detection. Each user has training data consisting of 10 days with the most frequent mouse usage, and reminder days of data are used for internal attack test data. This dataset is highly suitable for testing the under-development procedures against the insider threat, remote unauthorised access, and physical access.

Web Bot Detection Using Mouse Movement

“…In this project the dataset created by Kiliç et al,.2021 called Bogazici mouse dynamics dataset is used. This dataset contains real user mouse movement when using a browser [16]. The preprocessing of the dataset is later explained.…”

“…• Legitimate mouse movement: This part of the dataset was created by using Bogazici mouse dynamics created by AA Kiliç et. al [16]. It is a public dataset that contains real free human mouse movement.…”

“…The legitimate sequences of the dataset are extracted from the public database Bogazici mouse dynamics [16]. It contains free mouse movements of 43 users recorded during 10 days.…”

“…The first dataset created is called Base dataset. The legitimate sequences are extracted from the Bogazici mouse dynamics dataset [16] as it have been explained in the Methodology. In total there are 2385 legitimate labeled sequences.…”