User Identification with Deep Learning on MEU-Mobile KSD

Karadeli M. I., KAHRAMAN N.

2025 Innovations in Intelligent Systems and Applications Conference, ASYU 2025, Bursa, Türkiye, 10 - 12 Eylül 2025, (Tam Metin Bildiri) identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Doi Numarası: 10.1109/asyu67174.2025.11208354
  • Basıldığı Şehir: Bursa
  • Basıldığı Ülke: Türkiye
  • Anahtar Kelimeler: Deep Learning, Keystroke Dynamics, MEU-Mobile KSD Dataset, Mobile Authentication, User Identification
  • Yıldız Teknik Üniversitesi Adresli: Evet

Özet

Traditional security mechanisms on mobile devices - such as passwords, facial recognition, or fingerprint scanning - rely on one-time authentication and are vulnerable to spoofing, imitation, or theft. In contrast, keystroke dynamics provide a behavioral biometric modality that is significantly harder to replicate or steal, enabling continuous and implicit user authentication throughout device usage. A user identification system leveraging keystroke dynamics was developed using the MEU-Mobile KSD 2016 dataset, which contains biometric typing behavior data from 56 individuals. To determine the most effective classification strategy, various machine learning and deep learning models were evaluated following feature extraction via deep neural network models. Among the evaluated models, a Random Forest (RF) classifier achieved the highest accuracy of 89.86% on the test set with Convolutional Neural Network (CNN) feature extraction technique. These findings highlight the potential of combining feature extraction and machine learning for robust behavioral biometric-based user recognition.