The article presents a study on keystroke authentication using Transformer-based networks. Two architectures, bi-encoder and cross-encoder, are compared for their effectiveness in keystroke authentication. Various loss functions and distance metrics are also explored to optimize the training process and improve the model’s performance. The model is evaluated using the Aalto desktop keystroke dataset. The bi-encoder architecture with batch-all triplet loss and cosine distance yielded the best performance. The study also explores alternative algorithms for calculating similarity scores to enhance accuracy. The findings indicate that this model surpasses the previous state-of-the-art in free-text keystroke authentication.
Publication date: 19 Oct 2023
Project Page: Not provided
Paper: https://arxiv.org/pdf/2310.11640
