HPCANet:一种用于身份认证的异构并行神经网络

HPCANet: a heterogeneous parallel neural network for identity authentication

摘要

网络安全事件频发,使得以键盘作为输入设备的击键动力学身份认证方式成为目前信息安全领域的研究热点。已有的研究在不同尺度下击键特征提取以及处理大量数据时仍然面临较大挑战。本文设计了基于击键身份认证的数据特征提取算法,提出基于击键动力学的HPCANet身份认证模型,该模型采用双轨并行结构,分别提取击键数据的局部特征和长期依赖关系,并通过骨干网络实现特征融合,引入了通道注意力机制重新为数据特征分配权重,实现突出用户关键行为特征的效果。本文在UB和CMU公开数据集上进行实验,分别获得2.74%和3.50%的等错误率。相较于现有代表模型,性能有显著提升。此外,实验还验证了模型在用户使用不同类型的输入设备时仍然具有较高的可靠性。

Abstract

In an era of escalating cybersecurity threats, keystroke dynamics has become a pivotal technology for identity authentication. Despite its potential, existing research struggles with the challenges of extracting features across different scales and managing massive datasets. This paper introduces a comprehensive solution, beginning with a specialized data feature extraction algorithm for keystroke authentication. We then present HPCANet, a novel model featuring a dual-track parallel architecture designed to independently capture both local features and long-term temporal dependencies in typing behavior. These distinct features are integrated through a robust backbone network. A key innovation is the integration of a channel attention mechanism, which dynamically reassigns feature weights to accentuate the most discriminative user behaviors. Evaluated on the public UB and CMU datasets, HPCANet achieves impressive Equal Error Rates (EER) of 2.74% and 3.50%, marking a substantial improvement over current state-of-the-art models. Our findings also demonstrate the model′s robust reliability when users switch between various input devices.

关键词

击键动力学 / 身份认证 / 神经网络 / 通道注意力机制

Key words

keystroke dynamics / identity authentication / neural network / channel attention mechanism

[1] KIM J, KANG P. Freely typed keystroke dynamics-based user authentication for mobile devices based on heterogeneous features[J]. Pattern Recognit, 2020, 108: 107556.
[2] KIM J, PARK S, KANG P. Device-invariant keystroke dynamics authentication via distribution alignment[J]. IEEE Transactions on Information Forensics and Security, 2024, 19: 2200-2213.
[3] YANG H, MENG X, ZHAO X, et al. CKDAN: Content and keystroke dual attention networks for continuous authentication[J]. Computers & Security, 2023, 128: 103159.
[4] YOU Y, ZHAO X, CHEN L, et al. Graph-based keystroke representation learning for robust continuous authentication[J]. Pattern Recognit, 2024, 158: 110859.
[5] MONACO J V. Robust keystroke biometric anomaly detection[EB/OL]. arXiv, 2016[2025-10-17]. https://arXiv.1606.09075.
[6] LIAKAT A M, TAPPERT C C. POHMM/SVM: A hybrid approach for keystroke biometric user authentication[C]//2018 IEEE International Conference on Real-time Computing and Robotics (PCAR). Kandima: IEEE, 2018: 612-617.
[7] IVANNIKOVA E, DAVID G, HÄMÄLÄINEN T. Anomaly detection approach to keystroke dynamics based user authentication[C]//2017 IEEE Symposium on Computers and Communications (ISCC). Heraklion: IEEE, 2017: 885-889.
[8] ROY S, ROY U, SINHA D D. Performance perspective of different classifiers on different keystroke datasets[J]. International Journal of New Technologies in Science and Engineering (IJNTSE), 2015, 2(4): 64-73.
[9] SUN Y, CEKER H, UPADHYAYA S. Shared keystroke dataset for continuous authentication[C]//2016 IEEE international workshop on information forensics and security (WIFS). Abu Dhabi:IEEE, 2016: 1-6.
[10] BERNARDI M L, CIMITILE M, MARTINELLI F, et al. Keystroke analysis for user identification using deep neural networks[C]//2019 International Joint Conference on Neural Networks (IJCNN). Budapest:IEEE, 2019: 1-8.
[11] MAO R, WANG X Y, JI H M. ACBM: Attention-based CNN and Bi-LSTM model for continuous identity authentication[C]//Journal of Physics: Conference Series. IOP Publishing, 2022, 2352(1): 012005.
[12] LI X, ZHOU L, WANG S. BehaviorPrompt: Leveraging large language models for behavior-based user authentication[J]. Neurocomputing, 2024, 584: 127739.
[13] AYOTTE B, BANAVAR M, HOU D Q, et al. Fast free-text authentication via instance-based keystroke dynamics[J]. IEEE Transactions on Biometrics, Behavior, and Identity Science, 2020, 2(4): 377-387.
[14] CHANG H C, LI J W, STAMP M. Machine learning-based analysis of free-text keystroke dynamics[M]//Artificial Intelligence for Cybersecurity. Cham: Springer International Publishing, 2022: 331-356.
[15] ACIEN A, MORALES A, MONACO J V, et al. TypeNet: Deep learning keystroke biometrics[J]. IEEE Transactions on Biometrics, Behavior, and Identity Science, 2022, 4(1): 57-70.
[16] MHENNI A, CHERRIER E, ROSENBERGER C, et al. Analysis of Doddington zoo classification for user dependent template update: Application to keystroke dynamics recognition[J]. Future Generation Computer Systems, 2019, 97: 210-218.
[17] LIU Y, WANG J, CHEN S. Graph-Key: A graph neural network based model for keystroke dynamics authentication on free-text[J]. Computers & Security, 2023, 130: 103254.
[18] LU Y, SITOV A, BAJWA A. DeepKey: A new deep learning-based authentication system using keystroke dynamics[J]. Journal of Information Security and Applications, 2021, 62: 102957.

基金资助

新疆生产建设兵团重点领域科技攻关计划项目(2021AB023-4)