LFCA: A Lightweight Framework of Continuous Authentication with High Accuracy

With the widespread adoption of smart mobile devices, security concerns have become increasingly prominent. Traditional entry-point authentication methods struggle to cope with the complexities of modern network environments, heightening the risk of sensitive information leaks. As an effective implementation of Zero Trust security architecture, continuous authentication has gained significant attention. However, existing methods either struggle with high algorithmic complexity on mobile devices or lack consideration for the diversity of user characteristics, behavior patterns, and application scenarios in terms of dataset design. To address these problems, this study presents a large-scale open-source dataset featuring rich user characteristics, complex behavior patterns, and diverse application scenarios, and proposes a accurate and lightweight Mamba-based continuous authentication framework. Experimental results shows the high accuracy of the framework. Code and dataset are available at https://github.com/qinyawen/Dataset.