Smartphone Continuous Authentication via Deep Behavioral Features and Isolation Forest

Smartphone authentication typically relies on user input of passwords or biometric features such as facial recognition, fingerprint scanning, or iris recognition. However, these methods only authenticate the user at specific points in time, meaning that once the password is compromised or the device is unlocked, unauthorized individuals can access all information on the device or perform unauthorized actions. To resolve this issue, continuous authentication has emerged, continuously verifying the user's identity by covertly monitoring the ongoing interactions between the user and the device. This paper proposes a continuous authentication method that utilizes smartphone sensor data to extract users' deep behavioral features through Siamese Convolutional Neural Networks (Siamese CNN). These features are then input into the Isolation Forest (iForest) algorithm for classification. Experimental results demonstrate that the proposed method can authenticate the user within one second, achieving an average accuracy of 93.79% and an average Equal Error Rate (EER) of 8.37%, significantly outperforming other comparative methods used in this study.