基于用户行为特征的低能耗身份认证

Low Energy Identity Authentication Based on User Behavior Characteristics

目前,移动设备中传统的身份认证如密码、指纹等方式存在安全性低、易被破解等风险,不能完全保护使用者的隐私。本文提出基于用户行为特征的持续身份认证方案,通过调用移动终端传感器获得与用户行为相关的数据,采用最大互信息系数提取出能表征用户的特征,再利用机器学习算法进行模型训练,对用户身份进行识别。为了使持续的身份认证方式尽可能减少能耗,提出了低能耗的身份认证模型,对内置传感器的不同采样频率和识别算法进行能耗分析,实验结果表明,模型采用朴素贝叶斯算法,传感器采样率为25 Hz时,可以使认证精度达到97.94%,且显著降低认证模型的能耗。

At present, the traditional identity authentication in mobile devices, such as password and finger-print, has low security, easy to be cracked and has other risks, and cannot completely protect the user’s privacy. In this paper, a continuous identity authentication scheme based on user behavior characteristics is proposed. The data related to user behavior are obtained by calling mobile terminal sensors, and the features that can represent users are extracted by using maximum mutual information coefficient. Then the user identity is identified by machine learning algorithm for model training. In order to make the continuous authentication way as far as possible to reduce energy consumption, low energy consumption of the authentication model is put forward, with built-in sensors of different sampling frequency and identification algorithms for analysis of energy consumption, the experimental results show that the model uses naive Bayesian algorithm, the sensor when the sampling rate is 40 Hz, can make the authentication accuracy reach 97.94%, and significantly reduce the energy consumption of authentication model.