基于手机使用行为的身份认证方法研究

发布时间：2018-03-07 11:27

本文选题：身份认证　切入点：隐私保护　出处：《西北大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着手机智能化的普及,手机隐私保护问题变得日益关键。手机的隐私保护问题主要通过身份认证技术实现。现有的身份认证方案通过设置锁屏密码或利用用户的解锁手势特征来实现。然而,密码易通过被肩窥方式窃取或者被暴力破解,真实的机主用户和窃取密码的非法入侵用户无法被有效区分;此外,在机主解锁后,机主与借用者都可能是手机使用者,即解锁者与手机使用者并非一定相同,然而现有的基于手势的身份认证方案无法区分借用者用户和机主用户;同时,现有的方案中,均只实现了对机主与非机主两类身份的认证,而基于这种认证结果的隐私保护方案无法阻止借用者查看机主的隐私,不能满足用户对于隐私保护的需求。本文以隐私保护为背景,提出一种基于手机使用行为的身份认证方法,实现了对三类用户(机主、借用者、入侵者)的身份认证,针对以下三方面的挑战进行了探索并给出有效的解决方案:1)不同用户使用手机行为的相似度较高导致用户区别困难。2)用户的手机使用行为持续时间短,可供参考的信息量少。3)手机使用场景较多,且随机性大,如何设计高鲁棒性的身份认证方法是本文面临的另一挑战。不同于现有的基于行为特征的身份认证方案中只利用解锁单一动作的行为特征,本文的思路是同时考虑不同用户场景下动作序列和解锁子动作两方面的行为差异实现对三类用户的身份认证。本文将用户的手机使用行为分解成若干个独立存在的子动作,并用子动作组成的动作序列进行不同用户的手机使用习惯表达,利用马尔科夫链思想确定动作序列的发生概率,并以此为基础计算当前手机使用场景中的动作序列与机主用户的符合程度;通过有效的特征提取,提出基于逻辑回归的解锁行为差异性估计方法,确定当前解锁行为与机主的相似度;最终根据动作序列和解锁动作与机主的行为差异,以及动作序列的上下文信息提出身份认证算法,实现对三类用户的身份认证。本文对方法的有效性和准确性用真实实验进行了验证,实验证明本文方法准确率达到了 94%以上,且对各种手机使用场景均保持很高的鲁棒性,向手机的隐私保护迈进了关键的一步。本文提出的身份认证模型可扩展到其他基于动作差异的安全领域,如手机防盗等。
[Abstract]:With the popularity of intelligent mobile phones, The privacy protection of mobile phone is becoming more and more critical. The privacy protection of mobile phone is mainly realized through identity authentication technology. The existing identity authentication schemes are realized by setting the password of the screen lock or by using the features of the user's unlocked gesture. However, Passwords can easily be stolen by peeping or brutally cracked, and the actual owner and the illegal intruder who steal the password cannot be effectively distinguished; in addition, after the owner unlocks, both the owner and the borrower may be mobile phone users. That is, the unlocked user and the mobile phone user are not necessarily the same, however, the existing gesture-based authentication schemes can not distinguish between the borrower and the user. At the same time, in the existing scheme, But the privacy protection scheme based on this authentication result can not prevent the borrower from checking the privacy of the owner. This paper puts forward an identity authentication method based on mobile phone usage behavior, which realizes the identity authentication of three kinds of users (machine owner, borrower, intruder). This paper explores the following three challenges and gives an effective solution: 1) the high similarity of different users' mobile phone use behavior leads to the difficulty of user differentiation. 2) the user's mobile phone use behavior duration is short. The amount of information available for reference is small. 3) the mobile phone uses more scenarios and has a large randomness. How to design a robust authentication method is another challenge in this paper. It is different from the existing behavior based identity authentication scheme which only uses the behavior feature of unlocking a single action. The idea of this paper is to realize the identity authentication of three kinds of users by considering the behavior differences of action sequence and unlocked sub-action in different user scenarios at the same time. In this paper, the user's mobile phone usage behavior is decomposed into several independent sub-actions. The mobile phone usage habits of different users are expressed by the action sequence composed of sub-actions, and the probability of occurrence of the action sequence is determined by Markov chain thought. On the basis of this, we calculate the consistency between the action sequence in the current mobile phone usage scene and the main user. Through effective feature extraction, we propose a method to estimate the difference of unlock behavior based on logical regression. Finally, the identity authentication algorithm is proposed according to the behavior difference between the action sequence and the unlock action and the context information of the action sequence. In this paper, the validity and accuracy of the method are verified by real experiments. The experimental results show that the accuracy of this method is more than 94%, and the method is robust to all kinds of mobile phone usage scenarios. The authentication model proposed in this paper can be extended to other security fields based on action difference, such as mobile phone security.
【学位授予单位】：西北大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP309

【相似文献】