面向用户可信的移动云用户行为模式挖掘与在线识别策略研究

发布时间：2018-03-24 06:26

本文选题：移动云计算　切入点：用户可信　出处：《河南科技大学》2017年硕士论文

【摘要】：移动云计算是随着云计算和移动互联网的不断发展与融合而产生的一种新型应用模式。在可信移动云服务架构中,用户身份及行为的可信性判定与识别将成为后继服务的重要前提,只有在保证用户身份和用户操作可靠与安全的前提下,才可以进行后续的智能决策和服务资源调配,保障向用户提供满足需求的连续服务能力。可见,用户行为判别对移动云服务十分重要。本文中提到的用户可信是指用户对服务实施的操作和动作总是处于用户所属规则允许的范围内,对于用户的异常行为可识别、可控制。模式挖掘这个工具,使我们可以从大量的正常移动用户读写移动云服务的行为数据中挖掘出其行为模式特点,构造正常行为的参考集合。之后实时地将用户行为与正常行为进行对比,来检测用户是否异常,达到对用户行为的监督与识别,从而保障用户层数据读写环境和服务交付来往的安全性。本文围绕用户可信问题,着重开展用户行为模式挖掘与在线识别的研究。1.提出了用户时序行为序列的定义,作为进行用户异常行为识别的基础。首先,特化用户时序行为的用户-时序-操作形式化描述,使用“用户-时序-操作”时序图来表示用户一系列操作的时序集合。并给出用户时序行为序列的编码结构图,介绍了编码序列中各部分的含义和用途。2.提出了自适应的用户正常行为模式编码方法。在用户正常行为模式挖掘过程中,依据挖掘的时序行为模式特点,借鉴遗传算法的相关技术理论,使用自适应编码构建完整的用户正常行为模式集合,以提高后续用户时序行为实时判别的效率。通过仿真实验证明了算法在动态性能和运行效率方面的优越性。3.在得到完整用户正常行为模式集合的基础上,设计了基于NeedlemanWunsch算法的一种用户时序操作行为的实时判别方法。将用户实时行为转化为形式化编码,之后与正常行为模式集合进行比对,通过提前设定的阈值来确定用户行为是否异常,并讨论了异常检测过程中的一些细节。本文在理论研究成果的基础上进行了仿真验证,结果表明本文提出的检测算法相对传统算法在检测率和误报率方面表现优秀。本文提出的模式挖掘算法和在线异常检测算法的创新点在于:根据本文提出的行为序列编码方法对序列进行编码之后,它形成的二进制序列与DNA序列有非常高的相似性,所以本文创新性的引入了生物学中的遗传算法和DNA序列比对算法来进行模式挖掘和异常检测。同时,由于移动云用户终端存在系统复杂、接入环境多变等特点,移动云计算服务器相对于传统云计算服务器来说面临着种类更加多样、变化更加频繁的异常攻击。这就导致了如果使用一成不变的模式挖掘算法和参数模型,会使异常检测的准确率和误报率产生较大的浮动和偏差。为此,本文设计了自适应的编码方式和参数确定算法,可以使系统根据环境变化实时调整自身的挖掘和检测策略,提高运行效率和检测性能。
[Abstract]:Mobile cloud computing is a new application mode which is generated with the development and integration of cloud computing and mobile Internet. In the trusted mobile cloud service architecture, the credibility of user identity and behavior decision and recognition will become an important prerequisite for subsequent service, only to ensure user identity and user operation reliability and safety of the premise can, intelligent decision-making and follow-up service resource allocation, guarantee to provide users with continuous service capabilities to meet the needs of the user behavior. Therefore, discrimination is very important for mobile cloud services. Users can be mentioned in this letter refers to the user of the service operation and action is always the user belongs to rule allowable range. The abnormal behavior of the user can be identified, controlled pattern mining. This tool, we can make a large number of mobile users from the normal reading behavior of mobile cloud services according to write In digging out the behavior characteristics, structure of normal behavior. After the reference set in real time by comparing the user behavior and normal behavior, to detect whether the user is abnormal, to supervision and recognition of user behavior, so as to protect the user layer data read-write environment and service delivery. This paper focuses on the safety and credibility problem of users the definition of user behavior, timing sequence focuses on the research of.1. user behavior pattern mining and on-line identification, as the basis for the abnormal behavior of users recognition. Firstly, the temporal behavior of users - specialized users when the order operation of formal description, the use of "user sequential operation" sequence diagram to represent a series of sequential users the set of operations. Encoding structure diagram and gives the user the temporal behavior of the sequence, introduces the meaning and purpose of each part of the.2. encoding sequence in the proposed adaptive user normal line For pattern encoding method. In the normal user behavior pattern mining process, according to the characteristics of the temporal behavior pattern mining technology, from the relevant theory of genetic algorithm, the set mode of normal behaviors using adaptive encoding to build a complete, in order to improve the efficiency of subsequent user behavior in real time timing discrimination. The experimental results demonstrate the algorithm in dynamic performance and the operating efficiency of the advantages of.3. in the base of the complete set user normal behavior model, design a real-time identifying method for user behavior of sequential operation based on NeedlemanWunsch algorithm. The real-time user behavior into formal encoding, then compared with the normal behavior model set by set the threshold to determine the user behavior if abnormal, and discuss some details of the anomaly detection process. This paper basis on the theoretical research results for The simulation results show that the proposed detection algorithm than the traditional algorithm in the detection rate and false alarm rate of outstanding performance. This paper proposes the pattern mining algorithm and online anomaly detection algorithm innovation point is: according to the behavior sequence encoding method proposed in this paper for the encoding sequence has a high similarity with DNA binary sequence it forms the sequence, so this paper innovatively introduces the genetic algorithm and DNA algorithm in biology sequence alignment of pattern mining and anomaly detection. At the same time, because the mobile cloud users existing complex system, access environment and other characteristics, compared with the traditional mobile cloud computing server cloud computing server is faced with more variety. Abnormal attack change more frequently. This leads to if the algorithm and parameters of the model using the immutable and frozen pattern mining, will make the exception The accuracy and false positive rate of detection have great fluctuations and deviations. Therefore, this paper designs an adaptive coding method and parameter determination algorithm, which enables the system to adjust its mining and detection strategies in real time according to the environmental changes, so as to improve the operation efficiency and detection performance.

【学位授予单位】：河南科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN929.5;TP309

【相似文献】