Research On Text Analysis Of Smartphone Electronic Forensics

With the rapidly growing popularity of mobile phones and its memory increases, the challenges in processing the large amounts of SMS in the smartphone also grow. How to extract evidence from the phone effectively becomes a problem of the smartphone forensics technology, and short text automatic classification technology can partly solve the problem. Although the traditional text categorization technology is relatively mature, the short text format is not standard and sparse text information which leds to classification method suitable for long text cannot be applied to short text.Based on such background, the information classification method of SMS message in the cell phone is explored and compared. At first, this paper outlines the three factors of mobile phone forensics:source, principle and process, and analysis the basic process of extracting the electronic evidence by the example of SMS information in mobile phone. Then the paper focus on discussing the SMS messages automatic classification subsystem in mobile phone forensics system. My paper discusses the basic difference between long and short text, analysis of several commonly used in traditional classification method of feature weighting algorithm, and put forward an improved weight value algorithm for short text Then, in order to offset the disadvantages of the short text information scarce, this paper proposes the introduction of Wikipedia database as a repository for the short text feature extensions. At last, the thesis introduces the design and implementation process of each module of short text classification system, and compares the classification results of two improved methods mentioned in this paper with six groups of experiment.Six groups of the experimental results show that the two improved methods based on traditional classification methods all can improve the Chinese short text classification. Meanwhile, I found that if the noise in knowledge base is not eliminated clean, which will affect the results of the classification method based on feature extension.