| Compare to Linux, Windows has "close-source" nature and is more expansive. Many users have shown interest in migrating to Linux if their favorite applications could be well supported. Based on this, the project UnifiedKernel is set up by Insigma. As the project name UnifiedKernel, the core part of it is a kernel which implements full functions of Windows kernel in a Linux kernel. By emulation of the Windows kernel, Windows applications and drivers could be run on the Linux Operating System smoothly and efficiently. One of main works of the project is to compare the system function call sequences between the UnifiedKernel system and native Windows OS. By discovery pattern sequences of functions and recognize them, the differences and flaws of the UnifiedKernel system could be found. This idea could also be applied to other research fields, such as program understanding and source code analysis.This dissertation researches the pattern discovery of sequential patterns mining in database, and proposes a new algorithm of pattern discovery. To recognize similar function call sequence, a pattern recognition algorithm based on HMM is proposed. The main contributes are as follows:(1) Presents a survery of model of sequential pattern and classical mining algorithms. The applicable future and possible challenge of sequential pattern mining are also presented.(2) Proposes a new algorithm of sequential pattern discovery called BB-PrefixSpan. The struct of brief projection collections and bi-level projection are used in the BB-PrefixSpan algorithm. Results of our experiments show that BB-PrefixSpan has better time performance than traditional sequential pattern mining algorithms such as PrefixSpan.(3) Proposes a sequential pattern recognition algorithm based on the Hidden Markov Model. The HMM has been used successfully in speech recognition, and we introduced it in pattern recognition in function call sequences. The experiments showit can recognize patterns efficiently. |
PDF Full Text Request