In this course, I studied techniques for encrypting and authenticating H.264/Scalable Video Coding (SVC) streams. I customised the H.264/SVC Codec Opensvc, JSVM and RTSP (Real Time Streaming Protocol) software Live555 and implemented them in the Android platform. Then I tested the performance of the EStream fast stream encryption algorithms (Salsa20/12, HC-128/256, Rabbit, SOSEMANUK) and compared them with RC4 and AES in order to determine the best encryption algorithms for SVC encryption in mobile devices. I also implemented two SVC stream authentication schemes and studied their performance through computer simulation, where I used the Gilbert Channel Model to simulate packet loss statistics in wireless networks.

There is a continuing tension between the need to develop and test new features, and the need to provide a stable and secure execution environment. A compromise, adopted by the developers of the Linux kernel, is to release new versions, including bug fixes and new features, frequently, while maintaining some older “longterm” version. This strategy raises the problem of how to identify bug fixing patches that are submitted to the current version but should be applied to the long-term versions as well. The current approach is to rely on the individual subsystem maintainers to forward patches that seem relevant to the maintainers of the longterm kernel. The reactivity and diligence of the maintainers, however, varies, and thus many important patches could be missed by this approach. In this project, we propose an approach that automatically identifies bug fixing patches based on the changes and commit messages recorded in code repositories. We compare our approach with the keyword-based approach for identifying bug fixing patches used in the literature. We have applied our approach to patches from Linux. The results show that our approach can achieve a much higher recall (45.11% relative improvement in recall) as compared to keyword-based approaches, with similar precision.

My work in this semester is conducted around the theme of Twitter. Firstly, I surveyed the Twitter studies mainly from the event and the individual perspective. Considering the large volume of literature, while I have collected more than one hundred papers and sorted them into different categories according to the problems they addressed and methods used, there are still more works need to be included. I'm also responsible for the building of the frontend for the demo system of the Twitter users' real-life friend identification algorithm, the website of which is http://twitterbud2011.appspot.com/. Then I helped in strengthening the identification algorithm by doing experiments as well as finding social science work supporting the principles the algorithm based upon. After submission of the paper illustrating this algorithm to www2012 conference, I'm currently engaged in the design of the demo paper which is also expected to be submitted to www2012 conference. This demo paper is based on the aforementioned website, on which we'll do some improvements to make it more interesting for users to play with and, at the meantime, a platform to exemplify the concepts proposed in the related paper.

Sequential pattern mining is an important data mining problem with broad application. Long patterns reflect important nature of the large sequence database and they are more informative in characterising large sequential data. However, with regard to long patterns in large datasets like genome-scale datasets, some elegant algorithms like PrefixSpan would still take a great amount of time. In our project, we try to develop a new algorithm that can efficiently mine the top k longest sequential patterns. First, by using suffix trees, we can derive all the frequent common substrings. Then several probing strategies, for example, random probing and evenly probing, are proposed to generate the initial candidate set. Pairwise comparison is iteratively performed to merge the segments in the candidate set. In each iteration, the segments that can be merged would be merged to a new pattern and put in the candidate set for the next iteration, while those that can’t be merged would be abandoned unless they meet up some threshold, like the length of the pattern or the number of the segments in the pattern. Moreover, several heuristic methods are used to speed up the mining process.

Project Description

How it Began

• Began as a LARC initiative
• Project involving Resorts World Singapore
• Cooperate with Universal Studios as a platform
• Site visits & observations
• Interviews with Stakeholders

Which led me in thinking of how we can…

• Formalise the problems identified that are inherent to Universal Studios.
• Develop a project that can expand the use computational social science.
• Find a novel solution to use behavioral information observed and collected in order to solve the problem.

Project Description

The purpose of this project is to do one or more of the following:

1. Collecting information on these data sources
2. Visualising this information
3. Finding nice patterns on the social phenomenon behind this “developers” social network
4. Developing an automated data mining approach to recommending people on developer's social network

There is a wealth of software data available from years of software development efforts. These data are publicly available via systems like Sourceforge, Bugzilla, etc. The data revolves around the various activities that developers typically do in a software development effort, such as adding new code, modifying existing code, participating in multiple projects, forming connections with peers in the same project, creating new projects, fixing bugs, handling feature requests, handling bug reports, etc.

Plug-ins are gaining popularity because of the convenience they offer, and the size of plug-ins increases over time as changes are made by developers. If we can find out the specifications that a plug-in follows or group similar plug-ins together automatically, this will help developers save a great amount of time and energy when he/she takes on a legacy system, especially systems developed without any formal documentation.

We propose an effective approach of clustering plug-ins into several groups according to their similarity and doing some specification mining to get some common functions about each of the clustered plug-ins. In this way, developers can pick out plug-ins more easily, because one just needs to read the document about every cluster to get the main point of these plug-ins.

Current GIS tools are no longer adequate to support an efficient analysis of space-time-attribute datasets. Hence we have developed an interactive analytical application that has adopted various geo-visualisation techniques, such as heat maps, coordinated linked view and brushing. At the end of the paper, we will use this application to investigate the time-series patterns of taxi trips at different locations.

This report is aimed at understanding the big picture of virtualisation on the x86 platform. Classification of virtualisation is presented first, and then the architecture, then more details of the implementations are summarised. For a long time, virtualisation had remained a privilege for mainframes. Only recently, when x86 based microcomputers gradually took over the world of business computing, did people realise the need for virtualisation on x86 based machines. However, the x86 micro architecture had never been designed with support for virtualisation in mind. Virtualising an x86 box had proven to be intrinsically difficult, but not impossible. In the effort to bring virtualisation to x86 platforms, various techniques have been discovered and developed. Each approach exhibits varying aspects of advantages and disadvantages.