Master's Projects

Lost Decades of Japan

Nishant Makhija — Fri, 07 Jun 2013 09:03:46 PDT

YIOOP FULL HISTORICAL INDEXING IN CACHE NAVIGATION

Akshat Kukreti — Fri, 24 May 2013 21:06:35 PDT

This project adds new cache-related features to Yioop, an Open Source, PHP-based search engine. Search engines often maintain caches of pages downloaded by their crawler. Commercial search engines like Google display a link to the cached version of a web page along with the search results for that particular web page. The first feature enables users to navigate through Yioop's entire cache. When a cached page is displayed along with its contents, links to cached pages saved in the past are also displayed. The feature also enables users to navigate cache history based on year and month. This feature is similar in function to the Internet Archive as it maintains snapshots of the web taken at different times. The contents of a web page can change over time. Thus, a search engine caching web pages has to make sure that the cached pages are fresh. The second feature of this project implements cache validation using information obtained from web page headers. The cache validation mechanism is implemented using Entity Tags and Expires header. The cache validation feature is then tested for effect on crawl speed and savings in bandwidth.

Santa Clara County’s Juvenile Hall Transgender Policy: A First Look

Etelvina Alvarez — Fri, 24 May 2013 15:31:38 PDT

MongoDB Performance In The Cloud

Tudor Matei — Thu, 23 May 2013 07:43:58 PDT

Web applications are growing at a staggering rate every day. As web applications keep getting more complex, their data storage requirements tend to grow exponentially. Databases play an important role in the way web applications store their information. Mongodb is a document store database that does not have strict schemas that RDBMs require and can grow horizontally without performance degradation. MongoDB brings possibilities for different storage scenarios and allow the programmers to use the database as a storage that fits their needs, not the other way around. Scaling MongoDB horizontally requires tens to hundreds of servers, making it very difficult to afford this kind of setup on dedicated hardware. By moving the database into the cloud, this opens up a possibility for low cost virtual machine instances at reasonable prices. There are many cloud services to choose from and without testing performance on each one, there is very little information out there. This paper provides benchmarks on the performance of MongoDB in the cloud.

Automated RTL generator

Rohit Kulkarni — Thu, 23 May 2013 07:43:56 PDT

Code generation is a vast topic and has been discussed and implemented for quite a while now. It has been also been a topic of debate as to what is an ideal code generator and how an ideal code generator can be created. The biggest challenge while creating a code generator is to maintain a balance between the amount of freedom given to the user and the restrictions imposed on the code generated. These two seemed to be very conflicting requirements while designing the Automated RTL Code Generator. If the code generator tries to be rigid and sticks to well-defined paths and restricted code, the flexibility provided to the also reduces. It is a very interesting task to strike the right amount of balance and generate code of high quality and well-defined standards. Verilog code is a type of RTL (Register Transfer Level) that itself has fewer constructs and variety as compared to pure software languages like Java, or Python so it makes sense to generate it automatically so that the hardware designers are relieved from the mundane tasks of writing repetitive verilog code modules. Also code generator provides a nice introduction to the much wider topic of compiler design. This project also tries to delve deeper into the latest IPXACT XML standard IEEE 1865-2009 which is used for hardware description and will provide means of generating verilog code directly from it.

Analysis of Parallel Montgomery Multiplication in CUDA

Yuheng Liu — Thu, 23 May 2013 07:43:54 PDT

For a given level of security, elliptic curve cryptography (ECC) offers improved efficiency over classic public key implementations. Point multiplication is the most common operation in ECC and, consequently, any significant improvement in perfor- mance will likely require accelerating point multiplication. In ECC, the Montgomery algorithm is widely used for point multiplication. The primary purpose of this project is to implement and analyze a parallel implementation of the Montgomery algorithm as it is used in ECC. Specifically, the performance of CPU-based Montgomery multiplication and a GPU-based implementation in CUDA are compared.

User Profiling in GUI based Windows Systems for Intrusion Detection

Arshi Agrawal — Thu, 23 May 2013 07:43:52 PDT

Intrusion detection is the process of identifying any unauthorized access to a sys- tem. This process inspects user behavior to identify any possible attack or intrusion. There exists two type of intrusion detection systems (IDSs): signature-based IDS and anomaly-based IDS. This project concentrates on anomaly-based intrusion detection technique. This technique is based on the deviation of intruder’s actions from the authenticated user’s actions. Much previous research has focused on the deviation of command line input in UNIX systems. However, these techniques fail to detect attacks on modern GUI- based systems, where typical user activities include mouse movements and keystrokes. Our project aims to create a dataset suitable for testing intrusion detection strate- gies on GUI-based operating systems. We have developed an event logging tool to capture GUI-based user data on Windows systems. We have collected a large dataset which we analyze using a intrusion detection strategy based on hidden Markov models (HMM).

Analysis of Kullback-Leibler Divergence for Masquerade Detection

Geetha Ranjini Viswanathan — Thu, 23 May 2013 07:43:49 PDT

A masquerader is an attacker who gains access to a legitimate user’s credentials and pretends to be that user so as to avoid detection. Several statistical techniques have been applied to the masquerade detection problem, including hidden Markov models (HMM) and one class na ̈ Bayes (OCNB). In addition, Kullback-Leibler ıve (KL) divergence has been used in an effort to improve detection rates. In this project, we develop and analyze masquerade detection techniques that employ KL divergence, HMMs, and ONCB. Detailed statistical analysis is provided to show that our results outperform previous related research.

Metamorphic Code Generation from LLVM IR Bytecode

Teja Tamboli — Thu, 23 May 2013 07:43:47 PDT

Metamorphic software changes its internal structure across generations with its functionality remaining unchanged. Metamorphism has been employed by malware writers as a means of evading signature detection and other advanced detection strate- gies. However, code morphing also has potential security benefits, since it increases the “genetic diversity” of software. In this research, we have created a metamorphic code generator within the LLVM compiler framework. LLVM is a three-phase compiler that supports multiple source languages and target architectures. It uses a common intermediate representation (IR) bytecode in its optimizer. Consequently, any supported high-level programming language can be transformed to this IR bytecode as part of the LLVM compila- tion process. Our metamorphic generator functions at the IR bytecode level, which provides many advantages over previously developed metamorphic generators. The morphing techniques that we employ include dead code insertion—where the dead code is actually executed within the morphed code—and subroutine permutation. We have tested the effectiveness of our code morphing using hidden Markov model analysis.

Facebook Headquarters: Thumbs Up or Thumbs Down for Housing Prices in East Palo Alto?

Veronica A. Flores — Wed, 22 May 2013 09:09:10 PDT

HTTP Attack Detection using N-gram Analysis

Adityaram Oza — Tue, 21 May 2013 10:09:17 PDT

Previous research has shown that byte level analysis of HTTP traffic offers a practical solution to the problem of network intrusion detection and traffic analysis. Such an approach does not require any knowledge of applications running on web servers or any pre-processing of incoming data. In this project, we apply three n- gram based techniques to the problem of HTTP attack detection. The goal of such techniques is to provide a first line of defense by filtering out the vast majority of benign HTTP traffic. We analyze our techniques in terms of accuracy of attack detection and performance. We show that our techniques provide more accurate detecting and are more efficient in comparison to a previously analyzed HMM-based technique.

Cloud Based Complaint Management Service

Ajinkya Amrute — Tue, 21 May 2013 09:55:31 PDT

Complaint Management is important from both customer as well as business point of view. Complaints contain direct voice of the customer which provides companies a huge volume of data which can be used to improve the quality of the product the company is manufacturing. Hence it is necessary for the organizations to harness the data received via complaints. However as the data received via complaints is enormous it is not an easy task to manage the data received via complaints as the data keeps on expanding and multiplying. For implementing this difficult task of complaint management, websites were built but it becomes necessary for the customer to have a desktop/ laptop for using the system built for complaint management. However with time, mobiles have taken place of the Laptops / Desktops due to their on the fly nature which helps an individual to communicate or work on his personal tasks while he/ she is travelling. Hence in my project I have provided mobile support to these kinds of systems which has to deal with huge volumes of data as it will be beneficial. However mobile devices on their own do not possess sufficient level of resources for dealing with such level of computations that are done on the data. Hence I have cloud computing utilized for this kind of systems which will take care of the heavy computational stuff leaving mobile phone to efficiently function while people are using such a system. Hence my project aims to develop a mobile based complaint management system which will be provided as a service to the users of the system. I have also added the feature of the multi-tenancy to the system using which the system can be configured according to the requirements of an organization.

CLASSIFICATION OF WEB PAGES IN YIOOP WITH ACTIVE LEARNING

Shawn C. Tice — Tue, 21 May 2013 09:55:29 PDT

This thesis project augments the Yioop search engine with a general facility for automatically assigning meta words (e.g. advertising) to web pages based on the output of a logistic regression text classifier. Users can create multiple classifers using Yioop's web-based interface, each trained first on a small set of labeled documents drawn from previous crawls, then improved over repeated rounds of active learning.

Cloud Services for an Android Based Home Security System

Karthik Challa — Tue, 21 May 2013 09:55:27 PDT

This report talks in detail about an android based application designed for a home security system. The home security system is a tablet device developed using the android framework. The home security system makes use of sensors and a central device to secure an area. Currently the devices are standalone and require the users to be physically present to operate the devices with no interaction possible between two different devices. The system is also limited by its computational resources and storage capacity. For this project, I have developed a cloud based client server architecture to address these limitations and also to provide security and sharing functionalities along with remote diagnostics. This project has a device level framework to communicate with and exchange information with a cloud server. The project also addresses the primary limitations of cloud computing namely security, privacy and user control.

Entropy and State Visualization for Automation Design and Evaluation Prototyping Toolset

Rohit Deshmukh — Tue, 21 May 2013 09:55:24 PDT

Automation Design and Evaluation Prototyping Toolset (ADEPT) is a plugin developed on the Eclipse Rich Client Platform(RCP). ADEPT can be used by domain expert designers to create and modify testable prototypes. The aim of the project is to enhance ADEPT by adding dynamic visualizations to the ADEPT user interface. Three types of visualizations are implemented in this project. Table view is helpful to view the hierarchy and nesting of Logic Tables. The State visualization displays all the states in a selected Logic Table. Entropy visualization is a subset of State visualization and displays limited number of states having lowest Entropy value.

Recommendation System for News Reader

Shweta Athalye — Tue, 21 May 2013 09:55:21 PDT

Recommendation Systems help users to find information and make decisions where they lack the required knowledge to judge a particular product. Also, the information dataset available can be huge and recommendation systems help in filtering this data according to users‟ needs. Recommendation systems can be used in various different ways to facilitate its users with effective information sorting. For a person who loves reading, this paper presents the research and implementation of a Recommendation System for a NewsReader Application using Android Platform. The NewsReader Application proactively recommends news articles as per the reading habits of the user, recorded over a period of time and also recommends the currently trending articles. Recommendation systems and their implementations using various algorithms is the primary area of study for this project. This research paper compares and details popular recommendation algorithms viz. Content based recommendation systems, Collaborative recommendation systems etc. Moreover, it also presents a more efficient Hybrid approach that absorbs the best aspects from both the algorithms mentioned above, while trying to eliminate all the potential drawbacks observed.

Brownfield Cleanup in a Post-Redevelopment Era: Using a Collaborative, Community-Based Approach to Green the Spartan-Keyes Neighborhood

Sara Lynn Sichley — Mon, 13 May 2013 08:47:26 PDT

Intelligent Personalized Searching

Wing Lau — Thu, 09 May 2013 19:52:22 PDT

Search engine is a very useful tool for almost everyone nowadays. People use search engine for the purpose of searching about their personal finance, restaurants, electronic products, and travel information, to name a few. As helpful as search engines are in terms of providing information, they can also manipulate people behaviors because most people trust online information without a doubt. Furthermore, ordinary users usually only pay attention the highest-ranking pages from the search results. Knowing this predictable user behavior, search engine providers such as Google and Yahoo take advantage and use it as a tool for them to generate profit. Search engine providers are enterprise companies with the goal to generate profit, and an easy way for them to do so is by ranking up particular web pages to promote the product or services of their own or their paid customers. The results from search engine could be misleading. The goal of this project is to filter the bias from search results and provide best matches on behalf of users’ interest.

There’s No Place like Home? Telecommuting as a Viable Commuting Alternative for San Jose Government and Companies to Reduce Vehicle Miles Traveled and Greenhouse Gas Emissions

Jennifer Marie Piozet — Tue, 30 Apr 2013 07:56:11 PDT

A Comparison of Basic At-Large and At-Large By-Seat Elections Systems in the State of California

Rod Diridon Jr. — Tue, 09 Apr 2013 09:26:36 PDT