Konversatorium on Friday, April 27, 2018 - 10:30

Analyzing large amounts of data is becoming an ever increasing problem. Bitcoin as an example has produced more data than is possible to analyze. In order to compensate for these difficulties, creative ideas that employ data aggregation or minimization have been proposed. Other work also focuses on introducing novel visualization types that are geared towards visualization of blockchain data. However, visualization of graphs through node-link diagrams remains a difficult challenge. Analysis of the Bitcoin transaction graph to follow bitcoin transactions poses a difficult problem due to the Bitcoin protocol and the amount of data. This thesis combines two data processing strategies to visualize big network data on commodity hardware. The idea is to use visualization as a technique to analyze a data-set containing Bitcoin transaction information. Criminals use Bitcoin as a means of payment because of this guaranteed pseudonymity. However through visualization we aim to identify patterns that will allow us deanonymize transactions. To do so we use a proxy server that does data preprocessing before it is visualized on a web client. The proxy leverages parallel computing to be able to do top-down and bottom-up data processing fast enough for interactive visualization. This is done through incremental loading (bottom-up) which enables to visualize data immediately without a (pre-)processing delay. The database containing the public Bitcoin ledger is over 163 gigabytes in size. The resulting graph has more than 800 million nodes. As this information is too much to be visualized we also employ a top-down approach of data aggregation and graph minimization of the transactional graph. Through this methodology we intend to solve performance problems of long processing delays and the problem of fractured data where it is shown only partially in the visualization. We collaborate with security experts who share insights of their expertise through a continuously ongoing dialog. Exploratory analysis on a big data-set such as the Bitcoin ledger, enabled through the methodology presented in this thesis, will help security experts analyze the money flow in a financial network that is used by criminals for its anonymity. We evaluate the result through the performance and feedback of these security experts as well as benchmark the performance against current best practice approaches.
 

Slots for Talks still available! Please contact the KV administration.
Date: 
Friday, April 27, 2018 - 10:30
Location: 
Seminar room 193-2 (Favoritenstraße 9, Stiege 1, 5th floor)

Bitstream / Coinstream: A bottom-up/top-down approach to data loading for interactive bitcoin visualizations. (DAEV )

Speaker: 
Matthias Gusenbauer (Inst. 193-02 CG)
Duration: 
20 + 10