Edward Takahashi ‘14
Undeterred by this, he jumped into the research experience and was able to develop efficient algorithms for detecting anomalies in large network traffic data sets.
The area that Ed worked on is called streaming algorithms. A streaming algorithm is one that is able to process its input as a stream of data—hence the name—without being able to go back to see what came earlier in the stream. This class of algorithm is especially relevant to networking data since there are gigabytes of data passing through Internet core routers every second and it is infeasible to store all of it. In the age of Big Data, these types of algorithms are extremely important for being able to cope with the flood of digital information.
The specific problem that Ed worked on was to detect anomalies, or significant changes, in streaming network data. These anomaly events can have a number of causes, including sudden interest in a website or video, attacks on the network by hackers, or the spread of viruses and worms. The specific technique that Ed focused on was using a mathematical tool called a divergence to detect changes between normal and abnormal traffic. A divergence is common statistical tool for measuring the difference in the probability distributions of two sources. Ed’s idea was to design a streaming algorithm for measuring such a divergence on network traffic in an efficient manner. By the end of the summer, he was able to reliably detect anomalies in the data sources available to us.
Ed, Yuting Chen ’14, and Dr. Lall published this work at the Midstates Conference for Undergraduate Research in Computer Science and Mathematics (MCURCSM) 2012. Ed gave a talk based on this paper at the conference, which was held at Ohio Wesleyan University.
While at Denison, Ed was actively involved in the Asian Cultural Club, helping to organize the annual Lunar New Year Festival celebration. He was also on the university’s competitive programming team. Ed plans to work in industry for a few years before going to graduate school in computer science.