Info For
Skip to Content
Entrance to Taylor University

Computer Science Student Set to Present Original Research at MAICS Conference

  • By: Meredith Sell
  • Published:
Euler Science Complex

Senior computer science major Mitchell Mays will present his original research paper at the 28th Modern Artificial Intelligence and Cognitive Science (MAICS) conference in Fort Wayne on April 28 and 29. Mays conducted the research for his paper, entitled Feature Selection for Malware Classification, as part of a two semester research course last year.

“The idea was to combine two different fields of computer science, one being cybersecurity and the other being artificial intelligence,” Mays said. “The hope is that some of the techniques that are used in artificial intelligence could be applied … to areas of cybersecurity.”

Mays explored the idea of using machine learning to identify malware. (Machine learning is a branch of artificial intelligence in which a program recognizes patterns in data and uses those patterns to classify new data.)

To carry out his research, Mays found a dataset of 10,000 files on Kaggle, a data science web site. Each file within the dataset had one of nine kinds of known malware. Mays wrote a script to run a static analysis of the files, essentially sorting through the files and logging occurrences of certain features. He worked with Assistant Computer Resource Manager Nate White ’13 to split the work between more than 50 of the Computer Science and Engineering Department’s computers and run the script overnight. It took about three nights to run the script through all 10,000 files.

Mays used two classifiers—a feedforward neural network and a convolutional neural network—to classify the features and then apply those classifications to the files in question. Through the project, he reached a 97 percent accuracy of identifying each file’s particular malware based on the features his classifiers “learned” were part of the malware.

“There’s enough granular detail that’s being picked up and recognized that it can distinguish with really high accuracy between these kinds of malware,” Mays said. “That would lead me to believe that it would be able to also distinguish those same granular, minute details between a good software and a bad software.”

Mays submitted his work to MAICS last spring and, this March, received word that his paper was accepted for presentation at the conference on April 28 and 29. This past January, Mays presented his work to security researchers at Lockheed Martin’s Advanced Technology Labs in Arlington, Va. He will also be presenting on the topic at Purdue’s 18th Annual Information Security Symposium, CERIAS 2017, on April 18 and 19.

Mays is graduating this spring with a Bachelor of Science in Computer Science/Systems and a Psychology minor. He plans to eventually pursue graduate studies in the area of cognitive science.