Baishakhi Ray (Columbia University)

(slides)

Software bugs cost millions of dollars to US economy. Improving software reliability has been one of the primary concerns of Programming Language (PL) and Software Engineering (SE) research over decades. Researchers developed different techniques, e.g., new languages, automatic bug finding tools, and code review processes to reduce software defects. However, the adoption of these methods in the real-world is still limited, partly because most of them require a significant amount of manual work from developers and have a steep learning curve.

To automate the bug-finding process, in this talk, I will discuss how we can leverage a large number of open source projects collected in software forges like GitHub. Thanks to such rich source of Software Engineering data that has become available to the researchers, we can now learn from common coding mistakes and automate them. I will further show how such data-driven knowledge can be leveraged to build new bug-finding and fixing tools to improve software reliability. Finally, I will discuss how such techniques can be adopted for emerging machine-learning-based software.