The Missing Piece of Test Automation Puzzle: Explainable Deep Oracles
31/03/2022, 10:30am
Speaker
Abstract
Given an input to the software, the challenge of distinguishing the expected, correct behavior from the incorrect one is called the test oracle problem. The absence of automated test oracles demands human to decide whether observed behavior and generated output is correct, increasing the cost of testing to a great extent. Furthermore, it demonstrates a significant bottleneck that not only inhibits test automation and uptake of automated testing mechanisms, but also complicates debugging, i.e., root causing the bug found through testing and fixing it. The more expressive and informative is the decision of the test oracle, i.e., explains why a decision is made, the more effective and efficient localizing detected bugs and fixing them is. In this talk, I will introduce the concept of Explainable Deep Oracles. Such oracles not only overcome the test oracle problem, but also bridge the gap between testing and debugging, integrating software reliability assurance pipeline.
Speaker Bio
Reyhan Jabbarvand is an Assistant Professor of Computer Science at the University of Illinois at Urbana-Champaign. Her current research focuses on explainable neural software analysis, including (1) the use of explainable neural models to enhance software analysis and (2) leveraging software analysis techniques to improve the reliability of neural software, including ML-intensive mobile apps and self-driving software systems. She is a recipient of the Google Ph.D. Fellowship for her work on advancing energy testing of mobile applications and has been recognized as a rising star in EECS by UIUC in 2019.