Despite significant advances in software engineering during the last half century, debugging remains as one of the most difficult, labor-intensive, tedious activity to date. Debugging demands that human developers acquire deep understanding of the target program and reason about complicated cause-effect relations among an enormous number of code entities in real-world software. Researchers had developed automated techniques that infers the location of a fault by statistically analyzing observed testing results, however, they could not achieve high accuracies for effective real-world applications. In this talk, I will present a new automated debugging approach called mutation-based fault localization (MBFL). The key idea of MBFL is to acquire useful debugging information by systematically applying mutations to the target programs. The case studies with real-world open-source projects will show how MBFL successfully assists developers debugging complicated real-world faults with multilingual features.