Fuzzing currently has two flavors—an existing generic, domain-agnostic, solution such as AFL, or developers often build a custom generator such as “X”-Smith that is more effective for a specialized domain such as C, SQL, and MLIR. However, constructing a custom fuzzer such as “X”-Smith generally requires significant developer or engineering effort, measured in person-months.

In this talk, I will reflect on my group’s experience of designing custom fuzzers for data-intensive computing and heterogeneous hardware domains. I will discuss how we had to encode domain-specific constraints, custom feedback guidance, custom search strategies, and custom mutation operators to make the fuzzing solutions effective for a specialized domain. Then, reflecting on this manual specialization effort, I will propose a new direction on how we should strive to bootstrap a custom fuzzer, automatically or semi-automatically, without too much manual effort.

Toward this vision of “bootstrapping a custom fuzzer without paying too much”, I will share several ongoing effort to find the right balance between the universality of a fuzzer and its effectiveness in a specialized domain: (1) custom mutation synthesis from examples, (2) automated grammar refinement to constrain fuzzing, (3) LLM-guided constraint-generation for mutation, and (4) a lightweight DSL for context-guided input generation.

Miryung Kim is a Professor and a Vice Chair of Graduate Studies in UCLA Computer Science. She directs Software Engineering and Analysis Laboratory. She helped define the new area of Software Engineering for Data Intensive Computing (SE4DA and SE4ML). She works on automated testing and debugging for Apache Spark and developer tools for heterogeneous computing. She conducted the first systematic study of refactoring practices in industry and quantified rearchitecting benefits at Microsoft using Windows version history. She conducted the largest scale study of data scientists in industry. Her group’s Java bytecode debloating JDebloat made a tech transfer impact to Navy.

She received her BS from KAIST and MS and PhD from University of Washington under the supervision of David Notkin. She was an Assistant Professor at the University of Texas at Austin, moved to UCLA as an Associate Professor with tenure in 2014, and was promoted to a Full Professor in 2019. She spent time as a visiting researcher at Microsoft Research. She is an Amazon Scholar at Amazon Web Services.

She produced 6 professors (Columbia, Purdue, two at Virginia Tech, etc). For her impact on nurturing the next generation of academics, she received the ACM SIGSOFT Influential Educator Award. She is a Program Co-Chair of ESEC/FSE 2022, one of top 2 conferences in SE. She is a Keynote Speaker at ASE 2019 and ISSTA 2022. She gave Distinguished Lectures at CMU, UIUC, UMN, UC Irvine, etc. She is an ACM Distinguished Member.