@inproceedings{f581a45514324a4aa040ff6aa613384a,
title = "An eBPF-Based Scheduler for the AFL++ Fuzzer",
abstract = "Fuzzing is one of the most effective software vulnerability discovery techniques. It consists of repeatedly executing a target program with mutated inputs. Fuzzing campaigns are usually run with multiple fuzzer instances running in parallel to explore the target in different ways simultaneously. This accelerates the bug discovery process. eBPF is a Linux kernel technology that allows user-space programs to be loaded into the kernel to modify kernel behavior at runtime. The Linux SCX framework allows eBPF programs to be attached to scheduling events to implement custom CPU scheduling policies. In this study, we present an approach for prioritizing AFL++ fuzzing processes using eBPF -based scheduling. Our results show that the improvement in crash discovery with process-level scheduling is target-dependent, not universal.",
keywords = "AFL++, CPU Scheduling, eBPF, Fuzzing, sched ext",
author = "Ernesto Ortiz and Clemente Izurieta and Reinhold, \{Ann Marie\}",
note = "Publisher Copyright: {\textcopyright} 2025 IEEE.; 2025 Cyber Awareness and Research Symposium, CARS 2025 ; Conference date: 27-10-2025 Through 30-10-2025",
year = "2026",
month = jan,
day = "21",
doi = "10.1109/CARS67163.2025.11337453",
language = "English",
series = "2025 Cyber Awareness and Research Symposium, CARS 2025",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
booktitle = "2025 Cyber Awareness and Research Symposium, CARS 2025",
}