一句话：提出基于本福德定律检测模糊测试饱和点的方法，并在AFL++上验证其可行性。
模糊测试活动需及时判断是否饱和，以避免资源浪费或漏掉漏洞，但现有基于覆盖率的饱和判定往往不可靠。论文提出一种新思路：利用本福德定律刻画自然数分布的特性，假设模糊测试过程中的重复数字（如变异字节数）在达到饱和时会呈现受本福德定律支配的模式，此时灰盒模糊器因无种子可优先而趋于无偏，接近随机黑盒测试。作者计划用先进模糊器 AFL++ 在 29 个真实程序上检验该假设，并实证表明可用本福德定律检测模糊测试饱和。
Knowing when a fuzzing campaign has reached saturation is crucial for practitioners to avoid unnecessarily lengthy campaigns without missing bugs within given resources. Unfortunately, existing solutions for determining the saturation point rely on coverage measurements, which are often error-prone and unreliable. In this paper, we present a novel way to detect saturation in fuzzing based on Benford's law, which describes the characteristics of naturally occurring numbers. Specifically, we hypothesize that repeatedly occurring numbers in the fuzzing process, such as the number of mutated bytes, should show specific numerical patterns dictated by Benford's law when the fuzzer reaches saturation, thereby the fuzzing process becomes less biased (hence, more natural). The key observation is that grey-box fuzzers become less biased as in random black-box testing when they reach saturation because there will be no seed to prioritize. We aim to test our hypothesis on 29 real-world programs using the state-of-the-art fuzzer, AFL++, and empirically show that one can use Benford's law to detect fuzzing saturation.
@inproceedings{paperbot9596,
title = {On the Applicability of Benford’s Law to Detect Saturation in Fuzzing (Registered Report)},
author = {Jungwoo Lee and Haeun Lee and Sangjun Park and Sang Kil Cha},
booktitle = {ISSTA},
year = {2025},
doi = {10.1145/3713081.3731723}
}