NEUZZ: Efficient Fuzzing with Neural Program Smoothing — HN submission

Sat, 16 Mar 2019 23:58:03 GMT

Neuzz: A neural-network-assisted fuzzer(S&P 19)

I am lead author of NEUZZ, a neural network assisted fuzzer. We build a NN model to learn dynamic program behaviors and use the NN model to guide mutation of fuzzer. NEUZZ outperforms AFL and other fuzzers by 3X more edge coverage on the 10 real world programs evaluated in the paper. Please find more detail of this work below. Paper URL: https://arxiv.org/abs/1807.05620 Code URL: https://github.com/Dongdongshe/neuzz

2 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

NEUZZ: Efficient Fuzzing with Neural Program Smoothing · HackerPapers discussions

NEUZZ: Efficient Fuzzing with Neural Program Smoothing — HN submission