FuzzInsight: Optimizing the Efficiency-Effectiveness Trade-off in Coverage-Guided DNN Fuzzing
In brief
- FuzzInsight highlights biases in seed-selection strategies for DNN fuzzing.
- Equalized Sampling increases average seed utilization by 13 percentage points.
- LSCD, LSC and MDSC criteria reduce computational costs while maintaining effective coverage.
An analysis of deep neural network (DNN) fuzzing techniques reveals that current seed-selection strategies underutilize available data. FuzzInsight introduces Equalized Sampling, a method that boosts seed queue utilization by 13 percentage points. The study evaluates 84 configurations across four datasets and identifies non-structural coverage criteria that deliver efficiency gains without sacrificing coverage.
This summary gives you the essentials. The full article, its figures and quotes are on cs.AI updates on arXiv.org.
Read the original articleAI-generated summary of the source, checked by rules. Report an error