Proj CDeepFuzz Paper Reading: DeepStellar: Model-based quantitative analysis of stateful deep learning systems
Abstract
背景:关于test RNN-based stateful system的研究较少
本文:DeepStellar
Method:
- model RNN as an abstract state transition system
- design 2 trace similarity metrics and 5 coverage criteria
- STSim: state-based similarity
- TTSim: transition-based similarity
- BSCov: basic state coverage
- n-SBCov: n-Step Basic Boundary Coverage
- WSCov: Weighted State Coverage
- BTCov: Basic Transition Coverage
- WTCov: Weighted Transition Coverage
- adversarial detection classifier(linear regression classifier), coverage guided test
- represent RNN as Discrete-Time Markov Chain to capture its statistical behaviors
实验
Competitors: FGSM, BIM, DeepFool
datasets: DeepSpeech 0.1.1, DeepSpeech 0.3.0, MNIST-LSTM, MNIST-GRU
结果:
- the similarity metrics could effectively capture the differences between samples with very small perturbations(Q: why neeed to capture the difference in perturbations?)
- the coverage criteria are useful in revealing erroneous behaviours.