… deep neuron networks, ie, DeepHunter [55] and TensorFuzz [38], both of which adopt
coverage-based fuzzing … MODE: automated neural network model debugging via state differential …

… Taking steps towards debugging these issues, TensorFuzz adapts coverage-based fuzzing
… debugging and providing explanations of neural network output during the training process. …

… Following standard software fuzzing, we will examine how to use SUT performance to
guide mutation type and parameter selection to more quickly find inconsistencies. We will also …

… using an 8-layer convolutional neural network. Their study was to examine the effectiveness
of visual posthoc model explanations in diagnosing model errors and debugging. The bugs …

Deep neural networks (DNN) are increasingly applied in safety-critical systems, eg, for face
recognition, autonomous car control and malware detection. It is also shown that DNNs are …

… testing, file format testing or testing neural networks [7], [8]. … Coverage guided fuzzing
tools (CFG) usually consist of an input mutation generator, feedback guidance, and a fuzzing …

… Neural Network (DNN)-based software, such as metamorphic testing [75, 97], mutation testing
[57], and fuzzing [… Based on these criteria, some studies apply coverage-guided fuzzing on …

… seeds, such coverageguided fuzzers usually develop strategies … Notably, a number of recent
coverage-guided fuzzers (eg, AFL [… further use neural networks to guide the fuzzing process. …

… Here DL software includes infrastructure code that performs core neural network computations
and … Deephunter: a coverage-guided fuzz testing framework for deep neural networks. In …

… the architecture of the deep neural network (DNN), and … To test the data-driven deep
neural networks, a common way … Coverageguided testing is a representative and widely used …