A Survey of Automatic Software Vulnerability Detection, Program Repair, and Defect Prediction Techniques

Table 2

Various feature parameters selected of deep learning technology on software vulnerability detection, program repair, and defect prediction.

Method type	CNN network	RNN network	DNN network
Code similarity vul detection	Iteration, dropout, hidden_layer, gradient_rate	hidden_layer, network depths, dropout, batch_size, iteration	hidden_layer, iteration, dropout, layer_size, learning_rate
Code pattern vul detection	filter_size, filter_num, hidden_layer	Dropout, batch_size, iteration, learning_rate, vector_dimension	—
Grammer program patching	—	learning_rate, batch_size, hidden_layer, token_length, hidden_unit, iteration, embedding_size, gradient_optimizer	—
Semantic program patching	—	hidden_unit, gradient_optimizer, vector_dimension, learning_rate, iteration, dropout, hidden_dimension	—
Within-project defect prediction	filter_num, filter_size, hidden_node, batch_size, iteration, embedding_dimension	Dropout, vector_dimension, hidden_layer, hidden_node, batch_size, iteration, learning_rate	hidden_layer, hidden_node, iteration, learning_rate, gradient_optimizer, batch_size
Crossproject defect prediction	filter_size, filter_num, learning_rate, vector_size	Dropout, hidden_layer, hidden_node, batch_size	Iteration, hidden_layer, hidden_node
Crossproject defect prediction	—	—	hidden_layer, hidden_node, activation_function