Recurrent Neural Networks for Malware Analysis

摘要: Using a recurrent neural network (RNN) that has been trained to satisfactory level of performance, highly discriminative features can be extracted by running sample through the RNN, and then extracting final hidden state hh where i is number instructions sample. This resulting feature vector may concatenated with other hand-engineered features, larger classifier on hand- engineered as well automatically determined features. Related apparatus, systems, techniques articles are also described.