The RNN has a recursive connection in a temporal manner that allows it to effectively understand sequential patterns, while the CNN finds arbitrary local patterns and generates a high-level representation by summarizing the found patterns.

The neural networks of recent studies can be categorized into two types: recurrent neural network (RNN) and convolutional neural network (CNN). Deep neural network is one of solutions to address such a limitation, as it automatically extracts effective features. If the features are poorly defined, the overall performance will be severely degraded. Although these models have shown successful performance (e.g., accuracy) on various classification tasks, they have a common limitation that they require an intensive effort to define high-quality features.

There are various machine learning models (e.g., logistic regression, support vector machine, random forest ) known to be effective for the classification problem. The detection of malicious action is basically a binary classification problem it determines whether a given file contains any malicious actions or not.

Hwp file open app