Today, the use of machine leaning techniques for intrusion detection based on anomaly detection is one of the most important topics in the field of information security. One of the challenges in this field is restriction of labeled samples. Thus, training learning models that is capable of providing sufficient accuracy to detect attacks, is difficult. One of the methods that have been proposed to address this challenge is transfer learning. The purpose of the transfer learning is improving learning models using knowledge transfer from one environment to the other environments. The environment from which knowledge is extracted, known as source domain and the environment in which extracted knowledge is used, known as target domain. In intrusion detection field, if there is no labeled sample in network, we can use labeled samples from other networks. The aim of this thesis is to develop a transfer learning method in the field of network security. For this purpose, a general structure for transfer learning is presented in the context of intrusion detection. By following this structure, two different methods have been proposed. Both methods are designed to learn a model in target network where there are no labeled samples in. To build these models, the labeled samples from source network are used. Key Word s : Intrusion Detection, Transfer Learning, Source Network, Target Network
