Paper Title
The Development of Man-In-The-Middle Threat Detection Models with Neural Networks

Abstract
In the digital era, the spontaneous growth of Internet technologies has led to the significant importance of the information security field. The complexity of online threats and attacks puts the users’ computer infrastructure at great risk. Man-in-the-Middle attacks are among the most common threats to users’ data integrity, confidentiality, and availability. Although traditional methods still give opportunities to analyze the network traffic to detect Man-in-the-Middle attacks, they are becoming less relevant with the rapid development of intelligent tools. Therefore, new approaches based on neural networks offer powerful instruments to reveal anomaly traffic and discover hidden patterns in attacks. In this work, Dense, Convolutional, Long short-term memory, and a hybrid model of Convolutional and long short-term memory neural networks are utilized to detect Man-in-the-Middle attacks. The dataset of 2504267 instances (1145272 normal and 1358995 malicious) is used to conduct the experiments. All four types of neural networks showed efficiency with accuracy scores of 0.96 and above, and the CNN-LSTM neural network achieved the best score of 0.998. These experiments demonstrated how a deep learning approach could be beneficial in detecting Man-in-the-Middle attacks. These models are also important in analyzing other kinds of threats and attacks, which will be covered in subsequent research. Keywords - Internet technologies, Man-in-the-Middle attacks, Dense neural network, Convolutional neural network, Long short-term memory network, CNN-LSTM