Prediction of middle box-based attacks in Internet of Healthcare Things using ranking subsets and convolutional neural network

Middle-box based attacks create serious functional defects in the devices such as firewalls, address translators, load balancers, server units, and other data inspecting devices. Middle-box based attacks are more severe than terminal node-based attacks. These attacks are mainly initiated to malfunction the internal events of middle-level devices. Notably, Internet of Healthcare Things (IoHT) is a completely distributed and heterogeneous environment that leads into more vulnerable events. In the distributed medical system field, the middle-box devices manage secret medical data transactions, internal network communication, patient data protection principles and packet inspection procedures. The medical data collected from each patient needs secrecy and stability in various aspects in the IoHT environment. The middle-box based attacks injected into the IoHT nodes create inefficiency in maintaining patient data, Denial of Service (DoS) at middle-box nodes, excessive diagnosis time, and lack of data protection. For handling these issues, a new security architecture and a new attack detection model are proposed in this paper which has been developed using ranking subsets and Convolutional Neural Network (CNN) principles. The proposed CNN with ranking principles model is designed with binary class fuzzy fisher face optimization technique and flower pollination optimization method to initiate feature extraction. The extracted features of each data flow are analyzed using CNN and ranking subset methodologies. In this proposed model, the continuous involvement of middle-box events in respective devices are classified under various anomaly cases and legitimate cases. The proposed IoHT model attains maximum success rate than the existing models as indicated in implementation section.