Multi-Layer Perceptron Model on Chip for Secure Diabetic Treatment

Abstract

Diabetic patients use therapy from the insulin pump, a type of implantable medical device, for the infusion of insulin to control blood glucose level. While these devices offer many clinical benefits, there has been a recent increase in the number of cases, wherein, the wireless communication channel of such devices has been compromised. This not only causes the device to malfunction but also potentially threatens the patient's life. In this paper, a neural networks-based multi-layer perceptron model was designed for real-time medical device security. Machine learning algorithms are among the most effective and broadly utilized systems for classification, identification, and segmentation. Although they are effective, they are both computationally and memory intensive, making them hard to be deployed on low-power embedded frameworks. In this paper, we present an on-chip neural system network for securing diabetic treatment. The model achieved 98.1% accuracy in classifying fake versus genuine glucose measurements. The proposed model was comparatively evaluated with a linear support vector machine which achieved only 90.17% accuracy with negligible precision and recall. Moreover, the proposal estimates the reliability of the framework through the use of the Bayesian network. The proposed approach enhances the reliability of the overall framework by 18% when only one device is secured, and over 90% when all devices are secured.

Keywords

• Security
• machine learning
• insulin pumps
• deep learning
• implantable medical devices