Extraction of Computer-Inherent Characteristics Based on Time Drift and CPU Core Temperature

Abstract

The number of computers that provide information services via the Internet is constantly increasing, particularly for IoT applications. Compared to the servers in managed data centers, IoT computers have an increased risk of contamination from unidentified computers. It is therefore important for applications that utilize IoT to identify the appropriate computers to use. However, it is difficult to assign digital identifiers with adequate protection to a huge number of IoT computers. In this work, we have devised a method to extract computer-specific features from the characteristics of the CPU core temperature and the drift of the time information. This feature data can be treated as computer-specific information, just like human biometric information. We performed experiments on two types of computer (three of each) with the same software settings and obtained a regression linear equation for each that represents the time deviation per temperature. The correlation coefficients of these equations were greater than 0.9 for all, and a strong positive correlation was obtained. From the equation and the temperature values, we found that it is possible to estimate the computer-specific time deviation. Our method does not require the implementation of a special temperature sensor. Therefore, it shows good potential for future applications.