Abstract
The electromagnetic radiation of electronic equipment carries information and can cause information leakage, which poses a serious threat to the security system; especially the information leakage caused by encryption or other important equipment will have more serious consequences. In the past decade or so, the attack technology and means for the physical layer have developed rapidly. And system designers have no effective method for this situation to eliminate or defend against threats with an absolute level of security. In recent years, device identification has been developed and improved as a physical-level technology to improve the security of integrated circuit (IC)-based multifactor authentication systems. Device identification tasks (including device identification and verification) are accomplished by monitoring and exploiting the characteristics of the IC’s unintentional electromagnetic radiation, without requiring any modification and process to hardware devices, thereby providing versatility and adapting existing hardware devices. Device identification based on deep residual networks and radio frequency is a technology applicable to the physical layer, which can improve the security of integrated circuit (IC)-based multifactor authentication systems. Device identification tasks (identification and verification) are accomplished by passively monitoring and utilizing the inherent properties of IC unintended RF transmissions without requiring any modifications to the analysis equipment. After the device performs a series of operations, the device is classified and identified using a deep residual neural network. The gradient descent method is used to adjust the network parameters, the batch training method is used to speed up the parameter tuning speed, the parameter regularization is used to improve the generalization, and finally, the Softmax classifier is used for classification. In the end, 28 chips of 4 models can be accurately identified into 4 categories, then the individual chips in each category can be identified, and finally 28 chips can be accurately identified, and the verification accuracy reached 100%. Therefore, the identification of radio frequency equipment based on deep residual network is very suitable as a countermeasure for implementing the device cloning technology and is expected to be related to various security issues.
Electromagnetic radiation-based IC device identification and verification using deep learning