Research On Device Recognition Based On Cross-browser Fingerprinting

Device identification is mainly used to identify users and equipment to enhance the system authentication ability and security.Browser fingerprint is used for device identification,and browser fingerprint is convenient and quick.But in the evolution of network attack and defense,it is easier for attackers to circumvent traditional fingerprint identification.Thus,cross-browser fingerprint technology emerged.Cross-browser fingerprints can generate fingerprints using hardware information of the device,which can be used for device authentication,scene identification,track tracking,etc.Unlike the defect that a browser fingerprint is easy to modify,a cross-browser fingerprint has a better recognition effect and stronger robustness.However,many cross-browser fingerprint research still mainly uses static information of hardware,which gives attackers the possibility to modify static hardware information.To solve these problems,this paper proposes an effective cross-browser fingerprint based on clock frequency information,which makes up for some shortcomings of crossbrowser fingerprints.The main work of this paper is as follows:(1)This paper proposes a Clock Fingerprint Model Based on Adaptive Pareto Principle(CFMAP).Clock fingerprints are generated based on clock skew theory.CPU clocks and GPU clocks contain quartz Crystal oscillators,and small changes in these crystals can lead to small but measurable differences in clock frequency.In this paper,the function execution time is used as the primary expression of CPU clock skew,and the frame rate of 3D model rendering is used as the primary expression of GPU clock skew.This paper proposes a Pareto principle of self-adaptation according to the CPU and GPU loads.It then constructs a clock fingerprint model with a better recognition effect and stronger robustness.(2)This paper proposes a CFMAP recognition algorithm based on the KNN algorithm.Compared with the mode-based clock fingerprint recognition algorithm,the classification method based on machine learning is more in line with the current research direction of computer science.At the same time,KNN is a distance-based algorithm,which can be well integrated with CFMAP.In this paper,the effectiveness of the KNN algorithm and CFMAP is verified by comparing several machine learning algorithms with other researchers’ methods.At the same time,the time stability and temperature stability of the model and method are studied in this paper.(3)This paper proposes a threat model of clock fingerprint and analysis the security.From the attacker’s point of view,this paper studies the possible security threats against clock fingerprints.Experiments of virtual machine attacks and long-time high CPU load compression attacks are carried out,and the security of the proposed method and model is verified by comparing it with other researchers’ models and methods.