A cross domain access control model for medical consortium based on DBSCAN and penalty function

Abstract

Graded diagnosis and treatment, referral, and expert consultations between medical institutions all require cross domain access to patient medical information to support doctors' treatment decisions, leading to an increase in cross domain access among various medical institutions within the medical consortium. However, patient medical information is sensitive and private, and it is essential to control doctors' cross domain access to reduce the risk of leakage. Access control is a continuous and long-term process, and it first requires verification of the legitimacy of user identities, while utilizing control policies for selection and management. After verifying user identity and access permissions, it is also necessary to monitor unauthorized operations. Therefore, the content of access control includes authentication, implementation of control policies, and security auditing. Unlike the existing focus on authentication and control strategy implementation in access control, this article focuses on the control based on access log security auditing for doctors who have obtained authorization to access medical resources. This paper designs a blockchain based doctor intelligent cross domain access log recording system, which is used to record, query and analyze the cross domain access behavior of doctors after authorization. Through DBSCAN clustering analysis of doctors' cross domain access logs, we find the abnormal phenomenon of cross domain access, and build a penalty function to dynamically control doctors' cross domain access process, so as to reduce the risk of Data breach. Finally, through comparative analysis and experiments, it is shown that the proposed cross domain access control model for medical consortia based on DBSCAN and penalty function has good control effect on the cross domain access behavior of doctors in various medical institutions of the medical consortia, and has certain feasibility for the cross domain access control of doctors.

Keywords: Access control; Blockchain technology; Medical consortium; Medical informatization.

Fig. 1

Schematic diagram of graded diagnosis and treatment

Fig. 2

Schematic diagram of referral between hospitals

Fig. 3

Schematic diagram of experts’ consultation

Fig. 4

Cross domain access control model for medical consortium based on DBSCAN and penalty function

Fig. 5

Workflow diagram of the proposed model

Fig. 6

Example diagram of cross domain access data by doctors in Hospital A

Fig. 7

Plots of uploading of doctors’ cross domain access log are generated under 4G (a) and 5G (b). The value X-axis represents the number of access logs, and the value of Y-axis represents time consumption of uploading to the blockchain distributed ledger. The latency of 4G is in the range of (0.05,0.08) second, while the latency of 5G is in the range of (0.005,0.01) second

Fig. 8

DBSCAN clustering sub-plots were generated under different parameters (eps, min_s). In the sub-plots, the value X-axis represents the access time of a doctor, and the value of Y-axis represents department labels. Sub-plot (a)’s generation depends on parameter (15, 20), sub-plot (b)’s generation depends on parameter (30, 20), sub-plot (c)’s generation depends on parameter (25, 20), and sub-plot (d)’s generation depends on (30, 15)

Fig. 9

DBSCAN clustering sub-plot generated under parameter (eps=30, min_samples=15). The coordinate values of each point represent the positioning of the doctor’s access behavior in the figure. The value X-axis represents the access time of a doctor, and the value of Y-axis represents department labels

Fig. 10

DBSCAN clustering sub-plot generated under parameter (eps=30, min_samples=15). The coordinate values of each point represent the positioning of the doctor’s access behavior in the figure. The value X-axis represents the access time of a doctor, and the value of Y-axis represents department labels. The black line is the boundary line to distinguish normal behavior and abnormal behavior

Fig. 11

The function fitting sub-plots were automatically generated by Python through polynomial regression. The value X-axis represents the access time of a doctor, and the value of Y-axis represents department labels. The blue dots represent 79 random points distributed along the boundary line of Fig. 10. Sub-plot (a) illustrates the result of fitting with a 5th-degree polynomial. Sub-plot (b) illustrates the result of fitting with a 6th-degree polynomial. sub-plot (c) illustrates the result of fitting with a 7th-degree polynomial. Sub-plot (d) illustrates the result of fitting with a 8th-degree polynomial

Fig. 12

Schematic of doctor risk measurement across domains. The value X₁-axis represents the access time of a doctor, and the value of X₂-axis represents department labels. In order to adapt to the construction of penalty function, we use X₂ to substitute Y

Fig. 13

Interior point penalty function optimizes results for cross domain access. The value X₁-axis represents the access time of a doctor, and the value of X₂-axis represents department labels. In order to adapt to the construction of penalty function, we use X₂ to substitute Y. The red points stands for access points

Fig. 14

Cross domain access situation sub-plots of comparison before and after control. The value X-axis represents the access time of a doctor, and the value of Y-axis represents department labels