Improvement on PDP Evaluation Performance Based on Neural Networks and SGDK-means Algorithm.

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning

Background:

• Policy decision point (PDP) evaluation is crucial for access control systems.
• Existing PDP engines face performance challenges with large-scale policy sets.
• Efficient rule matching and evaluation are necessary for real-time decision-making.

Purpose of the Study:

• To propose a novel and efficient evaluation engine, XDNNEngine, for improving PDP performance.
• To leverage neural networks and a stochastic gradient descent K-means (SGDK-means) algorithm for enhanced policy evaluation.
• To reduce the computational cost associated with PDP evaluation on servers.

Main Methods:

• Clustering policy sets using K-means and asynchronous SGDK-means algorithms.
• Training neural networks to identify applicable rules based on policy features.
• Utilizing a quantitative neural network to minimize server computational load.
• Simulating request arrivals to compare XDNNEngine with existing systems (Sun PDP, XEngine, SBA-XACML).

Main Results:

• XDNNEngine achieved an evaluation time of approximately 2.5 ms for 10,000 rules with 10,000 requests.
• Compared to Sun PDP, XEngine, and SBA-XACML, XDNNEngine demonstrated significant reductions in evaluation time (98.27%, 90.36%, and 84.69%, respectively).
• The proposed engine shows superior performance in handling large-scale policy sets.

Conclusions:

• XDNNEngine offers a highly efficient solution for PDP evaluation, outperforming established systems.
• The integration of neural networks and SGDK-means provides a scalable and performant approach to policy management.
• This research contributes to advancing the efficiency of access control and policy enforcement mechanisms.