Network Host Cardinality Estimation Based on Artificial Neural Network

Cardinality estimation plays an important role in network security. It is widely used in host cardinality calculation of high-speed network. However, the cardinality estimation algorithm itself is easy to be disturbed by random factors and produces estimation errors. How to eliminate the influence of these random factors is the key to further improving the accuracy of estimation. To solve the above problems, this paper proposes an algorithm that uses artificial neural network to predict the estimation bias and adjust the cardinality estimation value according to the prediction results. Based on the existing algorithms, the novel algorithm reduces the interference of random factors on the estimation results and improves the accuracy by adding the steps of cardinality estimation sampling, artificial neural network training, and error prediction. The experimental results show that, using the cardinality estimation algorithm proposed in this paper, the average absolute deviation of cardinality estimation can be reduced by more than 20%.