Loading patent…
AssigneeTohoku University
InventorsYuji Waizumi, Hiroshi Tsunoda, Yoshiaki Nemoto
Estimating the term…
A system provides definitions of network states, and identifies a cause for the anomaly upon detection. A traffic measuring portion (characteristic quantity generating portion) counts the number of packets/time slot classified by traffic type with respect to network traffic, generating a characteristic quantity; a memory portion stores information about the characteristic quantity; a portion calculates correlation coefficients between each pair of characteristic quantities classified by traffic type; a portion generates a histogram from correlation coefficients; a portion for determining the severity of an anomaly based on the histogram; a portion evaluates the similarity of an anomaly of interest to a reference anomaly using the occurrence probabilities of correlation coefficients; and a portion assigns a color to each pixel according to its occurrence probability value, generating an n×n picture. Visualization of network state is achieved using probability distribution vectors derived from correlation coefficients obtained from each characteristic quantities pair.
Method for detecting an anomaly in a network comprising a feature value generating step for counting the number of packets per each time slot for each traffic type with respect to the traffic of a network, and generating it as a feature value;
System for detecting an anomaly in a network comprising a feature value generating portion for counting the number of packets per each time slot for each traffic type with respect to the traffic of a network and generating it as a feature value;
Dependencies are inferred from the claim text.