Generalized likelihood ratio test (GLRT) based network intrusion detection system in wavelet domain
View Patent ↗An improved system and method for detecting network anomalies comprises, in one implementation, a computer device and a network anomaly detector module executed by the computer device arranged to electronically sniff network traffic data in an aggregate level using a windowing approach. The windowing approach is configured to view the network traffic data through a plurality of time windows each of which represents a sequence of a feature including packet per second or flow per second. The network anomaly detector module is configured to execute a wavelet transform for capturing properties of the network traffic data, such as long-range dependence and self-similarity. The wavelet transform is a multiresolution transform, and can be configured to decompose and simplify statistics of the network traffic data into a simplified and fast algorithm. The network anomaly detector module is also configured to execute a bivariate Cauchy-Gaussian mixture (BCGM) statistical model for processing and modeling the network traffic data in the wavelet domain. The BCGM statistical model is an approximation of α-stable model, and offers a closed-form expression for probability density function to increase accuracy and analytical tractability, and to facilitate parameter estimations when compared to the α-stable model. Finally, the network anomaly detector module is further configured to execute a generalized likelihood ratio test for detecting the network anomalies.
1. A system for detecting network anomalies comprising:
a computer device; and
a network anomaly detector module executed by the computer device;
wherein:
the computer device is arranged to electronically collect and sniff network traffic data in an aggregate level using a windowing approach,
the windowing approach is configured to view the network traffic data through a plurality of time windows, each of which represents a sequence of a feature including packet per second or flow per second,
each time window of the plurality of time windows is analyzed individually, and configured to have overlap with a neighboring time window to increase accuracy in specifying positions of the network anomalies and preventing sudden changes of an inference step, and
the network anomaly detector module is configured to execute:
a wavelet transform for capturing properties of the network traffic data,
a bivariate Cauchy-Gaussian mixture statistical model for processing and modeling the network traffic data in a wavelet domain, and
a generalized likelihood ratio test for classifying patterns in the network traffic data and detecting the network anomalies, and reporting the network anomalies to a user of the computer device.
2. The system of claim 1 , wherein the computer device sniffs the network traffic data in real time.
3. The system of claim 1 , wherein the computer device sniffs the network traffic data periodically.
4. The system of claim 1 , wherein the wavelet transform is a multiresolution transform, and is configured to capture one or more properties of the network traffic data, including long-range dependence and self-similarity, and the wavelet transform is a multiresolution transform is configured to decompose and simplify statistics of the network traffic data into a simplified algorithm.
5. The system of claim 4 , wherein the wavelet transform is Daubechies wavelets, and is configured to solve of one or more problems, at least one of which includes capturing self-similarity.
6. The system of claim 1 , wherein the bivariate Cauchy-Gaussian mixture statistical model is an approximation of α-stable model, and offers a closed-form expression for probability density function to increase accuracy and analytical tractability, and to facilitate parameter estimations when compared to the α-stable model.
7. The system of claim 6 , wherein compatibility between the bivariate Cauchy-Gaussian mixture statistical model and the network traffic data in the wavelet domain is investigated by at least one of comparing histogram of the network traffic data and the probability density function or applying a Kolmogrov-Smirnov test.
8. A method for detecting network anomalies comprising the steps of:
electronically sniffing network traffic data at a computer device in an aggregate level by using a windowing approach;
executing a network anomaly detector module on the computer device, the network anomaly detector module including:
(a) a wavelet transform for capturing properties of the network traffic data;
(b) a bivariate Cauchy-Gaussian mixture statistical model for processing and modeling the network traffic data in a wavelet domain; and
(c) a generalized likelihood ratio test for classifying patterns in the network traffic data and determining detecting the network anomalies; and
reporting the network anomalies to a user of the computer device,
wherein:
the windowing approach is configured to view the network traffic data through a plurality of time windows each of which represents a sequence of a feature including packet per second or flow per second, and
each time window is analyzed individually, and configured to have overlap with a neighboring time window to increase accuracy in specifying positions of the network anomalies and preventing sudden changes of an inference step.
9. The method of claim 8 , wherein the computer device sniffs the network traffic data in real time.
10. The method of claim 8 , wherein the computer device sniffs the network traffic data periodically.
11. The method of claim 8 , wherein the wavelet transform is a multiresolution transform, and is configured to capture one or more properties of the network traffic data, including long-range dependence or self-similarity, and the wavelet transform is a multiresolution transform is configured to decompose and simplify statistics of the network traffic data into a simplified algorithm.
12. The method of claim 11 , wherein the wavelet transform is Daubechies wavelets, and is configured to solve one or more problems, at least one of which includes capturing self-similarity.
13. The method of claim 8 , wherein the bivariate Cauchy-Gaussian mixture statistical model is an approximation of α-stable model, and offers a closed-form expression for probability density function to increase accuracy and analytical tractability, and to facilitate parameter estimations when compared to the α-stable model.
14. A computer readable medium having program instructions stored thereon for detecting network anomalies which, when executed by a computer device, causing the computer device to perform the steps of:
electronically sniffing network traffic data at the computer device;
executing a network anomaly detector module on the computer device, the network anomaly detector module including:
(a) a wavelet transform for capturing properties of the network traffic data;
(b) a bivariate Cauchy-Gaussian mixture statistical model for processing and modeling the network traffic data in the wavelet domain; and
(c) a generalized likelihood ratio test for classifying patterns in the network traffic data and determining anomalies; and
reporting the anomalies to a user of the computer device,
wherein the bivariate Cauchy-Gaussian mixture statistical model is an approximation of α-stable model, and offers a closed-form expression for probability density function to increase accuracy and analytical tractability, and to facilitate parameter estimations when compared to the α-stable model.
15. The computer readable medium of claim 14 , wherein a windowing approach is used to electronically sniff the network traffic data and the windowing approach is configured to view the network traffic data through a plurality of time windows each of which represents a sequence of a feature including packet per second or flow per second.