In-situ trainable intrusion detection system
A computer implemented method detects intrusions using a computer by analyzing network traffic. The method includes a semi-supervised learning module connected to a network node. The learning module uses labeled and unlabeled data to train a semi-supervised machine learning sensor. The method records events that include a feature set made up of unauthorized intrusions and benign computer requests. The method identifies at least some of the benign computer requests that occur during the recording of the events while treating the remainder of the data as unlabeled. The method trains the semi-supervised learning module at the network node in-situ, such that the semi-supervised learning modules may identify malicious traffic without relying on specific rules, signatures, or anomaly detection.
1. A computer implemented method that detects intrusions using a computer by analysing network traffic comprising:
coupling a semi-supervised learning module to a network node that uses labeled and unlabeled data to train a semi-supervised machine learning sensor;
recording events that comprise a feature set that include unauthorized intrusions and benign requests;
identifying at least some of the benign behavior that occurs during the recording of the events while treating the remainder of the data as unlabeled; and
training the semi-supervised learning module at the network node in-situ, such that the computer implemented method including the semi-supervised learning module is configured to identify the malicious traffic without relying on specific rules, signatures, or an anomaly detection;
where the semi-supervised machine learning sensor comprises an extension of a plurality of pipelines and filters and the feature set is stored in one of the plurality of pipelines that the semi-supervised learning module is trained.
2. The computer method of claim 1 further comprising programming the semi-supervised machine learning sensor to transmit an alert when a threshold is exceeded.
3. The computer method of claim 1 where the semi-supervised machine learning sensor further comprises a plurality of sensors coupled to a plurality of network nodes that process network traffic.
4. The computer method of claim 1 where the feature set comprises data that identifies the time the data was received, the source and destination IP addresses, and the source and destination ports of the data seeking unauthorized intrusions.
5. The computer method of claim 1 where the semi-supervised learning module comprises a Laplacian Regularized Least Squares model.
6. The computer method of claim 1 further comprising storing the feature set comprising network data, packet data, and host data at the node it is received.
7. The computer method of claim 1 further comprising a plurality of semi-supervised learning modules that train a semi-supervised machine learning sensor staged at different locations of the plurality of pipelines and filters.
8. The computer method of claim 1 further comprising generating a ranking of the intrusion types and a corresponding probability of the respective occurrences.
9. The computer method of claim 1 further comprising programming a threshold that indicates a detected unauthorized intrusion.
10. The computer method of claim 9 where the automated detection causes an automated intervention against the intrusion.
11. The computer method of claim 1 further comprising retraining the semi-supervised learning module by interactively modifying training set data accessed by the semi-supervised learning module.
12. The computer method of claim 11 where the training set of data comprises a computer library comprising computer scripts.
13. The computer method of claim 11 where the semi-supervised learning module is executed on a parallel processor.
14. The computer method of claim 1 where the semi-supervised machine learning sensor communicates with an anomaly detector.
15. A system that detects intrusions using a computer by analysing network traffic, comprising:
a network node;
a semi-supervised learning module coupled to the network node that uses labeled and unlabeled data to train a semi-supervised machine learning sensor;
a recorder that records events that comprise a feature set that include unauthorized intrusions and benign requests; and
a computer readable medium programmed to:
identify at least some of benign behavior that occurs during the recording of the events while treating the remainder of the data as unlabeled; and
train the semi-supervised learning module at the network node in-situ, such that the system including the semi-supervised learning module is configured to identify malicious traffic without relying on specific rules, signatures, or anomaly detection;
where the semi-supervised machine learning sensor comprises an extension of a plurality of pipelines and filters and the feature set is stored in at least one of the plurality of pipelines that the semi-supervised learning module is trained.
16. The system of claim 15 where the semi-supervised machine learning sensor is configured to transmit an alert when a threshold is exceeded.
17. The system of claim 15 where the semi-supervised machine learning sensor further comprises a plurality of sensors coupled to a plurality of network nodes that process network traffic.
18. The system of claim 15 where the feature set comprises data that identifies the time the data was received, the source and destination IP addresses, and the source and destination ports of the data seeking unauthorized intrusions.
19. The system of claim 15 where the semi-supervised learning module comprises a Laplacian Regularized Least Squares model.