GoogleGranted patentSecurity & privacyInterest 6/10

Google patents method to spot network anomalies with less labeled data

Patent drawing
Patent drawing from the official filing.

Google describes a system called SPADE that helps computers find unusual patterns in network traffic. It uses a group of simple detectors to guess which unlabeled data points are problems, allowing the main system to learn without needing a large set of pre-marked examples.

Why it mattersThis approach could make it easier for organizations to monitor their networks for threats when they do not have enough historical data to train traditional security tools.

In the filing’s own words

Aspects of the disclosure are directed to a Semi-supervised Pseudo-labeler Anomaly Detection with Ensembling (SPADE) framework that is not limited by the assumption that labeled and unlabeled data come from the same distribution. SPADE utilizes an ensemble of one-class classifiers as the pseudo-labeler to improve the robustness of pseudo-labeling with distribution mismatch. Partial matching automatically selects critical hyper-parameters for pseudo-labeling without validation data, which is crucial with a limited amount of labeled data.

What exactly they claim (claim 1)

1. A method to train a binary classifier for anomaly detection, the method comprising: receiving, by one or more processors, unlabeled training data comprising data samples associated with anomaly detection; generating, by the one or more processors, pseudo labels for the unlabeled training data using a plurality of one-class classifiers, wherein a positive pseudo label is generated for a first unlabeled data sample of the unlabeled training data based on a threshold amount of the one-class classifiers indicating the first unlabeled data sample is anomalous and a negative pseudo label is generated for a second unlabeled data sample of the unlabeled training data based on the threshold amount of the one-class classifiers indicating the second unlabeled data sample is normal; assigning, by the one or more processors, the pseudo labels respectively to the data samples of the unlabeled training data to generate pseudo labeled training data; and training, by the one or more processors, the binary classifier to detect network anomalies based on minimizing a loss function that assigns a binary weight to each of the data samples of the pseudo labeled training data.

Facts

Company
Google (Google LLC)
Official title
Semi-supervised anomaly detection under distribution mismatch
Patent number
US12801846B2
Application number
18150277
Filed
Jan 5, 2023
Granted
Nov 17, 2026
Inventors
4
Claims
20

Classified under

G06F 40/216 Physics & computingG06N 5/022 Physics & computing

More from Google in Security & privacy

Summaries are simplified from the filing’s own abstract and first claim. A patent filing is not a product announcement.
Patent data and drawings: U.S. Patent and Trademark Office (USPTO). Nebula is not affiliated with or endorsed by the USPTO.