A Novel Method for Anomaly Detection Using Scan Statistics
Description:
Overview of Technology
Potential methods for detecting clusters/anomalies while simultaneously dealing with model misspecification
Description of Technology
Mississippi State researchers have developed scan-based methods to identify and statistically quantify abnormal and potentially adversarial behavior. The researchers propose to effectively extend the usefulness of scan-based statistics to a multitude of applied settings.
The goal is to identify anomalous activity while also allowing for a quantifiable degree of misspecification of normal activity. They present a novel tool called theta-filtering for scan-based methods. In essence, theta-filtering presents a host of potential methods for detecting clusters/anomalies while simultaneously dealing with model misspecification.
Applications
Fraudulent activities have caused losses of multi-million dollars in credit card, retail, service, and government businesses.
This invention is a novel statistical technique that is tailored to statistically identify and quantify clusters or anomalous activities as quickly as possible.
Seeking
- Development partner
- Commercial partner
- Licensing
IP Status