IGLOO SECURITY, INC. Acquired 6 patents related to AI and Security Monitoring

- IGLOO SECURITY acquires AI and Security monitoring related patents. Reinforcing position as a leading in AI and Cyber Security company in Korea.

- The AI will be applied to SPiDER TM AI Edition, the first AI security monitoring solution in Korea.

[April 28, 2020] IGLOO SECURITY secures its position as a leader in AI security monitoring by acquiring AI and security monitoring patents. IGLOO SECURITY announced that it has completed the registration of six patents related to AI and security monitoring including 'an intelligent security monitoring system and method using a combination of supervised learning-based alert analysis and non-supervised learning-based anomaly detection techniques.'

The patents aimed at minimizing false positives and increasing visibility into security threats that are under cover over the long term with increasing security events. IGLOO SECURITY plans to apply this patented technology to SPiDER TM AI Edition, an AI security monitoring solution.

The patent for 'an intelligent security monitoring system and method using a combination of supervised learning-based alert analysis and non-supervised learning-based anomaly detection techniques.' improves the accuracy of alert prediction and prevents unknown attacks by combining supervised learning and unsupervised learning. By analyzing high-risk events and abnormalities automatically identified by AI algorithms by time flow and attack stages, it will be able to accurately select high-risk events that need to be addressed first and detect new and mutated threats that are difficult to detect with existing security equipment to minimize security gaps.

The patent for 'the artificial intelligence-based security event analysis system and its method through semi-supervised learning' is focusing on minimizing the time required for labeling tasks that give security events the correct answer and improving the accuracy of security policies.

This is a method of labeling only some of the clustered data and letting the algorithm make judgments on the rest of the data based on the labeled data. The accuracy of the algorithm and security policy is improved through continuous feedback on the results of the supervised learning.

The patent for 'a model selection system and method for unsupervised learning anomaly detection' is a technique to select the best anomaly detection model by predicting and evaluating the learning results provided by the anomaly detection model when a new unlabeled dataset is received based on the criteria for evaluating the anomaly detection model that has learned the dataset labeled with the target value for the input value.

Through this, it is possible to solve the difficulty of evaluating the results of unsupervised learning, and to increase the visibility of abnormal behaviors that can develop into high-risk threats.​​