Detecting Novel Scans Through Pattern Anomaly Detection

We introduce a technique for detecting anomalous patterns in a categorical feature (one that takes values from a finite alphabet). It differs from most anomaly detection methods used to date in that it does not require attackfree training data, and it improves upon previous methods known to us in that it is aware when it is adequately trained to generate… CONTINUE READING