The proposed Differential Stimulus technology is an innovative solution component for AI and Machine Learning especially in real time intrusion detection and anomaly detection systems. It specifically addresses the correctness of the training data, which is critical for AI/ML. If the training data includes anomalies labeled as legitimate data or events, then later the system will not be able to identify as anomalies. In our practical work, we have seen that this is a major challenge, especially in already deployed systems, where the operators do not know that the system is already silently compromised, and therefore train the anomalies detection with incorrectly labeled data. Importantly, it can be used to find “hidden” attributes which were not known during the training process. Our technology address one of the key issues in using AI/ML for anomalies detection, and can provide great value to many AI/ML based efforts at the Air Force and beyond, in order to establish trust in training for intrusion and anomalies detection, and to provide further insights into already existing data sets. It can easily be integrated with existing and newly developed systems, and can also be used to evaluate already trained and deployed AI/ML based intrusion and anomalies detection systems. @objectsecurity | #AFWERXFusion2020 | #BOTFChallengeShowcase
