Accuracy & testing: On the whole, our detectors are trained via machine learning, which yields higher accuracy than traditional approaches, and works well on unstructured data like files and messages. Our machine learning detectors are geared towards false-positive reduction. It's important to note that when you are testing these detectors, you may explicitly know what you're looking for the detector to flag on, but in real-world situations, these detectors need to explore context and other factors to make determinations. Not to mention, true positives are significantly rarer in practice in proportion to their frequency in a test environment. That being said, accuracy rates you experience in testing may not necessarily reflect what you experience when these detectors are enabled in the wild.