The scale and complexity of networks supporting IoT will quickly outstrip human capabilities to defend them. Devices can be elusive; they will hide behind gateways and roaming arrangements. The challenge is how to monitor and learn the behavior of the different classes of devices. IoT devices are like flocks of birds: they move in large groups in the same direction; when one deviates, usually the rest of the group deviates. We can use unsupervised machine learning techniques with a 5-step process to help solve the IoT Security problem. 1. Observe; 2. Cluster; 3. Detect Anomalies; 4. Root Cause Analysis; 5. Remediate.