Methods, systems and computer program products for detecting flow-level network traffic anomalies via abstraction levels. An exemplary embodiment includes a method for detecting flow-level network traffic anomalies in a computer network, the method including obtaining current distributions of flow level traffic features within the computer network, computing distances of the current distributions' components from a distributions model, comparing the distances of the current distributions to distance baselines from the distributions model, determining if the distances are above a pre-determined thresholds and in response to one or more of the distances being above the pre-determined thresholds in one or more distributions, identifying the current condition to be abnormal and providing indications to its nature.