Estimation of data traffic flows from aggregate measurements

This paper considers the problem of estimating multiple different traffic flows moving on the same physical support from aggregate measurement. This problem is relevant in the area of Internet traffic management, in particular for anomaly detections, but it also appears of interest in many other applicative areas. In this paper, two on-line procedures are proposed, characterized by the search of an optimal trade-off between the conflicting requirements of achieving good estimation accuracy and of keeping the measurement costs sufficiently low. The procedures have been tested and validated against real data with satisfactory results.