Document Type


Date of Award

Spring 5-31-2003

Degree Name

Master of Science in Telecommunications - (M.S.)


Electrical and Computer Engineering

First Advisor

Constantine N. Manikopoulos

Second Advisor

George Antoniou

Third Advisor

Bin He


Network traffic is bursty in nature and exhibits the property of self-similarity, the degree of which is measured by the Hurst parameter. Now, in any network there is always the possibility of the occurrence of fault traffic that can be caused due to faults or malfunction of a network component. Fault Detection Systems that check for traffic anomalies can trigger off an alert on the detection of any traffic behavior that deviates from normal. Such deviation is usually caused when a fault occurs. But in network traffic sudden bursts may occur due to the inherent behavior of a network application. Due to this burst a fault detection system could generate an alert if the burst crosses a preset threshold even though it is not a fault condition. This paper investigates the use of a data filter to reduce such false positives that may be caused due to sudden bursts or spikes.