Method for detection of malign instrusions in a communication system and related detector

摘要

A method is proposed for detection of malign intrusions in a communication system in which flows of streaming data are monitored by means of a processing module in which a set of streaming data points has been sequentially inputted into a processing module in order to calculate a first value of a self-similarity degree. Furthermore, said first value of self-similarity degree is calculated by means of a plurality (index k) of auxiliary variables, wherein said index (k ‰¥ 1) is continuously incremented for each new fed data point. Furthermore, a new streaming data point is fed into a corresponding modified estimator (= mE) comprising the processing module that runs under a first mode wherein a second actualized value of the self-similarity degree is newly calculated. Now and according to a (possibly pre-definable) finite number n of data points, the processing module provides a windowed based calculating mode which is activated (or activable) by incoming of the new streaming data point if the index k is at least equal to the finite number n and wherein at least the oldest data point having been fed/inputted in the processing module is excluded from the processing module. According to the same scheme, by further new streaming and inputted data points, further actualized values of the self-similarity degree are calculated under the said windowed based calculating mode, wherein the finite number n has been selected in order to detect a (possibly presumed or predictable) suspected variation of actualized values within a "shifting" time interval wherein the variation and the time interval are deterministic for at least one kind of malign intrusion.