Graph theoretical approach for automated IP lifecycle management in telco networks

Abstract

Summary IP address management (IPAM) systems are commonly used in the telecommunication industry to serve resource management for large‐scaled IP networks. In resource management, IPAM provides network operators a live IP lifecycle support with an easier use on IP inventory while providing the scaling and high performance. Throughout the management of IP lifecycle, having up‐to‐date topology information plays a key role because of the rapidly changing usage of the IP addresses. Furthermore, if the usage information which corresponds to data taken from the real topology for a specific block does not match with the information in planned state, there arises an anomaly situation. And if this anomaly is not taken under control, especially for the large‐scaled networks, it causes a considerable increase in the operational cost (OPEX) of the internet service providers (ISPs). Thus, keeping track of this resource and checking the anomaly cases make a smooth functioning process through the lifecycle of an IP block in telecom applications. However, the techniques used in traditional IPAMs are mostly not covering track of IP usage. For this purpose, we propose an attribute‐based IP management module (KIP), which consists of three engines by covering the three active states of the IP lifecycle. In this way, KIP serves efficient IP management with minimum traversal time, anomaly rate, and planning time. Compared to previous methods, our method has addressed three additional challenges that significantly impact an effective IP lifecycle management by regarding device‐specific information, such as connectivity, anomaly situation, and planning.

Network-Aware AutoML Framework for Software-Defined Sensor Networks

“…O. Rahman et al evaluated J48, Random Forest (RF), SVM, and K-NN algorithms to detect and block the DDoS attack in an SDN network and the results showed that J48 has better performance than others [17]. K. Duran et al proposed an automated attribute-based IP management module with the three engines and they succeeded in a more accurate anomaly detection method by decreasing the anomaly scoring rate compared to IPAMv1 [18]. SVM, Naive Bayes, Decision Tree, and Logistic Regression techniques were evaluated to detect DDoS attacks in SDN and the results showed that SVM has better performance [19].…”