A Linked Data Quality Assessment Framework for Network Data

Abstract

For network analysts, understanding how traffic flows through a network is crucial to network management and forensics such as network monitoring, vulnerability assessment and defence. In order to understand how traffic flows through a network, network analysts typically access multiple, disparate data sources and mentally fuse this information. Providing some sort of automated support is crucial for network management. However, information about the quality of the network data sources is essential in order to build analyst's trust in automated tools. This paper presents SydNet, a novel Linked Data quality assessment framework which allows analysts to define quality dimensions and metrics which provide an accurate reflection of the quality of the data sources. The SydNet architecture also provides a number of novel fusion heuristics which can be used to fuse data from various network data sources. We demonstrate the utility of the SydNet architecture using CAIDA longitudinal topological data from a recent 24 months period and we demonstrate that our approach was able to detect dataset quality anomalies that would require further investigation.