Vulnerability Assessment and Classification based on Influence Metrics in Mobile Social Networks

Abstract

In emerging 5G wireless systems, Mobile Social Networks (MSN) will play an important role for providing data services and offloading data traffic from cellular networks. MSNs are vulnerable to various security attacks because of the ways users move and collaborate. Since most protocols for MSNs are designed based on social behaviors of users, it is important to understand the impact of user behavior on network vulnerability. This can provide valuable insights into crucial factors, such as how easily a network loses its connectivity, or a network's ability to form strong communities. We present a novel vulnerability assessment and classification scheme based on structural, social and influence distribution metrics in mobile social networks. We design a vulnerability index metric (VI) to investigate the level of damage inflicted when networks are subjected to both targeted and random attacks. Then, we use influence distribution metrics and various machine learning based classifiers to determine the vulnerability levels for various network states. Finally, we define a Mean Information Diffusion index to determine the information dissemination capability of a network, given the vulnerability state. Our results revealed that campus WLAN traces, represented by the Time Variant Community model, exhibit highly vulnerable states that reduce the network's ability to disseminate information by up to 16%.