A Performance Approach to Mobile Security

Signaling related attacks, such as the signaling storms, threaten the stability of mobile networks and are a frequent cause of multiple service degradations and even several network outages. This paper proposes a performance evaluation approach to the problem using techniques from probability theory, in general Markov Chains and queueing networks. It proposes a model of a mobile network under signaling storm, including more detailed models of the data channels and the data service times. Two classes of packets, also referred as "calls", are introduced in the model: normal and an attack ones, with the possibility of calls joining and leaving the network. The proposed model is used to provide insights on the effects of the storm on both network and user sides, such as the load on the signaling servers and the rate of blocking legitimate calls. Results show that the inactivity timer, set by the network operator, could be adjusted accordingly to lower the impact of such attacks and maintain higher network performance. Therefore, the paper further proposes a dynamic inactivity timer setup, which is represented as a function of the load on the signaling server. The proposed technique is introduced in the model and results show the improvement, as well as the significance of the timer selection. Such mechanism could be implemented in real mobile networks, as an improvement of the current static inactivity timer.