Secure real-time communications over 5G in an autonomous vehicle platooning scenario

Abstract

In 5G networks, gNodeB, periodically broadcast system information messages, including network identifiers, to enable the connection of user equipment (UE) to the network. Similar to previous generations, 5G system information messages are delivered in clear text without encryption, letting an adversary deceive a legitimate gNodeB.

The 5G has improved Rogue Base Station (RBS) detection in terms of subscription permanent identifier concealment, while other security mechanisms passed down to 5G security from previous generations include mutual authentication between UE and network, secure algorithm negotiations, and integrity-protected signalling. Despite these advances, the current position is that 5G remains vulnerable to RBS attacks. A rogue agent can send a signal with enough power to deceive the UE, causing the UE to connect to the Rogue and potentially leaking confidential information.

The scope of the thesis is restricted to the radio access network layer and describes a novel simulation of a 5G cellular system to generate a dataset of signal strength measurements. It also proposes an experimental simulation and then describes the development of a platform for the generation of realistic measurement reports.

The design, implementation, and evaluation of a network-based RBS detection system based on the filtering of measurement reports are presented. It proposes a novel architecture model and introduces the notion of a BS Probation Period, during which a BS is excluded from consideration for handover while its legitimacy is determined. This is supported by an enhanced handover protocol to facilitate the rejection of suspected rogue BS. It also proposes a BS Trust Analysis component to implement the RBS detection, initially using a rule-based approach.

This final contribution proposes an AI technique for RBS identification, called RBS-MLP. The performance results of the RBS-MLP model demonstrate that the ML approach is 99.999% accurate and provides a new standard procedure for RBS identification.

Date of Award	May 2023
Original language	English
Sponsors	BT
Supervisor	Philip Perry (Supervisor), Adrian Moore (Supervisor), Christopher Nugent (Supervisor), Ian Herwono (Supervisor) & Maria Cuevas (Supervisor)