Towards secure, privacy-preserving, and trusted analytic about ad hoc network data

Abstract

The boom of Big Data and Internet of Things (IoT) trends induces lots of hot topics including data security, users privacy, devices connectivity and usage of analytic. We put our focus on security and privacy of ad hoc network data. Ad hoc network can be applied in many different domains, especially vehicular network. Nowadays, even though the hardware capability of ad hoc network is ready, the design of systems are not mature enough when compared to traditional wired network. This creates the uncertainty of whether additional problems may arise in the security, privacy and trust in ad hoc network.

In this thesis, the diversified cases of the ad hoc network are summarized to a general ad hoc data analytic framework. We apply this framework to the vehicular network domain. Three major problems are defined: 1) Attacks on batch verifications, 2) Single Point of Failure (SPOF) problem and 3) Anonymous Counting Problem from Vehicular ad hoc network (VANET). We provide novel attacks and solutions for these problems. This study makes the design of ad hoc network data analysis systems move towards to the directions of secure, privacy-preserving and trusted analytic. Fine-grained data are becoming more reliable in the analysis, and it gives higher chance to produce accurate and meaningful analysis results.

The proposed solution for 1) False acceptance on batch verifications allows an adversary to pass verifications on pseudo identities and signatures. We show that the attacks are valid for some previous work. We propose a solution with randomized property to ensure the authenticity of the vehicular messages as well as protecting the privacy of drivers. Our proposed solution is secure and the additional overhead is small.

The proposed solution for 2) SPOF problem allows vehicles to broadcast vehicular data to adjacent vehicles as another backup. No specific node becomes the only connecting point of the system. This scheme enables trusted authority to reveal those data from recipients. We ensure the integrity and confidentiality of vehicular messages as well as the privacy of drivers. We show that our scheme is secure and the performance is feasible through probability analysis and network simulation. We also propose collision avoidance system as an example.

The proposed solution for 3) Anonymous Counting problem achieves conditional anonymity to avoid over-counting and applies flexible time-slotted pseudo identities to minimize the influence of the adversary. The major technical difficulty of designing this scheme is to avoid using any same hash identity digest to achieve high privacy standard but still allow counting at the same time. We explain how our design, which adopts the suggested trust level calculation methods in other publications, achieves a more meaningful trust level calculation. We also explain how to set the time slot and threshold by trust level performance simulation.