/Distributed-Privacy-Focused-Driver-Scoring-System

Prototype for my Bachelor's Thesis

Primary LanguageJavaGNU General Public License v3.0GPL-3.0

Prototype for my Bachelor's Thesis

Disclaimer:

This is a prototype under construction. You are (within the boundaries of the gnu general public license) free to do what ever you want with this code.

Contents:

  • /android_app contains a prototype app written in java
  • /obd2_sim contains an arduino based simulation tool for sending data via can-bus to an obd2 bluetooth-dongle
  • server_linux contains a python3 implementation of the server-component that manages a localy deployed burrow blockchain

Thesis proposal:

Introduction

In the automotive insurance industry there are currently various approaches to minimise the risk of bad driver behaviour. Data related to some characteristics of the individual driver are gathered through multiple ways, including inter- views, standard forms, etc. This data is then used to calculate the risk for insuring a specific person. An example would be that the insurant gets better conditions when he/she drives crash and incident free for a specific time period. Another approach at gathering individual information is to monitor the driving behaviour of a person while driving. Pricing the insurant on that basis min- imises the cost for the insurance company overall. According to [7] this is called ”usage-based pricing”.

State of the Art and Problems

Most current driver scoring systems are put in place on a small scale and orig- inate from the insurance companies themselves or contracted third parties [4]. This leads to several problems: First, there is always an information asymmetry between the insurant and the insurance. The consequence is that the insurance could withhold information, that might either not be profitable for the com- pany or even be beneficial for the driver, which might lead to unfair treatment towards the insurant. Secondly the insurance is gathering all the data and has no reason to share it with the insurant, his/her attorneys or other parties the insurant wants to share his information with. This puts the driver in subordi- nation to the insurance. Arising from these problems the mentioned systems fail to provide a decentralised and global way of sharing trusted information between parties with different interests, thus making it nearly impossible for two insurance companies to fight over a case, based on the same and trusted information.

Proposed Solution

How would one design a system that accounts for the problems mentioned above? To approach this question, let us consider a view different questions first: What measures could be taken to ensure immutability of data? Could we design a system that enables the driver to control access to his data? Is there a technology that could be the basis of all that? At first one would need to make sure that the capturing device is ”tamper-proof” and that captured data cannot be altered. This could be achieved by working with car manufacturers to develop a built-in system that uses information pro- vided by the on board computer. Secondly the data has to be stored in a place that is controlled by a third party with no specific interest in the data (neither the insurance, nor the insurant). This data would then need to be associated with the driver in some way. State of the art research and development on blockchains raise a solution to all the questions above. With the right implementation it is possible to have information on driving behaviour tied to the driver him-/herself. This enables the driver to share his information with an insurance company and other parties of interest (e.g.: car rental service, authorities, etc.). The information asymmetry could be overcome by designing a system that stores most of the gathered data on a blockchain. This makes sure that neither the insurant nor the insurance company can alter the stored information. At last the problem of sharing information in a decentralised way and on a global scale solves itself by the nature of blockchain technology. This could increase competi- tion between insurance companies on a global scale and be profitable for drivers. The proposed system allows us to imagine even more useful applications. For example, companies could start to monitor the behaviour of professional drivers, or with the use of smart contracts it is possible to tie the insurance contract not to a car, but solely to a person, enabling the possibility for easier and cheaper car sharing.

Placement of the Thesis

Currently there is no system which accounts the issues of protecting the driver and his/her data as well as immutability of data that might be relevant to the insurance. Another advantage of the proposed system using blockchain technol- ogy is better and easier interoperability among different parties of interest, e.g.: different car manufacturers and insurance companies. Driver scoring is nothing new and there is some research on the topic to be found e.g.: [3] or [2]. There is however no solution that would match this specific scenario, so tailoring of those solutions would be necessary. Currently blockchain technology is at a state where most applications with a fair amount of complexity are not viable for an enterprise solution. Thus the task at hand is more a proof-of-concept. The proposed thesis would include a software prototype with all described func- tionality that would run on a smart phone and make use (where possible) of the OBDII(On-board diagnostics two) port of the car. The proposed ”tamper- proof” hardware device is explicitly not part of this thesis. To evaluate the driver scoring part of the proposed system one would need a OBDII-port Bluetooth dongle (10-15don Amazon), as well as simulation- software/-hardware for the port. There is a software solution [1] that seems nearly good enough for the purpose. All the Hardware emulators are quite ex- pensive (at least 180d[6] and up to 2000d[8]), but would probably better suite the purpose. The goal would be to use open data sets [5] that could be replayed with the simulator in order to reliably test the system under specific conditions. The development of this evaluation system would take place before the thesis in a research project. To evaluate the viability of the whole system it is pro- posed to do stress testing of the whole system. For example there could be test cases with up to 1.000.000 cars. Thereby one could evaluate the capabilities in performance of the system.

Conclusion

Taking a step back we can clearly see the bigger picture now: The proposed solution aims to solve problems in the insurance and car industry in a new fashion, hopefully leading to innovation and progress in these fields. All in all this thesis could take a step towards decentralisation and help working towards fair treatment of customers.

References