Risk Assessment of Digital Assets – Insurance Applications in Cryptocurrencies and NFTs

Sammanfattning: The aim of the project is to develop a framework for an insurance policy for digital assets. The project comprised several stages, starting with the identification of risks associated with these assets. Policyholders were then categorized into two groups based on a predefined rating factor. Subsequently, data about previous thefts was gathered, two different approaches were explored. In the first approach, patterns in cyberattacks targeting the selected assets are detected based on the risks previously identified, and a Python script is developed to automate the whole process. However, practical limitations surfaced, impeding the success of this approach, therefore a decision was made to pursue an alternative strategy for data collection, involving manual retrieval from trusted sources. The collected data was used to fit various statistical distributions, enabling the prediction of the probability of policyholders experiencing loss of their digital assets. Additionally, a mathematical model was developed to provide a one-step forecast of the tokens prices, incorporating variables such as the floor price and token rarity. These predictions formed the basis for estimating the expected losses on a daily basis, which are utilized to calculate the company's potential liabilities. A real-world scenario was simulated, where a user takes out an insurance policy to cover the risks for one of their items during the month of April 2023. The lump expected losses are calculated at the end of the month, assuming a daily exchange of money between both parties, and the final value is compared for both groups of policyholders. Furthermore, an alternative approach was proposed, introducing a supplementary variable to the model based on the policyholder's behavior. The findings demonstrated consistency, as the expected losses fell within a reasonable range, with higher premiums for the riskier group of policyholders. However, it was observed that at a certain point, the perceived risk became higher for the safer group. Therefore, it is suggested to dynamically adjust the calculated parameters for the statistical distributions, taking this factor into account. This pricing model serves as a preliminary framework for insurance policies and can be further refined through iterative improvements by incorporating historical claims data gathered by the insurance company. Ultimately, these enhancements aim to develop a comprehensive insurance policy offering. This Master's thesis was written in collaboration with Trygg-Hansa through the Faculty of Engineering at Lund University.