Attacking cryptocurrencies with deep reinforcement learning

DegreeBachelor
StatusAvailable
Supervisor(s)Patrik Keller, MSc

Description

Selfish mining is a family of attacks that threaten the fairness and the security of proof-of-work cryptocurrencies such as Bitcoin. Previous results on optimal attack strategies are restricted to relatively simple protocols and artificial assumptions, e.g., with regards to the network connectivity of the attacker.

The objective of this thesis is to reproduce and potentially extend upon recent results in this line of research. The student working on this topic will be provided with an OpenAI Gym environment that simulates the execution of advanced blockchain protocols under realistic network conditions. The student will apply deep reinforcement learning tools to find optimal attack strategies in the provided environment.

This thesis is co-supervised by Jakob Hollenstein.

Prerequisites

Practical experience in machine learning, ideally reinforcement learning

References

  • Eyal, I. and Sirer, E.G. Majority Is Not Enough: Bitcoin Mining Is Vulnerable. In N. Christin and R. Safavi-Naini, eds., Financial Cryptography and Data Security. Lecture Notes in Computer Science 8437, Springer, Berlin Heidelberg, 2014, pp. 436–454.
  • Gervais, A., Karame, G.O., Wüst, K., Glykantzis, V., Ritzdorf, H., and Capkun, S. On the Security and Performance of Proof of Work Blockchains. In Proceedings of the ACM Conference on Computer and Communication Security (CCS). 2016, pp. 3–16.
  • Sapirshtein, A., Sompolinsky, Y., and Zohar, A. Optimal Selfish Mining Strategies in Bitcoin. In J. Grossklags and B. Preneel, eds., Financial Cryptography and Data Security. Lecture Notes in Computer Science 9603, Springer, Berlin, Heidelberg, 2016, pp. 515–532.
  • Hou, C., Zhou, M., Ji, Y., et al. SquirRL: Automating Attack Analysis on Blockchain Incentive Mechanisms with Deep Reinforcement Learning. In Network and Distributed System Security Symposium (NDSS). 2021.