Generic Selfish Mining MDP for DAG Protocols

Patrik Keller

2024-04-30

Abstract:Selfish Mining is strategic rule-breaking to maximize rewards in proof-of-work protocols [3] and Markov Decision Processes (MDPs) are the preferred tool for finding optimal strategies in Bitcoin [4, 10] and similar linear chain protocols [12]. Protocols increasingly adopt non-sequential chain structures [11], for which MDP analysis is more involved [2]. To date, researchers have tailored specific attack spaces for each protocol [2, 4, 5, 7, 10, 12]. Assumptions differ, and validating and comparing results is difficult. To overcome this, we propose a generic attack space that supports a wide range of DAG protocols, including Ethereum, Fruitchains, and Parallel Proof-of-Work. Our approach is modular: we specify each protocol as one program, and then derive the Selfish Mining MDPs automatically.

Cryptography and Security,Distributed, Parallel, and Cluster Computing

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is to provide a general Selfish Mining attack space model for blockchain protocols based on the Directed Acyclic Graph (DAG) structure. Specifically, the author points out:

1. **Limitations of existing research**: Most of the existing Selfish Mining analyses mainly focus on linear - chain protocols such as Bitcoin, while the analysis of protocols adopting non - sequential chain structures (such as DAG structures), for example Ethereum, Fruitchains and Parallel Proof - of - Work, is more complex and difficult to carry out.

2. **Lack of a unified comparison benchmark**: Different researchers have customized specific attack spaces for each protocol with different assumptions, making the results difficult to verify and compare.

To solve these problems, the author proposes a **general attack space model**, which supports multiple DAG protocols and can automatically derive the Markov Decision Process (MDP) of Selfish Mining. In this way, the following goals can be achieved:

- **Unified assumptions**: Ensure that all protocols are analyzed under the same assumptions, thus facilitating fair comparison.
- **Modular design**: Encode the specific rules of each protocol into a program and then automatically generate the corresponding MDP, simplifying the support for new protocols.
- **High extensibility**: It can be easily extended to more DAG protocols and support the analysis of more security indicators.

### Formula representation

The formulas involved in the paper are mainly expressions related to probability and state transition. For example, the attacker's behavior choice can be represented by the following formulas:

- The attacker's mining ability $\alpha$ represents its proportion relative to the total network computing power.
- The attacker's communication advantage $\gamma$ represents its advantage in the propagation speed in the network.

The probability of state transition can be calculated in the following way:

\[ P(\text{State}_{t + 1}|\text{State}_t,\text{Action})=\begin{cases}
\alpha\cdot\gamma & \text{if the attacker spreads quickly and mines a new block}\\
\alpha\cdot(1 - \gamma) & \text{if the attacker spreads slowly and mines a new block}\\
(1 - \alpha)\cdot\gamma & \text{if the defender spreads quickly and mines a new block}\\
(1 - \alpha)\cdot(1 - \gamma) & \text{if the defender spreads slowly and mines a new block}
\end{cases} \]

These formulas are used to describe the probability that the system will transition to the next state given the current state and action.

### Summary

This paper proposes a general Selfish Mining attack space model, aiming to overcome the complexity of DAG protocol analysis and the incomparability of results in existing research. Through modular design and automated MDP derivation, this model can support multiple DAG protocols and provide a unified framework for future security analysis.