Paper 2020/1290

FORTIS: Selfish Mining Mitigation by (FOR)geable (TI)me(S)tamps

Osman Biçer and Alptekin Küpçü

Abstract

Selfish mining (SM) attack of Eyal and Sirer (2018) endangers permissionless Proof-of-Work blockchains by allowing a rational mining pool with a hash power (a) much less than 50% of the whole network to deviate from the honest mining algorithm and to steal from the fair shares of honest miners. Since then, the attack has been studied extensively in various settings, for understanding its interesting dynamics, optimizing it, and mitigating it. In this context, Heilman (14) ''Freshness Preferred'', we propose a timestamp based defence if timestamps are not generated by an authority. To use this proposal in a decentralized setting, we would like to remove the timestamp authority, but due to two natural and simple attacks this turns out to be a non-trivial task. These attacks are composed of Oracle mining by setting the timestamp to future and Bold mining by generating an alternative chain by starting from a previous block. Unfortunately, these attacks are hard to analyze and optimize, and the available tools, to our knowledge, fail to help us for this task. Thus, we propose generalized formulas for revenue and profitability of SM attacks to ease our job in analysis and optimization of these attacks. Our analyses show that although the use of timestamps would be promising for selfish mining mitigation, Freshness Preferred, in its current form, is quite vulnerable, as any rational miner with a>0 can directly benefit from our attacks. To cope with this problem, we propose an SM mitigation algorithm Fortis with forgeable timestamps (without the need for a trusted authority), which protects the honest miners' shares against any attacker with a<27.0% against all the known SM-type attacks. By building upon the blockchain simulator BlockSim by Alharby and Moorsel (2019), we simulate our Oracle and Bold mining attacks against the Freshness Preferred and our Fortis defenses. Similar to our theoretical results, the simulation results demonstrate the effectiveness of these attacks against the former and their ineffectiveness against the latter.