Online Archive of University of Virginia Scholarship
Latency Makes You Look Honest: Fully Undetectable Selfish Mining in Bitcoin11 views
Author
Vejalla, Varun, Computer Science - School of Engineering and Applied Science, University of Virginia
Advisors
Xavier Ferreira, Matheus, EN-Comp Science Dept, University of Virginia
Abstract
Eyal and Sirer [2014] showed that a strategic Bitcoin miner can increase their share of Bitcoin rewards by deviating from the intended protocol through a selfish mining strategy. However, their strategy is statistically detectable, as there are unusually long forks which a system with natural network delays wouldn't produce. The exchange rate hypothesis has been put in place to justify why Selfish Mining does not occur in practice. Under this hypothesis detectable deviations would harm the price of Bitcoin. While detectable attackers may get a higher payoff in terms of BTC, it may be a lower payoff in terms of USD.
Bahrani and Weinberg [2024] have demonstrated that selfish mining strategies that are undetectable when looking at the shape of the blockchain are available. We take this further by showing that even in the presence of global clocks and precise measurements of end-to-end network delays, there is an undetectable selfish mining strategy. We accomplish this by showing that if the adversary adds a small delay to all messages it sends or receives, a sufficiently large miner would obtain unfair profits. This is surprising since our strategy is passive in the sense that miners are not adaptive and commit to time to release or receive blocks.
We develop a selfish mining strategy that is provably statistically undetectable under a realistic network model: block generation follows a Poisson process, blocks propagate with pre-defined latency distributions, and forks are resolved arbitrarily by miners. We show that an attacker who controls more than one-third of the total hash power can strictly increase their long-run share of blocks in the main chain while producing a distribution of blocks that is indistinguishable from that of an honest miner with slightly worse latency. Moreover, no augmentation of per-block metadata (headers, timestamps, signatures, etc.) can prevent this advantage in our model. Our results suggest that protocol changes limited to block metadata are insufficient to eliminate profitable, undetectable selfish mining under realistic network delays.
Vejalla, Varun. Latency Makes You Look Honest: Fully Undetectable Selfish Mining in Bitcoin. University of Virginia, Computer Science - School of Engineering and Applied Science, MS (Master of Science), 2025-12-12, https://doi.org/10.18130/eehr-am14.