tl;dr the best solution is simply not reward any of the competitive blocks. Meaning, miners will not be rewarded for blocks that have competitive orphan blocks.

Hi all,

I'm Lear, and I have discovered and analyzed the Block Discarding Attack, which is a generalization of the "Selfish Mining" strategy, independently (and actually before) Eyal and Sirer published their paper. You can find here my draft paper, in which I present a more comprehensive analysis of the attack and its possible countermeasures: http://eprint.iacr.org/2013/868.pdf. You might as well read the bitcointalk posts I wrote in response to Eyal and Sirer misleading paper: https://bitcointalk.org/index.php?topic=325824.msg3494144#msg3494144, https://bitcointalk.org/index.php?topic=325824.msg3504846#msg3504846.

In this post I would like to be focused on the possible countermeasures, yet I must note that the Block Discarding Attack (including the "selfish mining" strategy) is not applicable to pools, and only solo miners can perform it (Please read my paper or the first of my two posts for explanations). Thus, the current threat of the Bitcoin network is negligible. Nevertheless the currently theoretic attack might one day become feasible, in case a solo miner gains a relatively high fraction of the total hash-power. This is not completely unlikely scenario, as the arrival of new ASICs changes the distribution of hash-power. So I do believe that a countermeasure should be introduced.

Fortunately, there is a very simple change to the core protocol, that can be shown to make the attack unreasonable unless the attacker's fraction of the total hash-power is *very* close to 50%. Obviously, that is as far as we can get. The solution is to introduce a fork-punishment mechanism, as explained in the paper and in my first post. Basically, I suggest that when there are two competitive blocks, no matter which one of them is eventually confirmed, both the confirmed and the unconfirmed blocks will not reward their miners. A detailed possible implementation, incentive-oriented, is presented in my paper.

Lear.