Dan Kaminsky points out one of Bitcoin’s greatest weaknesses: the potential for a bad actor to gain control of more than 50 percent of the network’s computing power. If that happened, the bad actor could engage in all kinds of mischief, including reversing his transactions and preventing others’ transactions from being confirmed.
As Bitcoin’s computing power moves to specialized computers known as ASICs, the universe of persons who control the network shrinks—thus making it easier that a pool of just a few miners (or potentially just one), could take control of the network. Kaminsky suggests changing Bitcoin’s proof of work math to be more amenable to solution by general purpose PCs, thus ensuring the expansion of the the number of persons who control the network’s computing power—even if that means that criminal botnets will get in on the action.
Jim Harper argues that Kaminsky’s doomsday scenario is not likely to transpire because anyone who abuses control of the network will only be making themselves worse off. After all, what’s the use of controlling a network that no one trusts? Harper points to the fact that the largest Bitcoin mining pool voluntarily restricts itself to less than 50 percent. But this serves to make Kaminsky’s point. The fact is that what began as a decentralized system now relies on the voluntary action of a few individuals to avoid a crisis of confidence.
While I acknowledge Kaminsky’s critique, I tend to agree with Harper that, in the near-term at least, the individual self-interests of those in a position to overtake the network keep them in check. And in the long term, this episode of centralization may prove to be but a blip in Bitcoin’s evolution. As Jeff Garzik, one of Bitcoin’s lead developers, has pointed out, the market is responding to the growing demand for ASICs, with new entrants producing more and cheaper machines. This will have the effect of swinging the pendulum in the other direction, back toward greater democratization of computing power.
Why not just switch to math problems better suited to general purpose CPUs, the most democratized kind of processing power there is? Because specialized ASICs provide greater hashing power, and as Garzik says, “More mining power makes it more difficult to reverse bitcoin transactions. The more widely spread that is, the more difficult it is to shut down bitcoin itself.” In the end the network may come to enjoy greater computational power and greater decentralization.
Jim wonders what will be the next shoe to drop. “Will it be a flaw in the technical structure that results in calamitously skewed incentives? Or will it be the opposition of governments, who see the threats in unfettered value transfer and seek to censor it?”
My fear is that it could be both. In a time of $2 billion NSA data centers, and billions more in secret cybersecurity budgets, it would be trivial for certain governments to engage in a 51 percent attack with the intent of disrupting the Bitcoin network. I don’t mean to be conspiratorial. Such an attack by the U.S. government, for example, would be unprecedented, and at $1 billion the Bitcoin economy hardly threatens the dollar. So it’s more than unlikely that the U.S. government would resort to such a drastic measure anytime in the foreseeable future. That said, the greater total hashing power of the Bitcoin network, the more costly it is for such an attack to succeed, which further argues for a large market in ASICs.