Decentralizing Sybil Detection

The story goes that a woman, given the pseudonym “Sybil”, had 16 different personalities living in one head. Her story gave rise to the concept of multiple personality disorders, and although it later turned out to be partly fictional, it fascinated many. Multiple personality disorder has since been a popular theme in many books and movies.

In the world of cryptocurrency and computer science, the name is now associated with the biggest danger to any reputation system, called a “Sybil attack.”

A Sybil attack involves an attacker creating a large number of fake accounts, which then interact with each other to generate the impression of real activity. The FBI recently indicted perpetrators of two of the largest Sybil attacks in history, earning a combined $36 million dollars by faking internet ad traffic using bots.

It’s also widely known that millions of social media accounts operate under the sponsorship of different countries, including Israel, the United States, and Russia, among others, and that these fake accounts constantly try to influence public opinion in order to achieve policy goals, like influencing the outcome of elections. This is a form of Sybil attack on a living social network.

As the Internet of Things and distributed ledgers continue to increase in importance, so will Sybil attacks.

Don’t be the green node. Never be the green node.

Of course, a problem of this magnitude has all kinds of bright minds working on it. Satoshi Nakamoto’s solution was Proof of Work, originally developed by Cynthia Dwork and first implemented in a payment protocol by Adam Back’s Hashcash. There is no point in creating a fake account in Bitcoin, because you can’t fake hashpower.

Many different defenses have been devised for protecting against Sybil attacks- all of them are based on different methods of viewing the behavior of nodes in a network in order to detect certain behaviors consistent with Sybil behavior, and each approach has various strengths and weaknesses.

Within a cryptocurrency network, there may be a simple defense strategy can combine the strengths of all the various approaches, at no cost to the network.

Make Sybil detection a side gig.

The Gig Economy

As companies seek to become more flexible and automation eliminates more and more jobs, there are fewer and fewer full time jobs, and people are looking for side gigs.

I saw one cafe in Indonesia that was filled with young men who would sit up after work surfing their phones, looking for content to post on Steemit, a blockchain based blogging platform. For several hours of work, they’d get about $1 worth of Steem, the native currency of the platform.

I can’t help but think there has to be a better way for people like this to use their time.

When we think about the monstrous size of the Bitcoin network (and it’s very likely going to keep getting bigger), it gives you and idea of just how much power there is in the creation of money. Satoshi Nakamoto wisely chose to base the network on something that cannot be faked, but from the perspective of those of us (Muslims, for example) that are less well endowed when it comes to computing resources, it would be a major breakthrough if we could somehow use human resources to secure a network and provide a similar incentive structure.

Each Sybil defense has different strengths and weaknesses, so a simple solution is to make all transactional data available (which is the default state of a public blockchain), and provide a reward for successfully detecting an attack. This puts the market to work, with users competing to find the best methods to quickly and efficiently detect attacks. Such a system would consist of judges, flaggers and defendants.

Decentralized Audits

This is inspired partly by the fascinating work of the Kleros team.

Don’t think that because the dynamic is decentralized that it means everything, by any means, needs to be on a blockchain- rather, the goal is always to get as much data as possible off of the blockchain.

The system functions more or less as follows:

1. The flagger, using either insider information, manual, or programmatic analysis, detects Sybil behavior and flags it.
2. The accuser makes a deposit which initiates a process which freezes the accused account.
3. A judge is pseudo-randomly selected according to reputation weight.
4. The plaintiff presents the evidence to the judge, and the defendant is given an opportunity to disprove the accusation.
5. The judge makes a decision. If the decision is not appealed, the deposit is returned to the flagger, whose reputation increases as a result. The reputation of the judge also increases.
6. If the defendant wishes to appeal the decision, the defendant account must make a deposit.
7. Two judges are pseudo-randomly selected according to the same metric, with the first judge being excluded from the pool.
8. If the decision is overturned, the deposit of the flagger goes to the defendant, and both the flagger and the first judge receive a reduction in reputation. If the decision is upheld, the deposit of the defendant goes to the flagger.
9. The appeal can be repeated with an additional judge at each iteration.

Such an arbitration system could be applied to other abuses of community standards, and would naturally generate a sort of “meritocracy,” where accounts with a higher degree of integrity would tend to gain influence.

If reputation is directly linked to the reward associated with the issuance of a currency, this increases the stakes considerably. Reputation translates directly to income, so people would be lining up to try to catch potential Sybil behavior and to judge accusations. In this way, not only is it possible to get a concrete monetary reward by defending the network, it’s also possible to increase earning power.

A Need for Alternatives

Both Proof of Work and Proof of Stake have some pretty serious drawbacks. If it is possible to develop a better alternative, there is without any doubt a moral imperative to do so.

The blockchain is not the biggest innovation brought to the world by Bitcoin- it’s the block reward. In it lies the key to directing a monetary system that rewards adhering to moral values, which is precisely the opposite of the effect of the present system.

I would welcome any criticism to this proposal.