Summary: It should be possible to analyze the connectivity of members in a p2p credit network, and from this to determine which members are real and which are Sybils (fake or duplicate accounts).
Here is a simple, powerful idea: let’s build an open-source identity network, a social graph that does not require any personal information besides your connections to other real people. This network could be the basis for a decentralized system to probabilistically prove whether each member of the network is a real, unique human.
Such a feat is possible if we use a graph analysis of our social connections, with the additional input of a limited number of trusted seed identities. The interconnectivity of the graph should reveal which members are real and which are fake (aka “Sybils”), based on their position in relation to the trusted seeds. This is exactly the goal of the project known as BrightID.
The creators of BrightID additionally came up with a modified version of the well-known “SybilRank” algorithm, which they refer to as “SybilGroupRank” (M. Heydari, M. Khanmohammadzade, R. Bakhshandeh, & C.A. Stallard; publication forthcoming). This modification takes into account the interconnectivity of social groups, rather than just individual social connections. In its current implementation, you may join a “group” only if you are personally connected to over half the group’s members. Joining a group is akin to saying “I may not personally know everyone in this group, but I trust the good judgment of the people that I do personally know in this group, and therefore I believe that everyone in this group is a real, unique human.” The SybilGroupRank algorithm performs well at detecting Sybils in a variety of random social graphs.
Notably, the BrightID identity network is an open source public good, distinct from the (also open source) SybilGroupRank algorithm. That leaves the identity system open for innovation: Anyone can access the data and run their own Sybil-detection algorithm on the social graph, using their own hand-picked trusted seeds.
The weak point of such an identity network is this: How do we properly incentivize connections? What should compel users to connect to a friend, or not to connect to a stranger?
First of all, we should not entirely discount the possibility that people may simply act with reasonably good judgment and honesty in the aggregate. If we can communicate the gravity of honest participation in a robust identity system, then “connecting with caution” should be well within the average person’s capabilities. And after all, being a Sybil takes some amount of dedication, including lying to friends or group dishonesty if we assume collusion: all which is shameful and highly risky social behavior, and therefore unlikely from a psychological perspective. (How would you react if you realized a friend of yours was using you to commit voter fraud?) In addition, collusion becomes exponentially more unstable as dishonest groups grow larger.
Nevertheless, what if we don’t want to rely on the good judgment and honesty of the majority? The trustless answer may lie in peer-to-peer (p2p) credit networks (i.e. p2p lending systems, or an interest-free mutual credit or hawala system – see Ryan Fugger’s 2004 paper, the original idea behind Ripple). In a p2p credit system, participants extend lines of credit to those whose reputation they trust. This may be a good approximation for unique identity: most people are unlikely to extend a credit line to a stranger.
However, equally important is whether you would be sufficiently likely to extend a credit line to a friend. This hinges on whether p2p credit systems are useful in the life of the average person. But the question here is not how much people would lend each other, whether they would charge interest, or how reliably they would pay it back. The critical question seems to be whether they would make such a connection at all, to any degree.
There is evidence that a p2p credit system would be useful in many contexts. Microcredit or p2p lending platforms like Kiva Zip or Zidisha have garnered considerable press. Meanwhile, the interest-free LETS or “local exchange trading system” model has inspired several community currencies over the years. I have high hopes for projects like Trustlines Network, an implementation of mutual credit based on Fugger’s idea, and far more scalable than LETS. Notably, it is possible to reach high levels of liquidity based simply on p2p mutual credit. (This is an important point, as liquidity barriers tend to be a central concern among skeptics of the p2p currency model.)
Going even further, imagine if we could change people’s perception of the definition of debt. Sometimes I describe the Circles UBI project as a mutual credit system masquerading as a universal basic income (UBI) system. What if that’s exactly what we need? How would it change people’s relationship to mutual credit if we look at negative balances not as debts to be repaid, but instead as a basic income? In a UBI-as-mutual-credit system, you might be more likely to “lend” to a friend, even if you thought that friend might not pay you back. It may seem to go against common sense, but most economists will recognize that some level of debt reflects economic growth, and arguably must not be repaid (or we risk a regressive, deflationary monetary system). In this light, you could call such debts “deficit spending by the people”, a logically sustainable monetary system.
So we see that a system of p2p credit could be enormously useful in itself: both from the traditional viewpoint as a medium of exchange via p2p lending, but also framed as bottom-up deficit spending through (interest-free) mutual credit.
How does this lead back to identity? Simple. Imagine if we could draw the connectivity data from a system of p2p credit and use this graph as the basis for an identity network. Enabling credit lines is a decentralized way to organically incentivize trusted connections. To reiterate: you do not need a centralized identity register for a p2p credit system, because each person decides for herself whether or not to extend credit to her peers. These connections could ultimately lay the foundation for a decentralized, privacy-conscious, Sybil-proof identity system.