The federation of our digital identities
Facebook, Twitter, email, WordPress, Instagram, online banking, the list goes on… Offline, you’re one person maintaining (presumably) one identity. On the web, you have many of them. All of them might point at you, but they’re still distinct packets of data floating through different websites. Within each site, your identity is unified, but between them, you’re different people. For example, I can’t log into Twitter with my Facebook username/password because Facebook owns them. When digital information becomes federated like this, it drives down cross-network accountability because my identity doesn’t move around.
However, there are some popular exceptions to this. Facebook and Twitter don’t exchange my log-in credentials – the keys with which I unlock my identity – because they’re rivals, but many other services and these sites are not. For example, I can log into my YouTube account using my GMail credentials. When I hit ‘Submit’, YouTube banks on the validity of my identity on GMail to log me in. Suddenly, GMail and YouTube both have access to my behavioral information through my username now. In the name of convenience, my online visibility has increased and I’ve become exposed to targeted advertising, likely the least of ills.
The Crypto-Book
John Maheswaran, a doctoral student at Yale University, has a solution. He’s called it ‘Crypto-Book’, describing its application and uses in a pre-print paper he and his colleagues uploaded to arXiv on June 16.
1. The user clicks ‘Sign up using Facebook’ on StackOverflow.
2. StackOverflow redirects the user to Facebook to log in using Facebook credentials, 3. after which the user grants some permissions.
4. Facebook generates a temporary OAuth access token corresponding to the permissions.
5. Facebook redirects the user back to StackOverflow along with the access token.
6. StackOverflow can now access the user’s Facebook resources in line with the granted permissions.
Crypto-Book sits between steps 1 and 6. Instead of letting Facebook and StackOverflow talk to each other, it steps in to take your social network ID from Facebook, uses that to generate a username and password (in this context called a public and private key, respectively), and passes them on to StackOverflow for authentication.
OpenID and OAuth
It communicates with both sites using the OAuth protocol, which came into use in 2010. Five years before this, the OpenID protocol had launched to some success. In either case, the idea was to reduce the multiplicity of digital identities but in the context of sites like Facebook and Twitter that could own your identities themselves, the services the protocols provided enabled users to wield more control over what information they shared, or at least keep track of it.
OpenID let users to register with itself, and then functioned as a decentralized hub. If you wanted to log into WordPress next, you could do so with your OpenID credentials; WordPress only had to recognize the protocol. In that sense, it was like, say, Twitter, but with the sole function of maintaining a registry of identities. Its use has since declined because of a combination of its security shortcomings and other sites’ better authentication schemes. OAuth, on the other hand, has grown more popular. Unlike OpenID, OAuth is an identity access protocol, and gives users a way to grant limited-access permissions to third-party sites without having to enter any credentials (a feature called pseudo-authentication).
So Crypto-Book inserts itself as an anonymizing layer to prevent Facebook and StackOverflow from exchanging tokens with each other. Maheswaran also describes additional techniques to bolster Crypto-Book’s security. For one, a user doesn’t receive his/her key pair from one server but many, and has to combine the different parts to make the whole. For another, the user can use the key-pair to log in to a site using a technique called linkable ring sgnatures, “which prove that the signer owns one of a list of public keys, without revealing which key,” the paper says. “This property is particularly useful in scenarios where trust is associated with a group rather than an individual.”
The cryptocurrency parvenu
Interestingly, the precedent for an equally competent solution was set in 2008 when the cryptocurrency called bitcoins came online. Bitcoins are bits of code generated by complex mathematical calculations, and each is worth about $630 today. Using my public and private keys, I can perform bitcoin transactions, the records of which are encrypted and logged in a publicly maintained registry called the blockchain. Once the blockchain is updated with a transaction, no other information except the value exchanged can be retrieved. In April 2011, this blockchain was forked into a new registry for a cryptocurrency called namecoin. Namecoins and bitcoins are exactly the same but for one crucial difference. While bitcoins make up a decentralized banking system, namecoins make up a decentralized domain name system (DNS), a registry of unique locations on the Internet.
The namecoin blockchain, like its website puts it, can “securely record and transfer arbitrary names,” or keys, an ability that lets programmers use it as an anonymizing layer to communicate between social network identities and third-party sites in the same way Crypto-Book does. For instance, OneName, a service that lets you use a social network identity to label your bitcoin address to simplify transactions, describes itself as
a decentralized identity system (DIS) with a user directory made of entries in a decentralized key-value store (the Namecoin blockchain).
Say I ‘register’ my digital identity with namecoin. The process of registration is logged on the blockchain and I get a public and private key. If Twitter is a relying partner, I should be able to log in to it with my keys and start using it. Only now, Twitter’s server will log me in but not itself own the username with which it can monitor my behavior. And unlike with OpenID or OAuth, neither namecoin or anyone on the web can access my identity because it has been encrypted. At the same time, like with Crypto-Book, namecoin will use OAuth to communicate with the social networking and third-party sites. But at the end of the day, namecoin lets me mobilize only the proof that my identity exists and not my identity itself in order to let me use services anonymously.
If everybody’s wearing a mask, who’s anonymous?
As such, it enables one of the most advanced anonymization services today. What makes it particularly effective is its reliance on the blockchain, which is not maintained by a central authority. Instead, it’s run by multiple namecoin users lending computing resources that process and maintain the blockchain, so there’s a fee associated with staking and sustaining your claim of anonymity. This decentralization is necessary to dislocate power centers and forestall precipitous decisions that could compromise your privacy or shut websites down.
Services like IRC provided the zeroth level of abstraction to achieve anonymity in the presence of institutions like Facebook – by being completely independent and ‘unhooked’. Then, the OpenID protocol aspired, ironically, to some centrality by trying to set up one set of keys to unlock multiple doors. In this sense, the OAuth protocol was disruptive because it didn’t provide anonymity as much as tried to provide an alternative route by limiting the number of identities you had to maintain on the web. Then come the Crypto-Book and blockchain techniques, both aspiring toward anonymity, both reliant on Pyrrhic decentralization in the sense that the power to make decisions was not eliminated as much extensively diluted.
Therefore, the move toward privatization of digital identities has been supported by publicizing the resources that maintain those identities. As a result, perfect anonymity becomes consequent to full participation – which has always been the ideal – and the size of the fee to achieve anonymity today is symptomatic of how far we are from that ideal.
(Thanks to Vignesh Sundaresan for inputs.)
