A protocol for aggregating digital identity information that's broadly interoperable with existing, proposed, and hypothetical future digital identity standards.
This EIP proposes an identity management and aggregation framework on the Ethereum blockchain. It allows entities to claim an `Identity` via a singular `Identity Registry` smart contract, associate it with Ethereum addresses in a variety of meaningful ways, and use it to interact with smart contracts. This enables arbitrarily complex identity-related functionality. Notably (among other features) ERC-1484 `Identities`: are self-sovereign, can natively support [ERC-725](https://github.com/NoahZinsmeister/ERC-1484/tree/master/contracts/examples/Resolvers/ERC725) and [ERC-1056](https://github.com/NoahZinsmeister/ERC-1484/tree/master/contracts/examples/Resolvers/ERC1056) identities, are [DID compliant](https://github.com/NoahZinsmeister/ERC-1484/blob/master/best-practices/DID-Method.md), and can be fully powered by [meta-transactions](https://github.com/NoahZinsmeister/ERC-1484/tree/master/contracts/examples/Providers/MetaTransactions).
Emerging identity standards and related frameworks proposed by the Ethereum community (including ERCs/EIPs [725](https://github.com/ethereum/EIPs/issues/725), [735](https://github.com/ethereum/EIPs/issues/735), [780](https://github.com/ethereum/EIPs/issues/780), [1056](https://github.com/ethereum/EIPs/issues/1056), etc.) define and instrumentalize digital identity in a variety of ways. As existing approaches mature, new standards emerge, and isolated, non-standard approaches to identity develop, coordinating on identity will become increasingly burdensome for blockchain users and developers, and involve the unnecessary duplication of work.
The proliferation of on-chain identity solutions can be traced back to the fact that each codifies a notion of identity and links it to specific aspects of Ethereum (claims protocols, per-identity smart contracts, signature verification schemes, etc.). This proposal eschews that approach, instead introducing a protocol layer in between the Ethereum network and individual identity applications. This solves identity management and interoperability challenges by enabling any identity-driven application to leverage an un-opinionated identity management protocol.
-`Identity Registry`: A single smart contract which is the hub for all `Identities`. The primary responsibility of the `Registry` is to define and enforce the rules of a global namespace for `Identities`, which are individually denominated by Ethereum Identification Numbers (EINs).
-`Identity`: A data structure containing all the core information relevant to an identity, namely: a `Recovery Address`, an `Associated Addresses` set, a `Providers` set, and a `Resolvers` set. `Identities` are denominated by EINs (incrementing `uint` identifiers starting at 1), which are unique but otherwise uninformative. Each `Identity` is a Solidity struct:
-`Associated Address`: An Ethereum address publicly associated with an `Identity`. In order for an address to become an `Associated Address`, an `Identity` must either transact from or produce an appropriately signed message from the candidate address and an existing `Associated Address`, indicating intent to associate. An `Associated Address` can be removed from an `Identity` by transacting/producing a signature indicating intent to disassociate. A given address may only be an `Associated Address` for one `Identity` at any given time.
-`Provider`: An Ethereum address (typically but not by definition a smart contract) authorized to act on behalf of `Identities` who have authorized them to do so. This includes but is not limited to managing the `Associated Address`, `Provider`, and `Resolver` sets for an `Identity`. `Providers` exist to facilitate user adoption by making it easier to manage `Identities`.
-`Resolver`: A smart contract containing arbitrary information pertaining to `Identities`. A resolver may implement an identity standard, such as ERC-725, or may consist of a smart contract leveraging or declaring identifying information about `Identities`. These could be simple attestation structures or more sophisticated financial dApps, social media dApps, etc. Each `Resolver` added to an `Identity` makes the `Identity` more informative.
-`Recovery Address`: An Ethereum address (either an account or smart contract) that can be used to recover lost `Identities` as outlined in the [Recovery](#recovery) section.
-`Destruction`: In the event of irrecoverable loss of control of an `Identity`, `Destruction` is a contingency measure to permanently disable the `Identity`. It removes all `Associated Addresses`, `Providers`, and optionally `Resolvers` while preserving the `Identity`. Evidence of the existence of the `Identity` persists, while control over the `Identity` is nullified.
A digital identity in this proposal can be viewed as an omnibus account, containing more information about an identity than any individual identity application could. This omnibus identity is resolvable to an unlimited number of sub-identities called `Resolvers`. This allows an atomic entity, the `Identity`, to be resolvable to abstract data structures, the `Resolvers`. `Resolvers` recognize `Identities` by any of their `Associated Addresses`, or by their `EIN`.
The protocol revolves around claiming an `Identity` and managing `Associated Addresses`, `Providers` and `Resolvers`. Identities can delegate much or all of this responsibility to one or more `Providers`, or perform it directly from an `Associated Address`. `Associated Addresses`/`Providers` may add and remove `Resolvers` and `Providers` indiscriminately. `Associated Addresses` may only be added or removed with the appropriate permission.
The `Identity Registry` contains functionality to create new `Identities` and for existing `Identities` to manage their `Associated Addresses`, `Providers`, and `Resolvers`. It is important to note that this registry fundamentally requires transactions for every aspect of building out an `Identity`. However, recognizing the importance of accessibility to dApps and identity applications, we empower `Providers` to build out `Identities` on the behalf of users, without requiring users to pay gas costs. An example of this pattern, often referred to as a meta transactions, can be [seen in the reference implementation](https://github.com/NoahZinsmeister/ERC-1484/tree/master/contracts/examples/Providers/MetaTransactions).
Due to the fact that multiple addresses can be associated with a given identity (though not the reverse), `Identities` are denominated by `EIN`. This `uint` identifier can be encoded in QR format or mapped to more user-friendly formats, such as a `string`, in registries at the `Provider` or `Resolver` level.
The address management function consists of trustlessly connecting multiple user-owned `Associated Addresses` to an `Identity`. It does not give special status to any particular `Associated Address`, rather leaving this (optional) specification to identity applications built on top of the protocol - for instance, `management`, `action`, `claim` and `encryption` keys denominated in the ERC-725 standard, or `identifiers` and `delegates` as denominated in ERC-1056. This allows a user to access common identity data from multiple wallets while still:
Trustlessness in the address management function is achieved through a robust permissioning scheme. To add an `Associated Address` to an `Identity`, implicit permission from a transaction sender or explicit permission from a signature is required from 1) an address already within the registry and 2) an address to be claimed. Importantly, the transaction need not come from any particular address, as long as permission is established, which allows not only users but third parties (companies, governments, etc.) to bear the overhead of managing identities. To prevent a compromised `Associated Address` from unilaterally removing other `Associated Addresses`, it's only possible to remove an `Associated Address` by transacting or producing a signature from it.
All signatures required in ERC-1484 are designed per the [ERC-191](https://github.com/ethereum/EIPs/blob/master/EIPS/eip-191.md) v0 specification. To avoid replay attacks, all signatures must include a timestamp within a rolling lagged window of the current `block.timestamp`. For more information, see this [best practices document](https://github.com/NoahZinsmeister/ERC-1484/blob/master/best-practices/VerifyingSignatures.md) in the reference implementation.
While the protocol allows users to directly call identity management functions, it also aims to be more robust and future-proof by allowing `Providers`, typically smart contracts, to perform identity management functions on a user's behalf. A `Provider` set by an `Identity` can perform address management and resolver management functions by passing a user's `EIN` in function calls.
A `Resolver` is any smart contract that encodes information which resolves to an `Identity`. We remain agnostic about the specific information that can be encoded in a resolver and the functionality that this enables. The existence of `Resolvers` is primarily what makes this ERC an identity *protocol* rather than an identity *application*. `Resolvers` resolve abstract data in smart contracts to an atomic entity, the `Identity`.
If users lose control over an `Associated Address`, the `Recovery Address` provides a fallback mechanism. Upon `Identity` creation, a `Recovery Address` is passed as a parameter by the creator. Recovery functionality is triggered in three scenarios:
**1. Changing Recovery Address**: If a recovery key is lost, an `Associated Address`/`Provider` can [triggerRecoveryAddressChange](#triggerRecoveryAddressChange)/[triggerRecoveryAddressChangeFor](#triggerRecoveryAddressChangeFor). To prevent malicious behavior from someone who has gained control of an `Associated Address` or `Provider` and is changing the `Recovery Address` to one under their control, this action triggers a 14 day challenge period during which the old `Recovery Address` may reject the change by [triggering recovery](#triggerRecovery). If the `Recovery Address` does not reject the change within 14 days, the `Recovery Address` is changed.
**2. Recovery**: Recovery occurs when a user recognizes that an `Associated Address` or the `Recovery Address` belonging to the user is lost or stolen. In this instance the `Recovery Address` must call [triggerRecovery](#triggerRecovery). This removes all `Associated Addresses` and `Providers` from the corresponding `Identity` and replaces them with an address passed in the function call. The `Identity` and associated `Resolvers` maintain integrity. The user is now responsible for adding the appropriate un-compromised addresses back to their `Identity`.
*Importantly, the `Recovery Address` can be a user-controlled wallet or another address, such as a multisig wallet or smart contract. This allows for arbitrarily sophisticated recovery logic! This includes the potential for recovery to be fully compliant with standards such as [DID](https://decentralized.id/).*
The Recovery scheme offers considerable power to a `Recovery Address`; accordingly, `Destruction` is a nuclear option to combat malicious control over an `Identity` when a `Recovery Address` is compromised. If a malicious actor compromises a user's `Recovery Address` and triggers recovery, any address removed in the `Recovery` process can call [triggerDestruction](#triggerDestruction) within 14 days to permanently disable the `Identity`. The user would then need to create a new `Identity`, and would be responsible for engaging in recovery schemes for any identity applications built in the `Resolver` or `Provider` layers.
We considered many possible alternatives when devising the Recovery process outlined above. We ultimately selected the scheme that was most un-opinionated, modular, and consistent with the philosophy behind the `Associated Address`, `Provider`, and `Resolver` components. Still, we feel that it is important to highlight some of the other recovery options we considered, to provide a rationale as to how we settled on what we did.
**High Level Concerns**
Fundamentally, a Recovery scheme needs to be resilient to a compromised address taking control of a user's `Identity`. A secondary concern is preventing a compromised address from maliciously destroying a user's identity due to off-chain utility, which is not an optimal scenario, but is strictly better than if they've gained control.
**Alternative 1: Nuclear Option**
This approach would allow any `Associated Address` to destroy an `Identity` whenever another `Associated Address` is compromised. While this may seem severe, we strongly considered it because this ERC is an identity *protocol*, not an identity *application*. This means that though a user's compromised `Identity` is destroyed, they should still have recourse to whatever restoration mechanisms are available in each of their actual identities at the `Resolver` and/or `Provider` level. We ultimately dismissed this approach for two main reasons:
- It is not robust in cases where a user has only one `Associated Address`
- It would increase the frequency of recovery requests to identity applications due to its unforgiving nature.
**Alternative 2: Unilateral Address Removal via Providers**
This would allow `Associated Addresses`/`Providers` to remove `Associated Addresses` without a signature from said address. This implementation would allow `Providers` to include arbitrarily sophisticated schemes for removing a rogue address - for instance, multi-sig requirements, centralized off-chain verification, user controlled master addresses, deferral to a jurisdictional contract, and more. To prevent a compromised `Associated Address` from simply setting a malicious `Provider` to remove un-compromised addresses, it would have required a waiting period between when a `Provider` is set and when they would be able to remove an `Associated Address`. We dismissed this approach because we felt it placed too high of a burden on `Providers`. If a `Provider` offered a sophisticated range of functionality to a user, but post-deployment a threat was found in the Recovery logic of the provider, `Provider`-specific infrastructure would need to be rebuilt. We also considered including a flag that would allow a user to decide whether or not a `Provider` may remove `Associated Addresses` unilaterally. Ultimately, we concluded that only allowing removal of `Associated Addresses` via the `Recovery Address` enables equally sophisticated recovery logic while separating the functionality from `Providers`, leaving less room for users to relinquish control to potentially flawed implementations.
We find that at a protocol layer, identities should not rely on specific claim or attestation structures, but should instead be a part of a trustless framework upon which arbitrarily sophisticated claim and attestation structures may be built.
The main criticism of existing identity solutions is that they're overly restrictive. We aim to limit requirements, keep identities modular and future-proof, and remain un-opinionated regarding any functionality a particular identity component may have. This proposal gives users the option to interact on the blockchain using an robust `Identity` rather than just an address.
Preforms the same logic as `createIdentity`, but can be called by any address. This function requires a signature from the `associatedAddress` to ensure their consent.
Adds the `addressToAdd` to the `EIN` of the `approvingAddress`. The `msg.sender` must be either of the `approvingAddress` or the `addressToAdd`, and the signature must be from the other one.
`Identities` established under this standard consist of existing Ethereum addresses; accordingly, there are no backwards compatibility issues. Deployed, non-upgradeable smart contracts that wish to become `Resolvers` for `Identities` will need to write wrapper contracts that resolve addresses to `EIN`-denominated `Identities`.
