Verifiable Anonymous Identities and Access Control in Permissioned Blockchains

03/11/2019 ∙ by Thomas Hardjono, et al. ∙ MIT 0

In this paper we address the issue of identity and access control within shared permissioned blockchains. We propose the ChainAchor system that provides anonymous but verifiable identities for entities on the blockchain. ChainAchor also provides access control to entities seeking to submit transactions to the blockchain to read/verify transactions on the the permissioned blockchain. Consensus nodes enforce access control to the shared permissioned blockchain by a simple look-up to a (read-only) list of anonymous members' public-keys. ChainAnchor also provides unlinkability of transactions belonging to an entity on the blockchain. This allows for an entity to optionally disclose their identity when a transaction is called into question (e.g. regulatory or compliance requirements), but without affecting the anonymity and unlinkability of their remaining transactions.

READ FULL TEXT VIEW PDF
POST COMMENT

Comments

There are no comments yet.

Authors

page 1

page 2

page 3

page 4

This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.

I Introduction

The recent rise to prominence of the Bitcoin [1] decentralized digital currency system has generated broad interest in blockchains as a new form of infrastructure for maintaining a shared and cryptographically immutable ledger. Consequently interest has also peaked in the possible development of permissioned and private blockchain systems, in contrast to the permissionless and public blockchain in Bitcoin.

There are numerous use-cases for permissioned blockchain systems whose goal is to provide an immutable ledger that captures the existence of digital facts or artifacts (e.g. transactions, documents, etc) in a given moment in time in a non-repudiable manner. In order to understand the differences among these private permissioned blockchain systems, we believe it is useful to further distinguish between closed and shared permissioned blockchain systems. We define a

closed permissioned blockchain as one in which the entities having access to the private blockchain all belong to the same organization having a common business interest. We define a shared permissioned blockchain as a private blockchain where the transacting entities belong to distinct organizations with competing interests.

In looking at closed and shared permissioned blockchain systems there a number of challenges with regards to identity and access control to the blockchain:

  • Identity privacy: The issue of identity privacy can be acute when a blockchain is shared among competing entities. There is a potential that the behavior of an identity on a blockchain may inadvertently disclose or leak information to competitors transacting on the same blockchain.

  • Access control: In a shared permissioned blockchain, controlling access (both read-access and write-access) is crucial to the value of the blockchain. New approaches are needed beyond the classic Enterprise access control regime that prioritize the security of the infrastructure and services over the identity privacy of entities using the infrastructure.

  • Optional disclosure & transaction privacy: New approaches are needed to provide transaction privacy in the form of unlinkability of transactions. We believe that new solutions are needed for optional disclosure of identities relating to transactions that come into questions (e.g. AML or regulatory compliance). This feature allows an individual to own multiple unlinkable transaction keys on the blockchain, and disclose ownership of a key (e.g. upon legal challenge) without affecting the security and privacy of his or her remaining other keys on the same shared permissioned blockchain.

In this paper we propose the ChainAnchor system that addresses these challenges. The current work expands and generalizes our previous work [2] that addressed constrained devices in an IoT environment. In the next section we describe the ChainAnchor architecture and protocol steps. The design of ChainAchor aims to be functionally independent from the underlying blockchain system, with the goal of deployability of ChainAnchor on various blockchains.

The current paper seeks to be readable to a broad audience, and as such it does not cover in-depth the cryptography behind EPID [3] and DAA [4] schemes that provide for identity anonymity. We assume the reader is familiar with public-key cryptography and with the basic operations of the blockchain in the Bitcoin system. In order to assist the more curious reader, we provide a brief summary of the EPID scheme in the Appendix and provide pointers to the relevant equations in the Appendix. The current paper focuses on an RSA-based EPID scheme based on the Camenisch-Lysyanskaya signature scheme [5] and the DAA scheme of Brickell, Camenisch and Chen [4]. Readers are directed to the authoritative papers of [4] and [3] for an in-depth discussion. An EPID scheme using bilinear pairings can be found in [6]. It is based on the Boneh, Boyen and Schacham group signature scheme [7] and the Boneh-Schacham group signature scheme [8].

Fig. 1: Overview of ChainAnchor interactions

Ii ChainAnchor Architecture

As mentioned previously, we seek to address two important aspects of blockchain systems, namely that of the privacy of the identity of the user and that of providing access control to a permissioned blockchain. Figure 1 summarizes the ChainAnchor entities and interactions.

Our proposed ChainAnchor system generalizes our previous device-centric solution reported in [2], and provides the following features:

  • Anonymous but verifiable identities: Firstly, ChainAnchor allows participants in a shared permissioned blockchain to maintain their privacy by allowing them to use a verifiable anonymous identity when transacting on the blockchain. The anonymity (pseudonymity) of their identity is achieved using the EPID zero-knowledge proof scheme [3], which allows the owner to remain truly anonymous in transactions but allows other parties to verify that the identity is genuine. The verification process is such that no PII or other identifying information disclosed about the owner of the identity.   

  • Access control to blockchain: Secondly, the anonymous (but verifiable) identities make-up the “group” of entities allowed to access the permissioned blockchain. An entity who has proved their membership using the zero knowledge proof (ZKP) protocol can then “register” their self-asserted transaction public-key (for transacting on the blockchain). The consensus nodes collectively enforce access control to the blockchain by only processing transactions from the members’ transaction public-keys. Transactions originated from (destined to) unknown public-keys are simply ignored or dropped.   

  • Multiple unlinkable transaction keys: A member of a shared blockchain can execute the ZKP protocol as many times as they desire, each time registering a different self-asserted transaction public-key. No one on the blockchain (including the owner of the blockchain) has the ability to link these keys to the single owner. This allows a member to optionally disclose their identity (e.g. when challenged in a regulatory context) without endangering their other transactions public-keys.   

When a user requests membership to the permissioned blockchain, the user sends this request to the a Permissions Issuer entity who implements the permissioned group on behalf of the owner of the permissioned blockchain. For simplicity, we assume that the group owner and the owner of the permissioned blockchain are the same entity. In this step, the real-world identity of the user will be known to the group owner and Permissions Issuer entity. The user could represent himself or herself, or represent an organization who is a member of the group sharing the permissioned blockchain. The precise credentials and attributes required for membership approval is outside the scope of the current work.

Once the user is approved to join the permissioned blockchain by the group owner, the Permissions Issuer provides the user with a number of user-specific keying material. This keying material is crucial for the user to later prove membership and to then register the public-key intended to be used on the permissiond blockchain – referred to as the user’s transaction public key.

The entity to whom the user must prove membership in an anonymous fashion and register the user’s transaction public key is the Permissions Verifier entity. This entity must be distinct from the Permissions Issuer entity in order to protect the privacy of the user. Note that since the user’s transaction public-key pair is self-generated, only the user knows the transaction private-key.

The user proves membership to the Permissions Verifier by using the zero-knowledge proof ZKP protocol, which by design protects the anonymity of the user. Once a user successfully completes this protocol, the user’s transaction public-key is delivered by the user to the Permissions Verifier under a secure channel. In turn the Permissions Verifier will add the user’s transaction public-key to a Permissions Database for that group.

This permissions database – which can be a simple list – is maintained by the Permissions Verifier entity and is read-accessible by the consensus nodes (i.e. “miners”) in the permissioned blockchain. A user can have as many transaction public-keys as they wish on the same blockchain. This is achieved by the user executing a distinct run of the zero-knowledge proof protocol, each time registering a different transaction public-key.

The database only holds the transaction public-keys of the members and the timestamp of the completion of the zero-knowledge proof protocol execution. The Permissions Verifier is not able to distinguish one user from another. Furthermore, the Permissions Verifier is not able to know whether or not a user has multiple transaction public-keys registered in the permissions database.

Access-control for the permissioned blockchain is enforced by the consensus nodes in the blockchain, based on the list of (anonymous) public-keys in the permissions database. Prior to processing an unconfirmed transaction, a consensus node enforcing access-control must first verify that the public-key associated with the transaction is present in the permissions database for that blockchain. In other words, the consensus nodes in the permissioned blockchain must ensure that the blockchain contains transactions only from anonymous users whose transaction public-keys are listed in the permissions database.

In this paper we propose the functions of the Permissions Issuer and Permissions Verifier to be implemented by distinct identity provider (IdP) entities. This is because the IdP is the traditional issuer of a digital identities on the Internet and thus possess the necessary infrastructure implementing standard protocols relating to identity management and identity federation. Furthermore, many Enterprise organizations today are already operating an IdP service (or server) internally. As such, an Identity Federation (e.g. via SAML2.0 or OpenID-Connect) can be established for organizations which are sharing a permissioned blockchain.

Ii-a Entities in the System

  • Identity Provider and Permissions Issuer (IdP-PI):
    The IdP-PI is the identity provider entity that establishes the permissioned group on behalf of the group owner. For a given permissioned group, there is exactly one IdP-PI entity.   

  • Permissions Verifier (IdP-PV):
    The IdP-PV is the identity provider entity that performs the anonymous group-membership verification of a given a User by running the zero-knowledge proof protocol with that User. The IdP-PV maintains the Permissions Database. For a given permissioned group, there can be multiple independent IdP-PV entities (although only one IdP-PI for the group).   

  • User:
    The User is the entity wishing to join the permissioned blockchain (i.e. group permissioned) implemetted by the IdP-PI and the IdP-PV.   

  • Consensus node:
    The Consensus Node (“miner”) is entity processing transactions from valid group members of a permissioned blockchain.   

  • Owner:
    Although not shown explicitly in Figure 1, a permissioned group must be owned by an organization or individual. We use the term Owner for this entity.

Ii-B Keys in the System

A major feature of the EPID zero-knowledge proof scheme [3] is its “signature of proof” mechanism, which allows multiple distinct private-keys to be used with one public-key. This allows each distinct User to deploy individually unique EPID private-keys (which they keep secret), from which any signature can be verified by the IdP-PV using the single EPID public-key. We refer to this public-key as the membership verification public-key, and the multiple distinct private-keys as the user-member private-key.

More specifically, for a given permissioned group there is one (1) membership verification public key held by the IdP-PV entity. That one public-key is used by the IdP-PV to validate the membership of multiple () users whose corresponding user-member private keys are .

The keys in the ChainAnchor system are summarized below. We adopt the notational convention of [3] by denoting a public-key pair as , with the public-key being .

  • Membership Issuing Private Key:
    This key is denoted as and is is generated by the IdP-PI for each permissioned group that the IdP-PI establishes. This key is unique for each permissioned group. This key is used by the IdP-PI in enrolling or adding new Users to the permissioned group. (This key is shown in Eq. 2 in the Appendix).   

  • Membership Verification Public Key:
    This key is denoted as and is generated by the IdP-PI and is delivered over a secure channel to the Permissions Verifier entity (IdP-PV). This key is unique for each permissioned group. (This key is shown in Eq. 1 in the Appendix).   

  • User-Member Private Key:
    For a given User who is a member in a permissioned group , the user-member private key is denoted as . (This key is shown in Eq. 6 in the Appendix).   

  • User’s Transaction Public-Key Pair:
    This is the transaction public-key pair that the User employs to transact. This key pair is generated by the User (i.e. user’s computer or device). We denote this public-key pair as , with the public key being .   

  • User’s Identity Public-Key Pair:
    This is the public-key pair that the User employs to represent himself/herself as a member of the permissioned-group. This key pair is generated by the User (i.e. user’s computer or device). We denote this public-key pair as , with the public key being .   

  • User & IdP-PV Pairwise Shared Key (PSK):
    As part of proving group-membership, the User and the DB-PV will establish a pairwise shared key (PSK). The PSK is a symmetric key.   

  • IdP-PI and IdP-PV Certificates:
    These are traditional public-key pairs and X509 certificates:

    • IdP-PI public-key pair: We denote the public key pair of the IdP-PI as with the public key being .   

    • IdP-PV public-key pair: Similarly, we denote the public key pair of the IdP-PV as with the public key being .   

Ii-C ChainAnchor Protocol Steps

In the following, we describe the steps of the ChainAnchor design (see Figure 1).

[Step 0] IdP-PI Establishes Permissioned Group:

This step is not shown in Figure 1. As part of the creation of a permissioned group, the IdP-PI generates a number parameters that are unique to the permissioned group and are used to create two important keys related to the function of the IdP-PI as the Permissions Issuer:

  • Membership Verification Public Key:
    The IdP-PI creates this key to be used later by the Permissions Verifier entity (IdP-PV) when engaging the User in the zero-knowledge proofs protocol. (See Equation 1 in Appendix A).   

  • Membership Issuing Private Key:
    The IdP-PI creates this key in order to issue unique keys to Users in the system that allows the User later to prove membership to the IdP-PV. (See Equation 2 in Appendix A). This issuing private key is kept secret by the IdP-PI.


[Step 1] IdP-PI Shares Verification Public Key with IdP-PV:

In this step, the IdP-PI makes known the Membership Verification Public Key () to the IdP-PV. We assume a secure channel with mutual authentication is used between the IdP-PI and IdP-PV entities.


[Step 2] User Authenticates & Requests Membership

To join the permissioned group the User sends the request to the IdP-PI that manages the permissioned group of interest. The User must first authenticate itself to the IdP-PI. The method used to authenticate is outside the scope of the current paper.

At this point in the ChainAnchor protocol the User is not anonymous to the IdP-PI, and it knows the true identity of the User (e.g. has an account such as alice@idp-issuer.com at the IdP-PI).

A User who has successfully authenticated and obtained approval to join the group is then given a copy of the Membership Verification Public Key by the IdP-IP using a secure channel.   

[Step 3] User Delivers Blinded Commitment Parameters

After obtaining Membership Verification Public Key for the relevant group, the User perform the following tasks:

  • User validates the Membership Verification Public Key: Prior to using the User must verify that the components in are formed correctly (see Equation 1 in Appendix A).   

  • User generates blinded commitment parameters: The User employs some of the parameters in to create his/her own commitment parameters that “blinds” the User’s own secret keying material to the IdP-PI. (See Equations 3 and 4 in Appendix A).   

  • User sends blinded commitment parameters to the IdP-PI: The User sends the commitment parameters to the IdP-PI, who in-turn must verify that these parameters are formed correctly.   

It is important to note here that the cryptographic blinding (in the commitment values) is done to retain the anonymity of the User to the IdP-PI. The IdP-PI must unable to distinguish one user from another at this point based on a user’s blinded commitment parameters.   

Fig. 2: User Anonymously Proves Membership to IdP-PV

[Step 4] IdP-PI Responds with Group-Member Keying Parameters

In this step, after validating the blinded commitment parameters the IdP-PI generates a number of parameters associated with the Member Private Key and sends them to the User. (This is shown in Eq. 5 in the Appendix). In-turn the User uses these parameters to generate its own user-member private key (in the next step).   

[Step 5] User Generates User-Member Private Key

Upon receiving the user-specific group-member keying parameters, the User uses these parameters to generate his/her own User-Member Private Key, denoted as (See Equation 6 in Appendix A). Additionally, the User also generates its blockchain transaction key pair .   

[Step 6] User Proves Membership to IdP-PV

The anonymous membership verification protocol consists of a number of sub-steps following the challenge-response model. The User sends a request to the IdP-PV, and in-turn the IdP-PV challenges the User with some parameters that the User must respond to.

The sub-steps of the anonymous membership verification protocol are as follows (Figure 2):

  • Step 6.1: The User sends a request to the IdP-PV for an anonymous membership verification (Figure 2(a)).   

  • Step 6.2: The Permissions Verifier IdP-PV responds by returning a challenge message and a random nonce to the User (Figure 2(b)).   

  • Step 6.3: Upon receiving the challenge message and the random nonce from the verifier IdP-PV, the User must compute a “signature of knowledge” of the commitment parameter that the User supplied to the IdP-PI in Step 2. (See Figure 2(c)). The signature-of-knowledge is denoted as . (See Equation 8 in Appendix A).

    As input into the signature-of-knowledge computation, the User inputs:

    • The Membership Verification Public Key () for the group which the User obtained from the Permissions Issuer IdP-PI in Step 2. (See Equation 1 in Appendix A).

    • The User’s own User-Member Private Key which the User computed in Step 3. (See Equation 6 in Appendix A).

    • The challenge and the nonce obtained from the Permissions Verifier IdP-PV.   

  • Step 6.4: The User sends the computed signature-of-knowledge value to the IdP-PV as proof of the user’s membership in the group (Figure 2(d)).   

  • Step 6.5: The IdP-PV validates signature-of-knowledge , and returns an acknowledgement of a successful verification process to the User together with some parameters to establish a pair-wise shared key (PSK) between the User and the IdP-PV.   

  • Step 6.6: The User and the IdP-PV engage in a key agreement subprotocol that results in a pair-wise shared key (PSK) – (Figure 2(e)). This PSK is shared between the User (who is anonymous throughout Step 6) and the IdP-PV.   

  • Step 6.7: The User delivers his or her transaction public-key to the IDP-PV under a secure channel created using the shared PSK. The IdP-PV then adds the User’s transaction public key to the permissions database. (Figure 2(f)).   

[Step 7] IdP-PV Creates Anonymous Internet Identity

Optionally, as the result of a successful anonymous membership verification of the User in the previous step, the IdP-PV creates a new anonymous Internet identity for the user (e.g. anon123@idp-verifier.com). The IdP-PV returns a copy of this new Internet identity to the User (Figure 2(g)) under a secure channel created using the PSK.   

[Step 8] User Transacts using Transaction Key-Pair

In this phase the User transacts on the blockchain in the usual manner using the transaction private-key to sign transactions.   

[Step 9] Consensus Receives or Fetches Transaction

The consensus node fetches a transaction (i.e. from the pool of unprocessed transactions) and prepares to process that transaction.   

[Step 10] Consensus Node Validates User’s Public Key

Prior to processing a transaction, a consensus node participating in the ChainAnchor permissioned-group must check that the public-key found in the transaction has been approved to participate in the permissioned-group. That is, the consensus node must first look-up the permissions databases at the IdP-PV to ensure the public key is in the database.   

[Step 11 & 12] Consensus Node Processes Transaction

If the public key used in the User’s transaction exists in the permissions database at the IdP-PV, the consensus node can proceed with processing the transaction. Otherwise the consensus node can choose to ignore the transaction.   

Iii Verifiable Anonymous Identities

The identities in ChainAnchor – in the form of transaction public-keys – are anonymous because the transaction public-key pairs are self-generated by the user (just as in the Bitcoin systems and in the PGP system). Only the user knows the private key(s).

These identities are verifiable because the user at any time can execute the zero knowledge proof protocol with the IdP-PV in order to prove membership in the given blockchain. The unlikability of the multiple public-key pairs (belonging to a single user) is also derived from the zero knowledge proof protocol [3].

  • User remain anonymous to IdP-PV: The IdP-PV cannot distinguish among validated users. More specifically, if two Users and independently returns the challenge message with a signature-of-knowledge (see Equation 8) created using keys and then the IdP-PV can verify both signature using the one verification public key but it will not be able to distinguish between Users and .   

  • User remain anonymous to IdP-PI: When the User requests membership to the group, the User (person) is known to the IdP-PI. However, after Step 5 the User becomes anonymous even to the IdP-PI because the User injects a secret “blinding” parameter (in Step 3) when generating the User’s User-Member Private Key (see Equations 34 and 6 in the Appendix). Since the IdP-PI is not involved in Step 5 onwards, the IdP-PI has no knowledge of which transaction public-key pairs are owned by the User.   

  • Optional Disclosure of Transaction Keys: A User can deploy as many transaction keys as they wish within the blockchain. This is achieved by the User executing the ZKP protocol with the IdP-PV for each of the keys he or she wishes to register and use in the blockchain. This approach has the advantage that the User may reveal (to the IdP-PV) the User’s ownership of a given transaction key without affecting other transaction keys (i.e. unlinkability).   

Iv Access Control to the Shared Permissioned Blockchain

Fig. 3: Consensus node enforcing access control

In ChainAnchor the term “consensus” includes the notion of membership to a permissioned (private) blockchain. ChainAnchor employs the consensus nodes (miners) in the shared permissioned blockchain to collectively enforce access control, by way of a simple “filtering” of transaction belonging to members (see Figure 3). A consensus node needs to verify that the public-key of a transaction that it wishes to process belongs to a member of the permissioned blockchain. It does this verification by a simple look-up to the permissions database at IdP-PV. (Other list-sharing methods can also be deployed, such as the node maintaining a local copy of the permissions database).

However, in order to counter possible cheating from dishonest nodes, (who may allow non-member transaction to come into the blockchain) ChainAnchor also requires that the IdP-PV and IdP-PI be Validator nodes that checks that transactions belong to members of the blockchain. The issue of the weight of a decision (vote) coming from the IdP-PV or IdP-PI (versus coming from a consensus node) is outside the scope of the current paper and will be the subject of further study.

V Anonymous Identities: Beyond the Blockchain

ChainAnchor allows for further anonymous and addressable identities (e.g. email-based identities) to be derived based on the ZKP protocol that the user executed with the IdP-PV. Figure 4 illustrates as follows:

  1. User’s known Internet identity and the IdP-PI: When a user seeks to participate within a permissioned group, he or she must be approved by the group-owner (creator). This is a normal requirement, particularly for ledgers shared by business participants. In doing so the user must use a real world identity. This implies that some degree of personally-identifying information (PII) must be disclosed from the user to the group-owner (and possibly also the IdP-PI entity that is hosting the permissioned group for the owner). In Figure 4(a) we show this real-world identity as Alice’s Internet identity denoted by alice@idp-issuer.com. We assume here the IdP-PI knows all attributes relating to Alice.

  2. Regaining anonymity on the permissioned blockchain: After the User has been approved to join the permissioned group for the blockchain, the user is able to regain anonymity by using cryptographic blinding function in Step 3 of the protocol. This blinding function prevents the IdP-PI entity from knowing that the User-Member Private Key that was generated by the person (Alice) whose Internet identity is alice@idp-issuer.com in Figure 4.

  3. Anonymity of user’s transaction public-key: The anonymity of the User (obtained from the blinding function) is further protected by the zero knowledge proof protocol that is executed between the User and the IdP-PV (Figure 4(b)). The IdP-PV entity has no way of correlating between the User’s transaction public-key (which is self-generated by the User) and the Internet identity alice@idp-issuer.com employed by the User (Figure 4(a)) to initially engage the Permissions Issuer entity.

  4. Derived anonymous Internet identity for the key-holder: The cryptographically anonymous relationship between the key-holder (i.e. our User) and the IdP-PV in Figure 4(c) lends to the possible creation by the IdP-PV of a new anonymous Internet identity for the User, shown as anon123@idp-verifier.com in Figure 4(d).

Thus, in Figure 4(d) the IdP-PV can become an Identity Provider and can issue a new identity anon123@idp-verifier.com for the anonymous user (Alice). Furthermore, IdP-PV can bind (e.g. in an X509 certificate) this new identity with the transaction public-key whose private-key () is known only to the anonymous user (Alice). We believe this approach provides a more scalable solution than PGP.

Fig. 4: Deriving new Internet Addressable Identities

Vi Conclusions & Further Work

In this paper we have addressed the issue of identity and access control within shared permissioned blockchains. We proposed the ChainAchor system that provides anonymous but verifiable identities for entities on the blockchain.

ChainAnchor allows participants in a shared permissioned blockchain to maintain their privacy by allowing them to use a verifiable anonymous identity when transacting on the blockchain. The anonymous (but verifiable) identities make-up the “group” of entities allowed to access the shared permissioned blockchain.

In ChainAnchor the term “consensus” includes the notion of membership to a permissioned (private) blockchain. An entity who has proved their membership using the zero knowledge proof protocol can “register” their self-asserted transaction public-key (for transacting on the blockchain). The consensus nodes collectively enforce access control to the blockchain by only processing transactions from the members’ transaction public-keys. Transactions belonging to unknown public-keys (i.e. non-members) are simply ignored or dropped.

A member of a shared blockchain can execute the zero knowledge proof protocol as many times as they desire, each time registering a different self-asserted transaction public-key. No one on the blockchain (including the owner of the blockchain) has the ability to link these keys to the single owner. This allows a member to optionally disclose their identity (e.g. when challenged in a regulatory context) without endangering their other transactions public-keys.   

Appendix A Summary of EPID

ChainAnchor employs the EPID scheme [3] due to a number of advantages of the scheme. EPID is an extension of the Direct Anonymous Attestation protocol (DAA) [4] for user privacy in the TPMv1.2 hardware [9]. The EPID protocol can be deployed without Trusted Platform Module (TPM) hardware, with the option to add and enable a tamper-resistant TPM at a later stage. This option may be attractive to service providers who may wish to deploy TPM-based infrastructure in a phased approach (see [10, 11]). When a TPM hardware is deployed, it can be used to provide protected storage for the various keys used in the ChainAnchor system.

EPID is not the only anonymous identity protocol available today. The work of Brickell et al. [4] introduced the first RSA-based DAA protocol in 2004. A related anonymity protocol called Idemix [12] employs the same RSA-based anonymous credential scheme as the DAA protocol. However, Idemix cannot be used with the TPMv1.2 hardware (or the new TPMv2.0 hardware). Another related protocol called U-Prove [13] can be integrated into the TPM2.0 hardware (see [14]). However, the U-Prove protocol has the drawback that it is not multi-show unlinkable [15], which means that a U-Prove token may only be used once in order to remain unlinkable.

In the following we summarize the RSA-based EPID scheme as defined in [3].

A-a Issuer Setup

In order to create a group membership verification instance, the Issuer must choose a Group Public Key) and compute a corresponding Group-Issuing Private Key).

For the Group-Issuing Private Key the Issuer chooses an RSA modulus where and and where , , and are all prime.

The Group Public Key for the particular group instance will be:

(1)

The Group Issuing Private Key (corresponding to the Group Public Key) is denoted as:

(2)

which the Issuer keeps secret).

In order to communicate securely with a User, the Issuer is assumed to possess the usual long-term public key pair denoted as , where is publicly know in the ecosystem.

Any User who has a copy of the Group Public Key can verify this public key by checking the following:

  • Verify the proof that and .

  • Check whether and are primes, and check that , and

  • Check whether all group public key parameters have the required length.

A-B Join Protocol: User and Issuer

In the join protocol, a given User seeks to send to the Issuer the pair which are computed as follows.

  • The User chooses a secret and seeks to convey to the Issuer a commitment to in the form of the value .

  • The value is computed as

    (3)

    where is chosen randomly by the User for the purpose of blinding the chosen .

  • Next the User computes

    (4)

    where is derived from the basename of the Issuer (denoted as ).

The goal here is for the User to send to the Issuer and to convince the Issuer that the values and are formed correctly.

In the above Equation 4, a User chooses a base value and then uses it to compute . The purpose of the pair is for a revocation check. We refer to the base and as the pseudonym. To sign an EPID-signature, the User needs to both prove that it has a valid membership credential and also prove that it had constructed the pair correctly, all in zero-knowledge. In EPID and DAA, there are two (2) options to compute the base :

  • Random base: Here is chosen randomly each time by the User. A different base used every time the EPID-signature is performed. Under the decisional Diffie-Hellman assumption, no Verifier entity will be able to link two EPID-signatures using the pairs in the signatures.

  • Named base: Here is derived from the Verifier’s basename. That is, a deterministic function of the name of the verifier is used as a base. For example, could be a hash of the Verifier’s basename. In this named-base option, the value becomes a “pseudonym” of the User with regard to the Verifier’s basename. The User will always use the same in the EPID-signature to the Verifier.

A-C Issuer generates User’s Membership Private Key

In response, the Issuer performs the following steps:

  • The Issuer chooses a random integer and a random prime .

  • The Issuer computes such that

  • The Issuer sends the User the values

    (5)

Note that the CL-signature [5] on the value is . As such, the User then sets his/her Membership Private Key as:

(6)

where . Recall that is the secret chosen by the User at the start of the Join protocol.

A-D User proving valid membership

When a User seeks to prove that he or she is a group member, the User interacts with the Verifier entity. This is performed using the Camenisch-Lysyanskaya (CL) signature [5] on some value .

This can be done using a zero-knowledge proof of knowledge of the values , , , and such that

(7)

The User also needs to perform the following:

  • The User computes where is a random base (chosen by the User).

  • The User reveals and to the Verifier.

  • The User proves to the Verifier that the value is the same as in his/her private key (see Equation 4).

In proving membership to the Verifier, the User as the prover needs to send the Verifier the value

(8)

where each of the values are as follows:

  • : The value is a “signature of knowledge” regarding the User’s commitment to the User’s private key and that was computed using the User’s secret value .

  • : The value is a “signature of knowledge” that the User’s private key has not been revoked by the Verifier (i.e. not present in the signature revocation list sig-RL).

  • : The value is a “signature of knowledge” that the User’s private key has not been revoked by the Issuer (i.e. not present in the issuer revocation list Issuer-RL).

References

  • [1] S. Nakamoto, “Bitcoin: A Peer-to-Peer Electronic Cash System.” [Online]. Available: https://bitcoin.org/bitcoin.pdf
  • [2] T. Hardjono and N. Smith, “Cloud-Based Commissioning of Constrained Devices using Permissioned Blockchains,” in Proceedings of the 2nd ACM International Workshop on IoT Privacy, Trust, and Security (IoTPTS 2016), May 2016.
  • [3] E. Brickell and J. Li, “Enhanced Privacy ID: a Direct Anonymous Attestation Scheme with Enhanced Revocation Capabilities,” IEEE Transactions on Dependable and Secure Computing, vol. 9, no. 3, pp. 345–360, 2012.
  • [4] E. Brickell, J. Camenisch, and L. Chen, “Direct Anonymous Attestation,” in Proceedings of the 11th ACM Conference on Computer and Communications Security CCS2004.   ACM, 2004, pp. 132–145.
  • [5] J. Camenisch and A. Lysyanskaya, “A Signature Scheme with Efficient Protocols,” in Security in Communication Networks (SCN2002) (LNCS 2576), S. Cimato, G. Persiano, and C. Galdi, Eds.   Springer, 2002, pp. 268–289.
  • [6] E. Brickell and J. Li, “Enhanced Privacy ID from Bilinear Pairing for Hardware Authentication and Attestation,” in Proceedings of the IEEE International Conference on Social Computing (SocialCom 2010).   IEEE, 2010, pp. 768–775.
  • [7] D. Boneh, X. Boyen, and H. Shacham, “Short Group Signatures,” in In Advances in Cryptology - Proceedings CRYPTO 2004 (LNCS 3152).   Springer, 2004, pp. 41–55.
  • [8] D. Boneh and H. Shacham, “Group signatures with verifier-local revocation,” in Proceedings of the 11th ACM conference on Computer and communications security.   ACM, 2004, pp. 168–177.
  • [9] Trusted Computing Group, “TPM Main – Specification Version 1.2,” Trusted Computing Group, TCG Published Specification, October 2003, http://www.trustedcomputinggroup.org/ resources/ tpm_main_specification.
  • [10] T. Hardjono, “Infrastructure for Trusted Computing,” in Proceedings of ACSAC Workshop on Trusted Computing, December 2004, available at https://www.acsac.org/2004/workshop/Thomas-Hardjono.pdf.
  • [11] T. Hardjono and N. Smith (Eds), “TCG Infrastructure Reference Architecture for Interoperability (Part 1) – Specification Version 1.0 Rev 1.0,” June 2005, http://www.trustedcomputinggroup.org/ resources.
  • [12] J. Camenisch and E. Van Herreweghen, “Design and implementation of the Idemix anonymous credential system,” in Proceedings of the 9th ACM conference on Computer and communications security.   ACM, 2002, pp. 21–30.
  • [13] Microsoft, “U-Prove Cryptographic Specification v1.1 (Rev 3),” Microsoft Corporation, http://research.microsoft.com/en-us/projects/u-prove, 2014.
  • [14] L. Chen and J. Li, “Flexible and Scalable Digital Signatures in TPM 2.0,” in Proceedings of the 2013 ACM SIGSAC Conference on Computer and Communications Security CCS2013.   ACM, 2013, pp. 37–48.
  • [15] L. Chen and R. Urian, “DAA-A: Direct Anonymous Attestation with Attributes,” in Proceedings of TRUST 2015 (LNCS 9229).   Springer, 2013, pp. 228–245.