Patent Grant
8006086
The present invention relates to public key cryptography, and more particularly to digital certificate revocation.
Digital certificates 104 (
A certificate may have to be revoked prior to its expiration date D2. For example, the certificate owner U may change his affiliation or position, or the owner's private key SKU may be compromised. Other parties must be prevented from using the owner's public key if the certificate is revoked.
One approach to prevent the use of public keys of revoked certificates is through a certificate revocation list (CRL). A CRL is a signed and time-stamped list issued by CA 120 and specifying the revoked certificates by their serial numbers SN. These CRLs must be distributed periodically even if there are no new revoked certificates in order to prevent any type of replay attack. The CRL management may be unwieldy with respect to communication, search, and verification costs. Certificate revocation trees (CRTs) can be used instead of CRLs as described in [15] (the bracketed numbers indicate references listed at the end before the claims).
Instead of CRLs and CRTs, CA 120 could answer queries about specific certificates. In
Let f be a predefined public length-preserving function
f: {0,1}n→{0,1}n
where {0,1}n is the set of all binary strings of a length n. Let fi denote the f-fold composition; that is, fi(x)=x for i=0, and fi(x)=fi(fi-1(x)) for i>0. Let f be one-way, i.e. given f(x) where x is randomly chosen, it is hard (infeasible) to find a pre-image z such that f(z)=f(x), except with negligible probability. “Infeasible” means that given a security parameter k (e.g. k=n), the pre-image z cannot be computed in a time equal to a predefined polynomial in k except with negligible probability. Let us assume moreover that f is one-way on its iterates, i.e. for any i, given y=fi(x), it is infeasible to find z such that f(z)=y.
We can assume, without loss of generality, that CA is required to provide a fresh validity status every day, and the certificates are valid for one year, i.e. 365 days (D2−D1=365 days). To create a certificate 104 (
c365=f(x),c364=f(f(x)), . . . c1=f365(x), c0=f366(x). (1)
We will sometimes denote x as x(SN) for a certificate with a serial number SN, and similarly ci=ci(SN) where i=0, 1, . . . . The value c0 is called a “validation target”. CA 120 inserts c0 into the certificate 104 together with data 104D (
Every day i (i=1, 2, . . . 365), for each certificate 104, CA distributes to directories 210 a validity proof data structure which includes, in addition to a validity status indication (not shown in
1. the certificate's “i-token” ci if the certificate is valid on day i;
2. the revocation seed N0 if the certificate has been revoked.
This information is distributed unsigned. Each directory 210 provides this information, unsigned, to a requester system 110 in response to a validity status request 150 (
1. If the validity status is “valid”, the verifier 110 checks that fi(ci)=c0.
2. If the validity status is “revoked”, the verifier 110 checks that f(N0)=N1.
Despite the validity information being unsigned, the scheme is secure because given ci, it is infeasible to compute the subsequent tokens ci+1, ci+2, . . . .
To reduce the communication between CA 120 and directories 210, a hash chain (1) can be generated for a set of certificates 104, and a single i-token ci can be distributed for the set if the set is “unrevoked” (i.e. all the certificates are unrevoked in the set).
Every day i, if all the certificates are valid, CA 120 distributes to directories 210 only the i-token ci(F1). If only the set F3 has invalid certificates, CA 120 distributes the i-tokens for the set F4 and for all the valid certificates in the set F2. If only the set F2-F3 has invalid certificates, CA 120 distributes the i-tokens for the sets F3 and F4 and for all the valid certificates in F2-F3, and so on.
In response to a validity status request for a certificate 104, a directory 120 sends to the requester (the verifier):
1. an i-token ci for the certificate or for a set Fi containing the certificate if the certificate is valid;
2. the certificate's revocation number N0 if the certificate has been revoked.
If the response indicates that the certificate is valid, the verifier checks that fi(ci) is equal to one of the certificate's validation targets. If the response indicates that the certificate is revoked, the verifier checks that f(N0)=N1 for the certificate.
Clearly, for each set R of revoked certificates (
Also, it is desirable to find a system of sets {F} containing a small complement cover for any set R or at least for many possible sets R. If {F} contains a cover for each set R of the certificates, we will call {F} a complement cover for the set of all the certificates, and will denote this complement cover CC(U) or just CC.
For uniformity, we will assume that each certificate 104 corresponds to a singleton set consisting of that certificate. The hash chain for the singleton set is the same as for the certificate.
Clearly, if {F} contains the singleton set for each certificate, then {F} is a complement cover for the set of all the certificates.
Complement covers can be constructed using trees.
If a certificate is revoked, then the corresponding leaf is revoked, i.e. represents a set that cannot be used for the i-token distribution. Also, each node in the path from the leaf to the root is revoked. In the example of
This section summarizes some features of the invention. Other features are described in the subsequent sections. The invention is defined by the appended claims which are incorporated into this section by reference.
One aspect of the present invention is directed to reducing the certificate size. A certificate 104 in
Another aspect of the invention is directed to redacting the certificate by deleting unnecessary targets. In
Another aspect of the invention is directed to efficient distribution of certificate validity proofs. In some embodiments, the validity proofs (e.g. i-tokens) are distributed to the certificate owners. If a system 110 issues a request 150 for a validity proof, the validity proof is provided by the owner rather than the CA or a directory. A validity proof (e.g. an i-token) for a set F comprising multiple certificates can be distributed to the certificates' owners via a multicast transmission if the corresponding computer systems 110 form a multicasting group. In some embodiments the sets F and the multicasting groups are matched to facilitate the multicast transmissions. E.g., a multicasting group can be created for a set F, or a set F can be created by the CA in the setup phase for a multicasting group. Also, a complement cover CCR can be chosen to maximize the number of sets F for which the multicasting groups exist.
Another aspect of the invention is directed to efficient caching of validity proofs. In some embodiments, if a system 110 (e.g. 110.1) gets a validity proof for a certificate, the system 110.1 may cache the validity proof. Another system 110.2 may get the validity proof from the system 110.1 rather than the CA or a directory. In some embodiments, the certificate sets F are assigned caching priorities which are taken into account by the system 110.1 when making a decision as to whether or not to cache a validity proof for a set F. The caching priorities may be based on the number of certificates in the set F, and/or the sum of the remaining unexpired validity periods for the certificates in F, and/or other factors.
Another aspect of the invention is directed to generation of certificate validation data structures (such as the hash chain seeds x) by the CA. For a hash chain (1), for example, the data structures are generated for a predefined number of periods of time (e.g. 365 days), with each i-token corresponding to the period i. The number of periods of time is defined by the certificate's maximum validity period as defined by the certificate's issue and expiration dates D1 and D2. The number of periods of time is incorporated into the target c0. This complicates the addition of new certificates, especially if a complement cover is used since complement covers interrelate the validation proofs for multiple certificates. In some embodiments, in the setup phase, the CA generates the data structures for more periods of time than required by the certificates' maximum validity periods. For example, the CA can generate the data structures for some predefined number M of certificates for some number Td of periods of time where Td is greater than the maximum validity period. The actual number of the certificates created in the setup phase may be less than M. The CA can add new certificates after the setup phase as long as the new certificates will expire before or in the last time period Td. A validity proof may include the i-token ci(F) and, in addition, the number j of times needed to apply the function f to the i-token to obtain the target c0(F). The verifier checks that fj(ci(F))=c0(F). The hash chains can be replaced with other structures, e.g. hash trees as described below.
Another aspect of the invention is directed to providing a distributed certificate authority. The CA distributes validation data (e.g. i-tokens) to “Sub-CA” computer systems which in turn generate validation proofs and provide them to verifiers. The CA distributes secret data to the Sub-CAs in advance. For example, the CA can distribute all the i-tokens for all the certificates in the setup phase. Any one or more of a number of techniques are used to make the validation secure even if a Sub-CA is compromised.
One technique involves generating different data for the same certificate for different Sub-CAs. Thus, a separate validation seed x and a separate revocation seed N0 can be generated for each certificate for each Sub-CA. The certificate may include all the respective validation and revocation targets. Alternatively, the validation targets may be mapped into a single “super-target” by a public function, and the certificate may have only the validation “super-target”. The revocation targets can be handled in the same way. Alternatively, all the validation and revocation targets can be mapped into a single super-target.
Further, in each period i, a Sub-CA validity proof is made available by the CA for each Sub-CA for the period i. If a Sub-CA is compromised, the CA withholds the Sub-CA's validity proof. Therefore, the verifiers will know that the Sub-CA is invalid, and will not use the Sub-CA's certificate validity proofs. If the remaining Sub-CAs are not compromised, their data remain secret because each Sub-CA has its own certificate validation data as described above.
Further, a mechanism is provided for recovering control of a compromised Sub-CA. The mechanism involves encryption of the certificate validation data before transmitting the data to the Sub-CAs. For each period i, or a number of consecutive periods i (such consecutive periods i will be called a “partition”), separate encryption/decryption keys are used (the encryption can be symmetric or asymmetric). Further, different decryption keys are used for different Sub-CAs. For each partition, the decryption keys are provided to the Sub-CAs at or shortly before the start of the partition. If a Sub-CA is compromised in any partition, the adversary may get hold of the decrypted data (e.g. i-tokens) for the current partition, but the data for the future partitions are encrypted and thus secure. If the CA gets control of the Sub-CA again, the CA may reactivate the Sub-CA at or after the end of the last partition for which the Sub-CA received its decryption key. The CA does not provide the decryption keys to compromised Sub-CAs.
The invention is not limited to the features and advantages described above. Other features are described below. The invention is defined by the appended claims.
The embodiments described in this section illustrate but do not limit the invention. The invention is defined by the appended claims.
We will assume that the CA 120, the directories 210, the systems 110 are computer systems communicating with each other over a network or networks. Each of these systems may itself be a computer system having components communicating over networks. Each computer system includes one or more computer processors executing computer instructions and manipulating computer data as described above and below. The term “data” includes “computer data” and covers both computer instructions and computer data manipulated by the instructions. The instructions and data can be stored on a data carrier such as a computer storage, i.e. a computer readable medium (e.g. a magnetic or optical disk, a semiconductor memory, and other types of media, known or to be invented). The data carrier may include an electromagnetic carrier wave transmitted through space, via a cable, or by some other means. A “cache” can be any computer storage. The instructions and data are operable to cause the computer to execute appropriate algorithms as described above.
We will use the following notation. We let DS=(KG, Sign, Vf) denote a digital signature scheme. Here KG denotes a key generation algorithm, Sign(Sk, M) denotes the signing algorithm which outputs a signature σ on a message M under a signing key Sk. Vf(Pk, M, σ) denotes the verification algorithm which evaluates to a binary value indicating whether or not the signature σ on the message M is correct with respect to a public key Pk. We let {0,1}* denote the set of all bit strings. |s| denotes the length of a bit string s. We let H denote a cryptographic compression function that takes as input a b-bit payload and a v-bit initialization vector IV and produces a v-bit output. In some embodiments, b≧2v. We will assume that the cryptographic compression functions mentioned below can be collision resistant, i.e. it is difficult to find two distinct inputs m1≠m2 such that H(IV,m1)=H(IV,m2). We will assume that IV is fixed and publicly known, and we will sometimes omit it for notational simplicity. Practical examples of such cryptographic compression functions are SHA-1 [26] (output size is 20 bytes) and MD5 [28] (output size 16 bytes), both having a 64-byte payload. For simplicity, we will use the term “hash function” instead of compression function. The term “hash function” can also denote a mapping form {0,1}* into {0,1}v for some fixed v. Hash functions are typically one way and collision resistant, but the invention is not limited to such functions.
Hash Tree Over Targets.
In order to reduce the size of a certificate 104 of
r←A(c0) for all targets c0 for this certificate. (2)
Then the targets c0 can be deleted from the certificate and replaced with the value r.
In some embodiments, the algorithm A is defined using a hash tree data structure. A hash tree is a tree in which each node (“vertex”) v is assigned a value Value(v). A hash tree is created for each certificate 104.
Value(v3)=c0(5), Value(v4)=c0(5-6), and so on.
The values of the higher level nodes are computed using some public algorithm, for example:
Value(v)=H(Value(left child of v)oValue(right child of v)) (3)
where v is any higher level node, H is a public hash function, and “o” denotes string concatenation. The certificate 104 (
At the start of each period pi (we will use the expressions “period pi” and “period i” interchangeably), or shortly before the period pi, CA 120 determines a minimal complement cover CCR of the revoked certificates R (as in
In response to a request from a verifier system 110, a directory 210 responds with a possibly unsigned data structure 830 if the certificate is valid, and a possibly unsigned data structure 840 if the certificate is invalid. Even before issuing the request, the verifier can check the CA's signature on the certificate (step 1004 in
CoNodes(v)=Ø(empty set) if v is the root; (4)
In some embodiments, the structure 830 includes the co-nodes' values shown as Value(CoNodes(Leaf(F, SN))) in
(L,v3); (R,v2) (5)
Here L and R are values of a one-bit flag indicating if the co-node must be on the left or the right in the concatenation. For example, in the expression
Value(v1)=H(Value(v3)oValue(v4)),
Value(v3) is to the left of Value(v4), therefore the flag value for v3 is L.
Of note, the verifier does not need to know the geometry of tree 510 or 710 since the list (5) is sufficient to compute the root value Value(v0). The root value can be computed as follow:
Listing 1: Root Value Computation
1. Root value←starting value (i.e. Value(Leaf(F, SN)=c0(F)).
2. Traverse the list Value(CoNodes(Leaf(F, SN))) (such as the list (5)) from left to right.
For each co-node,
The verifier computes the root value at step 1030 (
The structure 840 for an invalid certificate can be as in prior art, and can have the same information as structure 820. The verifier checks at step 1040 that f(N0)=N1, as in prior art.
The invention is not limited to the hash tree 710 of the form of
The term “tree” denotes any computer data structure together with a method for determining the parent from a child node or the children from the parent. The data structure may include pointers from the children nodes to the parent and/or from the parent to the children. Alternatively, the nodes may be arranged as an array or some other structure, with no pointers, but with a method, implemented by computer instructions, which determines the parent from a child node or the children from the parent.
In some embodiments, the hash chains (1) are replaced with hash trees 1210 (
Each leaf gv(i) is assigned a random or pseudo-random value. The remaining nodes' values are defined by the child nodes' values, using the equation (3) for example, i.e. each parent node's value is a hash function of a concatenation of the children values; at the bottom level, the value of each node v7-v14 is a hash function of the corresponding child. The root value Value(v0) will be called a target, and denoted by c0. A separate tree 1210 is constructed by CA 120 for each set F of the complement cover CC (e.g. for CC 510), with different random leaf values generated separately for each tree 1210.
For each certificate 104, a hash tree 710 (as in
The CA set up procedure (
In each period pi, CA 120 determines a minimal complement cover CCR of the revoked certificates R using the same procedure as described above for
In response to a request from a verifier system 110, a directory 210 responds with a possibly unsigned data structure 830 if the certificate is valid, and a possibly unsigned structure 840 if the certificate is invalid. Structure 830 contains the certificate' serial number SN, an indication that the certificate is valid, and the values Value(gv(i,F)), Value(CoNodes(gv(i,F)), Value(CoNodes(Leaf(F,SN))). Here F is an element of CCR containing the SN certificate; Leaf(F,SN) is the leaf node corresponding to the set F of the certificate's tree 710. The verification procedure is shown in
Suppose that there are N=2k certificates. Then the tree 510 has (1+k) levels, and each certificate belongs to (1+k) sets F. This is the number of validation targets c0(F). Hence, there are about log2(1+k) levels in hash tree 710, with about the same number of co-nodes. Let us suppose that the certificate owner provides the certificate and the validity proof to a requester. Then the use of hash trees 710 provides an improvement by O(k/log2k). This is significant for large k.
In some embodiments, the certificate's targets c0 are sub-divided into groups of targets, and a separate hash tree 710 is defined for each group. The certificate contains the root value Value(v0) of each such hash tree. Other variations are also possible.
Redacting the Certificate
As illustrated in
In
At the start of, or shortly before, each period pi, the validity proof 1810 for the certificate 104 is provided (possibly pushed) to the certificate owner's computer system 110.1. Validity proof 1810 may have the same form as data 810 of
Certificate owner system 110.1 provides the validity proof 1810 to party (computer system) 110.2 together with the certificate 104. According to some aspects of the present invention, the party 110.1 redacts the certificate by deleting the unneeded targets to reduce the amount of data provided to party 110.2. The redacted certificate 104R includes a signature proof 104R-SigCA to allow the party 110.2 to verify the CA's signature on the certificate. Party 110.1 does not have the CA's secret key. CA 120 uses a redactable signature scheme to enable the party 110.1 to provide the signature proof 104T-SigCA. Suitable redactable signature schemes are described in [12], and other schemes may be suitable, whether known or to be invented. One scheme is illustrated in
Suppose each block of the message x has k bits (or less). Let G: {0,1}k{0,1}2k be a pseudo-random generator. We assign a k-bit number ks to each node s as follows:
Listing 2
1. Pick kE uniformly randomly from {0,1}k.
2. Recursively, for each node s, define k50 as the first k bits of G(ks), and ks1 as the remaining k bits.
End of Listing 2.
Each leaf node s as assigned to a corresponding block xs of the message x (where “s” is interpreted both as a string and as a binary number). We define a hash tree structure on tree 1910, using some predefined hash function H, with a value vs assigned to each node s as follows. For the leaf nodes,
vs=H(0,ks,xs) (6)
(the entities 0, ks, xs are concatenated). If there are more leaf nodes than blocks xs, the corresponding vs values can be generated randomly or pre-assigned (e.g. set to 0). For every non-leaf node s,
vs=H(1,vs0,vs1). (7)
The redactable signature Sig(x) on the message x is defined as
Sig(x)=<kE,Sig0(vE)> (8)
If the message x is redacted by a deletion of some block XL for some string L, the signature proof (which will also be denoted as Sig(x)) is:
Sig(x)=<k_Value(CoNodes(L)),vL,Sig0(vE)> (9)
where k_Value(CoNodes(L)) is the set of the ks values for the nodes s in CoNodes(L). Clearly, k_Value(CoNodes(L)) is sufficient for the verifier to compute ks for the leafs other than L, and hence to compute vs for these leafs. The value vL is part of signature proof (9), so the verifier can compute vE from equation (7) and verify the base signature Sig0(vE).
Multiple targets can be deleted sequentially. Alternatively, the following techniques can sometimes be used. Suppose we delete all the targets except Tar-1 (i.e. all the message blocks for the leafs 010 through 111 in
Sig(x)=<k00,v01,v1,Sig0(vE)> (10)
(The signature proof Sig(x) may also include the label s for each ks and vs value, and other data as needed to interpret the components of the signature proof). This signature proof has all the information needed for the verification. If all the targets except Tar-7 are deleted, the signature proof can be:
Sig(x)=<k000,v01,K111,v110,v01,Sig0(vE)> (10)
Here we delete six targets but add five k-bit values to the signature proof compared to the signature proof (8) for the original message x. The signature proof (11) can be shortened however if we provide kE instead of k000 and k111. In some prior applications, the value kE was not provided to hide the ks values for the deleted message blocks xs in order to make it harder for the verifier to determine the contents of the deleted blocks. In some embodiments, however, the certificate owner 110.1 does not try to hide the deleted targets, thus allowing a shorter signature proof to be constructed. Moreover, kE can be a public constant rather than randomly generated since the signature security is provided by the base signature scheme Sig0. In this case, the kE value does not need to be provided as part of the signature proof. Further, the function G does not have to be a pseudorandom generator.
In some embodiments, the certificate owner 110.1 first checks if the redacted certificate 104R is smaller than the non-redacted certificate 104. The certificate owner sends to party 110.2 the shortest of the two certificates. In another embodiment, the certificate owner 110.1 sends the shortest among the certificate 104 and all the redacted certificates containing the target Tar-i (deleting fewer than all of the unneeded targets may provide a shorter certificate than deleting all of the unneeded targets).
Multicast Distribution of Validity Proofs
Pushing validity status information can be advantageous in ad hoc networks. An ad hoc network is a self-configuring wireless network of systems 110 (
In some embodiments, CA 120 generates a complement cover CC(U) taking into account possible multicasting groups 2010. For example, a separate set F can be created for each multicasting group 2010. The set F will consist of certificates 104 owned by operators of stations 110 in the multicasting group 2010. Also, when CA 120 generates a complement cover CCR of the revoked certificates, CA 120 may preferentially choose the sets F corresponding to multicasting groups 2010. In some embodiments, the sets F are assigned priorities, with a higher priorities assigned to sets F corresponding to multicasting groups 2010. The priorities may also depend on other factors. In some embodiments,
Priority(F1)≧Priority(F2) if F1⊃F2 (12)
CA 120 first selects the sets F of the highest priority when generating the complement cover CCR, then the sets F of the next highest priority, and so on. In other embodiments, CA 120 maximizes the sum of the priorities of the sets F in CCR. Other schemes for taking the multicasting groups into account can also be used.
Suppose for example that the complement cover CC(U) includes the sets F of
In some embodiments, a set F may correspond only approximately to a multicasting group 2010. The set F may contain some, but not all, of the certificates associated with a group 2010. The set F may receive a higher priority if F contains more than one certificates associated with a multicasting group of systems 110. In some embodiments, Priority(F) is proportional to the number of such certificates. (“Proportional to” means “increases with” and does not necessarily mean “directly proportional”).
In some embodiments, CA 120 may or may not take the multicasting groups into account, but the network of stations 110 forms at least some of multicasting groups 2010 based on the sets F. For example, a multicasting group 2010 can be formed for each set F containing more than one certificates. In another embodiment, each set F containing more than one certificate is assigned a priority Prt(F) representing the expected frequency, or the expected number of periods pi, in which F will be an element of CCR. Multicasting groups 2010 are created only for the sets F of high priorities Prt. The priority Prt(F) may depend on, and be proportional to, a number of factors including, without limitation:
1. The number of certificates in the set F.
2. The sum Σ(D2−pi) of the remaining unexpired periods of the certificates in the set F in the period pi, or ΣD2. If D2 is infinite, some predefined high number can be assigned to the corresponding certificate when computing this sum.
When a system 110 joins the network, the system 110 may form a multicasting group 2010 for one or more of such sets F, e.g. for at least one set F containing a certificate 104 owned by the operator of system 110. Alternatively, the multicasting groups can be formed in advance, and the system 110 joining the network may subscribe to one or more, and possibly all, of multicasting groups 2010 associated with sets F containing the certificate corresponding to the system 110.
The invention is not limited to ad hoc networks.
Caching Validity Proofs
In ad hoc networks and other networks in which at least some of systems 110 serve as routers, when a verifier system 110 (e.g. 110.1) needs a certificate's validity status (possibly for a certificate not owned by the operator of system 110.1), the verifier's request 150 (
1. The number of certificates in the set F.
2. The sum Σ(D2−pi) of the remaining unexpired periods of the certificates in the set F in the period pi, or ΣD2. If D2 is infinite, some predefined high number can be assigned to the corresponding certificate when computing this sum.
When system 110.1 receives a proof 1810 (step 2120 in
The caching is not limited to the cases when the validity proofs are pushed to the certificate owners. Validity proofs 1810 may be identical to proofs 830 of
In some embodiments, the caching scheme of
Dynamic Introduction of New Certificates
It is desirable to allow addition of new certificates after the setup stage (e.g. the stage illustrated in
If hash chains (1) are used, the target c0=fTd(x) where x is the seed. The validity proof provided to verifiers 110 includes a token ci and the number j of times that the function f is to be applied to the i-token ci(F). The verifier checks that fj(ci)=c0.
The periods 1 though Tv will be called a “Tv window” herein. If a certificate is created in this window, it will expire on or before Td, and hence can be used with the data structures (hash chains (1) or hash trees 1210) created by CA 120 at step 918. More generally, a Tv window is a window in which a new certificate can be accommodated by the existing data structures for a predefined time interval of a number of the pi periods, i.e. the certificate will not last beyond the last pi period provided for by the data structures. In the embodiment of
A “Td window” starts at the same time as a corresponding Tv window and ends Td periods later, denoting the outside bounds of the maximum validity period of each new certificate created in the Tv window.
If CA 120 has not run out of empty slots until the end of the current Tv window, step 2360 is performed at the end of the window.
CA 120 and directories 210 provide the validity and proofs both for the current Td window and for all the previous Td windows which have not yet expired. When a Td window expires, the corresponding data structures can be re-used. The data structures associated with an expired certificate (e.g. the hash chains (1) or the hash trees 1210) can be re-used even before the corresponding Td window expires.
The invention is not limited to hash trees 1210 or hash chains.
Distributed Certificate Authority
A single CA 120 can be vulnerable to denial of service attacks and can be unavailable in the case of network errors. In some embodiments, a Distributed Certificate Authority (distCA) described below minimizes those drawbacks while maintaining the security of the overall system.
A distCA includes a CA 120 (
1. The verifier 110 can get a validity or invalidity proof for a certificate from any Sub-CA 2610. Therefore, in the case of network failures, chances are higher to obtain the proof of validity or invalidity.
2. Less vulnerability to successful denial-of-service attacks in each period pi because, as described below, less data need to be exchanged with CA 120 in each period pi (the decryption key distributed to Sub-CAs 2610 can be smaller than all of the period data such as i-tokens).
3. Compromising one of Sub-CAs 2610 does not compromise the whole system.
4. A compromised Sub-CA 2610 can be reactivated after it is back in control of CA 120.
5. The additional computational work performed by the verifier to validate a certificate is small compared to a digital signature computation.
Some embodiments do not provide all of the advantages described above. Some embodiments provide additional advantages, as described below.
This embodiment does not use revocation targets N1 for the Sub-CAs. Lack of a validity proof is taken as proof that the Sub-CA is invalid. Other embodiments use explicit revocation proofs for the Sub-CAs. As mentioned above, in some embodiments an invalid Sub-CA can be re-activated in a subsequent period pi. Therefore, in some embodiments, the revocation proof is constructed using a technique suitable for validity proofs. For example, trees like 2614, 2618 can be used to prove a Sub-CA invalidity in each period pi.
At step 2718, CA 120 creates data structures for certificate validation and revocation. Any of the certificate validation and revocation techniques can be used, including the techniques described above with respect to
CA 120 also defines a hash tree 2630 (e.g. like in
The certificate validation structures may also include structures common for the Sub-CAs, e.g. structures defining a complement cover CC(U) if the same complement cover is used for all the Sub-CAs.
In some embodiments, a separate structure 2622.j is created for each set F of a complement cover rather than each certificate. This could be done, for example, for the systems of
At step 2722, CA 120 generates encryption and decryption keys 2640.j for each Sub-CA 2610.j. In this embodiment, for each Sub-CA 2610.j, one decryption key is used for a number of consecutive periods pi. For example, suppose there are 365 periods pi subdivided into P=25 logical partitions. 365/25=14 10/25, so we can place the periods p1 through p15 into Partition 1, periods p16 through p30 into Partition 2, and so on. The last Partition 25 has periods p361 through p365. The invention is not limited to the partitions having any particular number of periods. For each Sub-CA 2610.j and each partition k, CA 120 generates a decryption key DK.j.k. The encryption scheme may be symmetric (i.e. the encryption keys may be the same as the decryption keys) or asymmetric.
At step 2726, CA 120 uses the data 2622, 2626 to create certificate validation data 2650.j for each Sub-CA 2610.j. CA 120 encrypts the data 2650.j and sends it to the respective Sub-CAs 2610.j. For each j, data 2650.j includes a data structure 2660 for each period pi. In some embodiments, the data 2660 are the same as the data 830V (
For each Sub-CA 2610.j and each partition k, the data 2660 are encrypted for decryption with the key DK.j.k.
Optionally, for each certificate 104, CA 120 distributes to each Sub-CA 2610.j, and/or the certificate owner and/or other parties, data 2662 needed to compute the validation target RC from the corresponding target Yj. Data 2662 may include Value(CoNodes(Leaf(Yj))) where Leaf(Yj) is the tree 2626 leaf corresponding to Sub-CA 2610.j. Optionally, for each certificate 104, CA 120 distributes to each Sub-CA 2610.j, and/or the certificate owner and/or other parties, data 2664 needed to compute the revocation target NC from the corresponding target N1j. Data 2664 may include Value(CoNodes(Leaf(N1j))) where Leaf(N1j) is the tree 2630 leaf corresponding to Sub-CA 2610.j. Optionally, CA 120 distributes to each Sub-CA 2610.j, and/or each certificate owner and/or other parties, data 2668 needed to compute the target Rsub from the corresponding target Wj. Data 2664 may include Value(CoNodes(Leaf(Wj))) where Leaf(Wj) is the tree 2618 leaf corresponding to Sub-CA 2610.j. Data 2662, 2664, 2668 need not be encrypted. Optionally, CA 120 can make public all or some of validation targets Wj, Yj, N1j. CA 120 can also provide these targets, for each j, to the respective Sub-CA 2610.j.
At, or shortly before, the start of each period pi, if pi is the first period in a partition k, CA 120 distributes, for each j, the decryption key DK.j.k to Sub-CA 2610.j. If a Sub-CA has been compromised and has not returned under control of CA 120, CA 120 may withhold the decryption key from the CA.
If a Sub-CA 2610.j is compromised, the adversary may have access to all the data structures 2650.j. However, the data 2660 are encrypted, and the adversary will hopefully not be able to decrypt the data 2660 for the future partitions. Therefore, once the Sub-CA 2610.j is back under the CA's control, CA 120 can reactivate the Sub-CA in the first period pi of the next partition. Reactivation implies that CA 120 will (a) provide the validity proof 2672 for the Sub-CA and (b) will provide the Sub-CA with (b1) the decryption key DK.j.k for the current partition k, and (b2) the decryption keys for all the preceding partitions if the keys were withheld when the Sub-CA was out of the CA's control and if the data 2660 for the preceding partitions are needed to construct certificate validity proofs.
At, or shortly before, each period pi, CA 120 sends to each Sub-CA 2610.j, and/or each certificate owner and/or other parties, a validity proof 2672 to prove the Sub-CA validity in the period pi. Validity proof 2672 includes Value(gv(i)) and Value(CoNodes(gv(i))) for the corresponding tree 2614.j. In addition, CA 120 sends to each Sub-CA 2610.j the revocation seed values N0j for each certificate 104 which has been revoked in the previous period pi (or in any of the previous periods pi).
At, or shortly before, each period pi which is the first period of a partition, each Sub-CA 2610 which has not been compromised uses the corresponding decryption key to decrypt the validation data 2660 for the partition.
In a period pi, a verifier 110 (
Certificate validity proof 2820 includes a validity proof 830. The validity proof 830 is constructed from decrypted data 2660. The form of proof 830 depends on the underlying certificate validity scheme, and may be as in
Certificate revocation proof 2830 includes a revocation seed N0j and data 2664. If the verifier receives revocation proof 2830, the verifier computes the revocation target N1j=f(N0j), and then the target NC from the data 2664. If the computed NC value matches the NC value in the certificate, the certificate is assumed to be revoked. Otherwise, the verifier may still assume the certificate to be revoked, or seek another proof, for example by contacting another Sub-CA.
In each case when a target is to be computed from the co-node values (e.g. when the target Wj needs to be computed), if the verifier has already verified the same certificate for a previous period pj and has cached the co-node values, and the grey vertices for the period pj and the current period pi are in a common sub-tree whose root value has been cached as the value of one of the co-nodes, then the verifier needs only to compute the root value for the sub-tree and compare it to the cached value rather than computing the root value for the tree. For example, if a sub-tree root value has been cached for a tree 2614.j, the verifier does not need to compute Wj or Rsub (provided that the verifier has all the pertinent information on the tree structure and not just a listing of type (5)).
Some embodiments use directories 210 (
In some embodiments, each partition consists of one period pi. Different Sub-CAs 2610 have different decryption keys DK.j.i for each period pi. Sub-CA validity verification is then omitted. If a Sub-CA is compromised, CA 120 revokes the Sub-CA by stopping to send the decryption keys to the Sub-CA. When the Sub-CA is back in control of the CA, the CA can reactivate the Sub-CA by sending to it the retained decryption keys. Hence, the data 2614, 2618, 2668 can be omitted.
The invention is not limited to the embodiments described above. The invention is not limited to any particular hash functions, or to cryptographic functions (which are easy to compute but are one-way or collision resistant). In some embodiments, it is desirable that a function f or H be collision resistant not in the sense that it is difficult to find different x and y with the same image but in the sense that if x and y are uniformly drawn from the function's domain, the probability is small that they both will have the same image:
P{H(x)=H(y)}≦α
where α is a small constant (e.g. 1/10, or 1/100, or 2−25, or 2−50, or 2−80, or 2−16, or some other value). Some or all of the techniques used for validity proofs can also be used for invalidity proofs and vice versa. The CA, the Sub-CAs, the directories and the systems 110 may include software-programmable or hardwired computer systems interconnected via a network or networks. Each function f or H represents an evaluation method performed by a computer system. The invention is not limited to the step sequences shown in the flowcharts, as the step order is sometimes interchangeable and further different steps may be performed in parallel. Other embodiments and variations are within the scope of the invention, as defined by the appended claims.
All of the following references are incorporated herein by reference.
The present application is a divisional of U.S. patent application Ser. No. 11/218,093, filed Aug. 31, 2005, which claims priority of U.S. Provisional Application No. 60/606,213 filed on Aug. 31, 2004, both of which are incorporated herein by reference.
| Number | Name | Date | Kind |
|---|---|---|---|
| 5666416 | Micali | Sep 1997 | A |
| 5687235 | Perlman | Nov 1997 | A |
| 5699431 | Van Oorschot | Dec 1997 | A |
| 5717757 | Micali | Feb 1998 | A |
| 5717758 | Micall | Feb 1998 | A |
| 5793868 | Micali | Aug 1998 | A |
| 5903651 | Kocher | May 1999 | A |
| 5960083 | Micali | Sep 1999 | A |
| 6044462 | Zubeldia | Mar 2000 | A |
| 6097811 | Micali | Aug 2000 | A |
| 6128740 | Curry | Oct 2000 | A |
| 6141347 | Shaughnessy et al. | Oct 2000 | A |
| 6226743 | Naor et al. | May 2001 | B1 |
| 6292893 | Micali | Sep 2001 | B1 |
| 6301659 | Micali | Oct 2001 | B1 |
| 6381695 | Kudo et al. | Apr 2002 | B2 |
| 6381696 | Doyle | Apr 2002 | B1 |
| 6385608 | Mitsuishi et al. | May 2002 | B1 |
| 6397329 | Aiello et al. | May 2002 | B1 |
| 6442689 | Kocher | Aug 2002 | B1 |
| 6487658 | Micali | Nov 2002 | B1 |
| 6532540 | Kocher | Mar 2003 | B1 |
| 6766450 | Micali | Jul 2004 | B2 |
| 7260572 | Min et al. | Aug 2007 | B2 |
| 20010034833 | Yagasaki et al. | Oct 2001 | A1 |
| 20020046337 | Micali | Apr 2002 | A1 |
| 20020165824 | Micali | Nov 2002 | A1 |
| 20020184504 | Hughes | Dec 2002 | A1 |
| 20030217265 | Nakano et al. | Nov 2003 | A1 |
| 20030221101 | Micali | Nov 2003 | A1 |
| 20030236976 | Wheeler | Dec 2003 | A1 |
| 20040049675 | Micali | Mar 2004 | A1 |
| 20040128504 | Kivinen | Jul 2004 | A1 |
| 20040148505 | Qiu | Jul 2004 | A1 |
| 20050053045 | Chmora | Mar 2005 | A1 |
| 20050055548 | Micali | Mar 2005 | A1 |
| 20050081037 | Kumagai | Apr 2005 | A1 |
| 20050278534 | Nadalin et al. | Dec 2005 | A1 |
| 20060129803 | Gentry | Jun 2006 | A1 |
| Number | Date | Country |
|---|---|---|
| 0 932 109 | Jul 1999 | EP |
| 11-289329 | Oct 1999 | JP |
| 2001-265216 | Sep 2001 | JP |
| 2008-524931 | Jul 2008 | JP |
| 9716905 | May 1997 | WO |
| 2005029445 | Mar 2005 | WO |
| WO200502944 | Mar 2005 | WO |
| 2006066143 | Jun 2006 | WO |
| Number | Date | Country | |
|---|---|---|---|
| 20090265547 A1 | Oct 2009 | US |
| Number | Date | Country | |
|---|---|---|---|
| 60606213 | Aug 2004 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 11218093 | Aug 2005 | US |
| Child | 12492898 | US |
