This application is a continuation of U.S. patent application Ser. No. 09/680,501 filed on Oct. 5, 2000.
The present invention relates generally to cryptographic schemes, and specifically to cryptographic schemes relating to wireless applications.
BACKGROUND OF THE INVENTION Information security is required to secure many types of transactions performed electronically using a wide range of computing and communication technologies. As consumers demand more flexible, convenient services, technologies such as wireless networks, paging infrastructures and smart cards are being deployed to support critical, information sensitive applications including account inquiries, electronic cash, secure communications and access control. One of the key features of each of these technologies is that they offer consumers the convenience of service anywhere, any time. The convenience offered to consumers results in a challenge for the vendors to create smaller and faster devices while providing a high level of security for information computed and transmitted.
Information security is provided through the application of cryptographic systems (commonly referred to as cryptosystems). The two main classes of cryptosystems are symmetric and public key. In a symmetric cryptosystem, two users wishing to participate in a secure transaction must share a common key. Therefore, each user must trust the other not to divulge the key to a third party. Users participating in a secure transaction using public key cryptosystems will each have two keys, known as a key pair. One of the keys is kept secret and is referred to as the private key, while the other can be published and is referred to as the public key. Typically, applications use a combination of both these classes of cryptosystems to provide information security. Symmetric technologies are typically used to perform bulk data encryption, while public key technologies are commonly used to perform key agreement, key transport, digital signatures and encryption of small messages.
Since the introduction of public key cryptosystems, there have been many implementations proposed. All of these public key systems are based on mathematical problems which are known to be hard, that is, it is thought that breaking a system is equivalent to solving a hard mathematical problem. These problems are generally easy to solve for numbers that are small in size, but become increasingly difficult as lager numbers are used. One of the differences among the systems is how large the numbers have to be so that the system is too hard to solve given present and anticipated computing power. This is typically linked to the length of the key and referred to as the key size. A system using a small key size while maintaining a high level of security is considered better, as it requires less information to be transmitted and stored.
Diffie-Hellman key agreement provided the first practical solution to the key distribution problem by allowing two parties to securely establish a shared secret over an open channel. The original key agreement protocol provides unauthenticated key agreement. The security is based on the discrete logarithm problem of finding integer x given a group generator α, and an element β, such that αx=β.
Rivest Shamir Adleman (RSA) was the first widely deployed realization of a public key system. The RSA system is a full public key cryptosystem and can be used to implement both encryption and digital signature functions. The security of the RSA cryptosystem depends on the difficulty of factoring the product of two large distinct prime numbers. To create a private key/public key pair, a user chooses two large distinct primes P and Q, and forms the product n=PQ. With knowledge of P and Q, the user finds two values e and d such that ((M)e)dmod n=M.
The public key of the user is the pair (e, n) while the private key is d. It is known that the recovery of d from and e and n requires the recovery of P and Q, and thus is equivalent to factoring n.
Elliptic curve cryptosystems are based on an exceptionally difficult mathematical problem, Thus, elliptic curve systems can maintain security equivalent to many other systems while using much smaller public keys. The smaller key size has significant benefits in terms of the amount of information that must be exchanged between users, the time required for that exchange, the amount of information that must be stored for digital signature transactions, and the size and energy consumption of the hardware or software used to implement the system. The basis for the security of the elliptic curve cryptosystem is the assumed intractability of the elliptic curve discrete logarithm problem. The problem requires an efficient method to find an integer k given an elliptic curve over a finite field, a point P on the curve, another point Q such that Q=kP.
In this system, the public key is a point (Q) on an elliptic curve (represented as a pair of field elements) and the private key is an integer (k). Elliptic curves are defined over an underlying field and may be implemented over the multiplicative group Fp, (the integers modules a prime p) or characteristic 2 finite fields (F2m where m is a positive integer).
There are typically three levels in a cryptosystem, which are encryption, signatures, and certificates. These three levels can be implemented using to above mentioned systems or a combination thereof.
The first level of a cryptosystem involves encrypting a message between correspondent A and correspondent B. This level is vulnerable to attack since there is no way for correspondent A to verify whether or not correspondent B sent the message, or if a third party in the guise of correspondent B sent the message.
Therefore, the second level of signing a message was introduced. Correspondent B can sign the encrypted message using, for example, a hashing function to hash the original message. If correspondent A uses the same hashing function on the decrypted message and it matches the signature sent by correspondent B, then the signature is verified. However, a third party may act as an interloper. The third party could present itself to correspondent A as if it were correspondent B and vice versa. As a result, both correspondents would unwittingly divulge their information to the third party. Therefore, the signature verifies that the message sent by a correspondent is sent from that correspondent, but it does not verify the identity of the correspondent.
To prevent this type of attack, the correspondents may use a trusted third party (TTP) to certify the public key of each correspondent. The TTP has a private signing algorithm and a verification algorithm assumed to be known by all entities. The TTP carefully verifies the identity of each correspondent, and signs a message consisting of an identifier and the correspondent's public key. This is a simple example as to how a TTP can be used to verify the identification of the correspondent.
Some of the most significant emerging areas for public key cryptosystems include wireless devices. Wireless devices, including cellular telephones, two-way pagers, wireless modems, and contactless smart cards, are increasing in popularity because of the convenience they provide while maintaining a low cost and small form factor.
However, implementing the above mentioned cryptosystems requires computational power, which is limited on such wireless devices. Therefore, there is a need for a cryptosystem that provides all of the advantages as described above, but requires less power from the wireless device.
SUMMARY OF THE INVENTION In accordance with the present invention there is provided a method of communicating between a pair of correspondents through an intermediary comprising the steps, registering one of said correspondents with said intermediary to share an identifier, preparing at said one correspondent a secure communication including a message between said correspondents, preparing a signature component including a derivation of said secure communication and said identifier forwarding said signature component to said intermediary and verifying said signature component at said intermediary, attaching to said communication a certificate of the public key and identity of the said one correspondent, and forwarding said communication and certificate to said other correspondent. BRIEF DESCRIPTION OF THE DRAWINGS
An embodiment of the invention will now be described by way of example only with reference to the following drawings in which:
FIG. 1 is a schematic drawing of a pager system;
FIG. 2 is a representation of a registration process for the ofFIG. 1
FIG. 3 is a representation of a message transfer system for the system ofFIG. 1
FIG. 4 is a schematic representation of an alternative embodiment of a communication system.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT For convenience, like numerals in the description refer to like structures in the drawings. Further, although the description refers only to pagers, it is intended that the description includes wireless devices in general.
Referring toFIG. 1, a paging system is represented generally by the numeral100. Afirst pager102 is operatively coupled with afirst home terminal104 through a wireless communication. Thefirst home terminal104 is operatively coupled to asecond home terminal106 via anetwork108 and thesecond home terminal106 in turn is operatively coupled to asecond pager110. Thepagers102,110 are typically coupled to theirrespective home terminals104,106 by radio frequency. Thenetwork108 is typically a public switched telephone network (PSTN), but can include a data network, and the Internet.
Before apager102 can communicate with thehome terminal104 it must be registered. Everypager102 contains a subscriber unit address and a public keyCof the pager manufacturer or service provider (herein referred to as the company public key). This information is loaded at the manufacture stage. The company public key QCis derived from a company private key dC.
Eachhome terminal104 has a private key dHand a public key QH. The public key QNis signed by the company private key dCto create a certificate denoted CM. The company public key QCcould be system wide or defined for a given region. A subscriber purchases apager102 from a retail outlet and the pager is then loaded with a home index112 and identifier ID using the protocol outlined below. The home index is typically a 32-bit index which uniquely identifies thepager102 and correlates it with aspecific home terminal104.
The subscriber calls a number, typically a toll-free number, to contact a service provider and ahome terminal104 is assigned. Thehome terminal104 sends thepager102 its public key QHand its certificate CM. The pager verifies QH, with the company public key QC. The pager generates a private key dpand a corresponding public key Qpwhich is communicated to thehome terminal104. Thepager102 sends to thehome terminal104 the necessary authorization information (including identification, credit card number, subscriber unit address, and the like) encrypted under the home terminal public key QH). The home terminal gets authorization from a central repository that this subscriber unit has not already been activated and thereby prevents counterfeiting of subscriber units. Thehome terminal104 sets up a subscriber account and sends thepager102 its home index and identifier ID encrypted under Qpand signed by the home terminal.
Eachpager102 in apaging infrastructure100 is registered with a home terminal using the registration protocol described above. The pagers have a private and public key pair, dp,Qp, each of which are approximately 20 bytes in length. Thehome terminals104 have a private and public key pair dh, QHeach of which are approximately 25 bytes in length. It is desirable to have a longer key length at the home terminal for providing additional security. Further, since thehome terminal104 does not have the same power constraints as thepager102, the extra computational power required for the longer key is not a significant issue. The additional security at thehome terminal102 is important since a compromise of the home terminal would permit counterfeiting of subscriber units.
To reduce the computational requirements on the pager thereby reducing the power required to encrypt a message M, each of thepagers102 has a certificate registered for it at thehome terminal104. The certificate, certca, validates the public key Qp, and identity ID. Each of the home terminals maintains a table for the pagers and their associated certificate. Rather than having the pager sign the certificate and send the message to the home terminal, the certificate certcais signed by the pager's home terminal. The transmission process used to implement such a protocol is described in detail below.
Referring once again toFIG. 1 andFIG. 3, the first pager P1wishes to send a message M to a recipient, e.g. a second pager P2having a public key QP1. The sender P1initially obtains an authentic copy of a recipient's public key QP2. The first pager P1calculates ciphertext with of a signed message M such that W=EQP2(SP1(M)), where EQP1is encryption under the public key QP1and SP1is the signature of the first pager on message M using the private key dp.
The first pager also calculates a signature ma=SP1(h(w)||CN||IDP1) where h(w) is a hash of W, such as SHA-1. CN is a timestamp or some other nonce, IDP1is the unique identifier of the first pager, and || represents concatenation. The first pager then transmits the signature, ma, and the signed, encrypted message, W, to the first home terminal.
The signature, ma, is used by thehome terminal104 associated with pager P1to verify that P1is a legitimate user. In order to avoid a challenge-response authentication to save time and bandwidth, the message W and a nonce CN, which is unique for each transmission, are coupled with the ID of P1and signed. The nonce is used to prevent replay of the transmission, W is a signed, encrypted form of the message M. Signing then encrypting is preferred over encrypting then signing.
The first home terminal receives maand W from P1and uses mato verify that P1is a legitimate user. IDP1is recovered from ma, and the first home terminal retrieves the certificate, certcafor P1from the corresponding table and attaches it to W. Certcais a full certificate such as X.509 and consists of 1 bytes. There is no loss of security in storing the certcacertificates at the first home terminal.
In addition to saving computational power on the pager, the bandwidth requirement of the transmission from the pager to the base are reduced since the pager does not have to transmit a certificate.
Thefirst home terminal104 stores a pre-computed table of values which allows it to increase the speed of verifying P1's signature. Alternately, if verification is fast enough, as would be the case with a hardware implementation, the table of values is not required.
The first home terminal then removes the signature component Maand transmits the signed, encrypted message W and the certificate Certcato the recipient. Since the recipient in this example is thesecond pager110, W and Certcaare sent to thesecond home terminal106 that has public and private keys QP3dP3respectively.
The second home terminal,106 receives the transmission and verifies QP1using Certca(QP1, IDP1). To save bandwidth, thesecond home terminal106 signs QP1according to the signature function Sdp1(W||QP1||IDP1) and sends it along with W to P2. A time stamp CNlmay be included to prevent replay attacks. P2trusts the second home terminal to do this honestly. The pager P2can then verify W and recover the message M using its private key dP2and the senders public key QP1. QP1has been validated by the signature of thehome terminal104 and therefore communicating between thesecond home terminal106 and thesecond pager110 in this manner keeps the certificates off the transmission channel and reduces bandwidth requirements.
An example of the bandwidth requirements for such a method is described as follows. Suppose M consists of t bytes. If the Nyberg-Rueppel protocol is used for signing the message, t+20 bytes are required for SP1(M). A further 20 bytes a used to encrypt SP1(M), therefore W is t+40 bytes in length. Hashing h(W) uses 20 bytes if SHA-1 is used. The nonce CN uses 4 bytes and the identification IDP1uses 4 bytes. Once again, if Nyberg-Rueppel is We for signing, 20 additional bytes are used. Hence mawill be 48 bytes. Therefore, the transmission between the first pager and the first home terminal uses t+92 bytes.
For the transmission from the first home terminal to the second home terminal, W uses t+40 bytes, Certcauses l bytes, and therefor the bandwidth required is t+l+40 bytes.
For the transmission from the second home terminal, W uses t+40 bytes, QP1uses 20 bytes, IDP1uses 4 bytes, and CN1uses 4 bytes. Therefore, using Nyberg-Rueppel for signing, the bandwidth used in sending W and Sdp3(W||QP1||IDP1) and the nonce CN1is a total of 25+(t+40)+20+4+4=t+93 bytes.
In the above example, the transmission is from pager to pager. However, the protocol may be used from the input devices, for example, a DTMF telephone as illustrated inFIG. 4. In this case, the transmission T, would be With and Certca(Qd; IDD) where QDand IDDare the public key and identity of the telephone.
The transmission T2 would be W and certca(Qd; IDD) and the transmission T3 to the pager, after verification of Certcawould be QD, With IDDand CN all signed by the home terminal.