AU2771500A - Cyclotomic polynominal construction of discrete logarithm cryptosystems over finite fields - Google Patents

Description

Regulation 3.2

AUSTRALIA

PATENTS ACT, 1990 COMPLETE SPECIFICATION r r r cc u FOR A STANDARD PATENT

ORIGINAL

e cc Name of Applicant: Actual Inventor: Address for service in Australia: Invention Title: CITIBANK, N.A.

LENSTRA, Arjen, K A J PARK, Level 11, 60 Marcus Clarke Street, Canberra ACT 2601 CYCLOTOMIC POLYNOMIAL CONSTRUCTION OF DISCRETE LOGARITHM CRYPTOSYSTEMS OVER THE FINITE FIELDS The following statement is a full description of this invention, including the best method of performing it known to me/us WO 98/36526 PCT/US97/17304 -1- CYCLOTOMIC POLYNOMIAL CONSTRUCTION OF DISCRETE LOGARITHM CRYPTOSYSTEMS OVER FINITE

FIELDS

BACKGROUND OF THE INVENTION The present invention relates to data security, encryption, and, generating and using electronic signatures to verify the identity of a communicating party.

Most public key cryptosystems involve either a factoring problem or a discrete logarithm (DL) problem. The factoring problem is, given a not-prime number, find its complete factorization into prime numbers.

10 The DL problem is, given a group G generated by g and an element h in G, t find an integer m such that g" h, that is, evaluate log, h. Several proposed schemes for public key cryptosystems rely on the computational difficulty of finding a DL in a multiplicative group of a finite field. Pu b l i c k e y 15 cryptosystems encompass public key encryption schemes and digital signature schemes. Assume each user has a public key and a private key, which is not necessarily true for all schemes, and that party A wishes to send a secure message to party B. In a public key encryption scheme, party A uses party B's public key to encrypt, and then party B uses its own public and private 2 keys to decrypt. In a digital signature scheme, party A uses its own public and private keys to prepare the message, and party B uses party A's public key to receive the message. That is, to prepare the message, in a public key encryption scheme, the sending party uses the receiving party's key information, whereas in a digital signature scheme, the sending party uses its own key information. To receive the message, in a public key encryption scheme, the receiving party uses its own key information, whereas in a digital signature scheme, the receiving party uses the sending party's key information.

Typical digital signature schemes have three steps: system setup, signature generation by a sending party, and signature verification by a receiving party.

System setup is assumed to occur well before signing or encryption of a message. Generally, during system setup of a DL based public key cryptosystem, a prime number is selected and used to obtain a generator key cryptosystem, a prime number is selected and used to obtain a generator WO 98/36526 PCT/US97/1I7304 -2for a group, then a random number is selected and used as an exponent for the generator to produce a resulting value in the finite field. Determining the random number when only the generator and resulting value are known is a DL problem.

The outcomes of system setup are a public key and a private key. A public key is assumed to be public knowledge and comprises the prime number, the generator, the resulting value and possibly other parameters. A private key is assumed to be known only to the sending party, and comprises the random number.

During signature generation of a DL based public key cryptosystem, a second random number is chosen and used as an exponent for the generator to produce a second resulting value in the finite field.

Determining the second random number when only the generator and second resulting value are known is a DL problem. Then a third value based on the private key, on the message to be signed, and second resulting value is obtained. The outcome of signature generation is a digital signature including the third value and at least one other parameter.

During signature verification of a DL based public key cryptosystem, the public key and third value portion of the signature are exponentially combined to produce a fourth result. If the fourth result is equal to at least one other parameter of the signature, then the signature is considered valid.

The exponentiation portions of system setup, signature generation and signature verification are computationally expensive and time consuming. Techniques are sought which will reduce the computational burden to an authorized user, particularly during signature generation, while maintaining computational difficulty for an unauthorized user.

SUMMARY OF THE INVENTION In accordance with an aspect of this invention, a method of and an apparatus for determining public and private keys for a public key cryptosystem comprises selecting a first prime number, obtaining a cyclotomic polynomial evaluated at the first prime number, obtaining a second prime WO 98/36526 PCT/US97/17304 -3number which is a factor of the cyclotomic polynomial evaluated at the first prime number, finding a generator of a subgroup of a multiplicative group of a finite field, the order of the subgroup being the second prime number, obtaining a public value based on the generator and a selected integer, forming the public key to include the first and second prime numbers, the generator and the public value, and forming the private key to include the selected integer.

In accordance with a further aspect of this invention, the finite field may be represented with an optimal normal basis.

In accordance with a different aspect of this invention, the second prime number q satisfies (log, q) 1 B, where B is a predetermined number of bits.

In accordance with another aspect of this invention, a control integer t' is selected, and the cyclotomic polynomial is the t'-th cyclotomic polynomial, and the public key includes the control integer t'.

In accordance with still another aspect of this invention, a method of generating a digital signature for a message additionally selects a second integer, obtains a first signature value based on the second integer and the generator, obtains a second signature value based on the first signature value and the message, and forms the digital signature to include the first and second signature values.

A method of verifying a thus-formed digital signature for a message finds an inverse integer which is the inverse of the second signature value, finds a first intermediate value based on the inverse integer and the message, finds a second intermediate value based on the inverse integer and the first signature value, finds a third intermediate value based on the generator, the public value, and the first and second intermediate values, and determines that the signature is valid when the third intermediate value is equal to the first signature value.

A method of determining a shared key for a public key cryptosystem selects a first prime number, obtains a cyclotomic polynomial evaluated at the first prime number, obtains a second prime number which is a factor of the cyclotomic polynomial evaluated at the first prime number, WO 98/36526 PCT/US97/17304 -4finds a generator of a subgroup of a multiplicative group of a finite field, the order of the subgroup being the second prime number, selects an integer, receives an intermediate value which is based on the generator, and forms the shared key as a function of the intermediate value and the integer.

A method for secure communication of a message selects a first prime number, obtains a cyclotomic polynomial evaluated at the first prime number, obtains a second prime number which is a factor of the cyclotomic polynomial evaluated at the first prime number, finds a generator of a subgroup of a multiplicative group of a finite field, the order of the subgroup being the second prime number, selects an integer, receives an intermediate value which is based on the generator, forms the shared key as a function of the intermediate value and the integer, and encrypts the message using the shared key.

A method for secure communication of a message receives an encrypted message which has been encrypted using a shared key formed as a function of an intermediate value and a selected integer, the intermediate value being based on a generator of a subgroup of a multiplicative group of a finite 20 field, the order of the subgroup being a second prime number which is a factor of a cyclotomic polynomial evaluated at a first prime number, and decrypts the encrypted message using the shared key.

A method for secure communication of a message selects a first prime number, obtains a cyclotomic polynomial evaluated at the first prime number, obtains a second prime number which is a factor of the cyclotomic polynomial evaluated at the first prime number, finding a generator of a subgroup of a multiplicative group of a finite field, the order of the subgroup 30 being the second prime number, obtains a public value based on the generator and a first integer, selects a second integer, finds a first encrypted value based on the generator and the second integer, finds a second encrypted value based on the message, the public value and the second integer, and forms an encrypted message from the first and second encrypted values.

A method for secure communication of a message receives an encrypted message formed of a first encrypted value and a second encrypted WO 98/36526 PCT/US97/17304 0 value, the first encrypted value being based on a first integer and a generator of a subgroup of a multiplicative group of a finite field, the order of the subgroup being a second prime number which is a factor of a cyclotomic polynomial evaluated at a first prime number, the second encrypted value being based on the message, the first integer and a public value based on the generator and a second integer, finds a first intermediate value based on the first encrypted value and a private key, the private key being based on the generator, and decrypts the encrypted message based on the second encrypted value and the first intermediate value.

It is not intended that the invention be summarized here in its entirety. Rather, further features, aspects and advantages of the invention are set forth in or are apparent from the following description and drawings.

BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1A is a flowchart illustrating system setup according to the ElGamal scheme; Fig. 1B is a flowchart illustrating signature generation according to the ElGamal scheme; 20 Fig. IC is a flowchart illustrating signature verification according to the ElGamal scheme; Fig. 2A is a flowchart illustrating system setup according to the Schnorr and DSA schemes; Fig. 2B is a flowchart illustrating signature generation according to the Schnorr scheme; Fig. 2C is a flowchart illustrating signature verification according to the Schnorr scheme; Fig. 2D is a flowchart illustrating signature generation according to the DSA scheme; Fig. 2E is a flowchart illustrating signature verification according to the DSA scheme; Fig. 3A is a flowchart illustrating system setup according to the ECDSA scheme; Fig. 3B is a flowchart illustrating signature generation according WO 98/36526 PCT/US97/17304

S

-6- 0 to the ECDSA scheme; Fig. 3C is a flowchart illustrating signature verification according to the ECDSA scheme; Fig. 4A is a flowchart illustrating system setup according to the present invention; Fig. 4B is a flowchart illustrating signature generation according to the present invention; Fig. 4C is a flowchart illustrating signature verification according to the present invention; Fig. 4D is a table of cyclotomic polynomial coefficients; Fig. 4E is a flowchart illustrating a DES system setup according to the present invention; Fig. 4F is a flowchart illustrating encryption for the DES system setup according to the present invention; Fig. 4G is a flowchart illustrating decryption for the DES system setup according to the present invention; Fig. 4H is a flowchart illustrating encryption for the ElGamal system setup according to the present invention; Fig. 4J is a flowchart illustrating decryption for the ElGamal system setup according to the present invention; Fig. 5A is a table of results for comparing signature generation performance of schemes for public key cryptosystems; Fig. 5B is a table of results for comparing signature verification performance of schemes for public key cryptosystems; Fig. 6 is a chart showing the message encrypted and decrypted to obtain the performance results of Figs. 5A and Figs. 7A-11D are charts showing the public key, private key, signature, and signature generation parameter k, for each of the public key cryptosystems in the examples used to obtain the performance results of Figs.

and 5B; and Fig. 12 is a block diagram of an environment in which the present invention may be implemented.

S

WO 98/36526 PCT/US97/17304 -7- DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Cyclotomic polynomials are used to construct subgroups of multiplicative groups of finite fields that allow very efficient implementation of discrete logarithm based public key cryptosystems, including public key encryption schemes and digital signature schemes. A field is represented with an optimal normal basis, and a generator of a subgroup of the multiplicative group of the field is used to form a public key. Depending on the type of application and implementation, public key encryption according to the cyclotomic scheme may be up to three times faster than schemes using more conventional choices of subgroups or finite fields.

Proposed digital signature schemes include the ElGamal scheme, as set forth in T. ElGamal, "A public key cryptosystem and a signature scheme based on discrete logarithms", IEEE Trans. Info. Tech., 31, 469-472, 1985; the Schnorr scheme, as set forth in C.P. Schnorr, "Efficient signature generation by smart cards", J. Cryptology, 4, 161-174, 1991; the digital signature algorithm (DSA) scheme, as set forth in U.S. Patent No. 5,231,668 (Kravitz), "Digital signature algorithm", July 27, 1993; and the elliptic curve digital signature algorithm (ECDSA) scheme, as set forth in Agnew et al., "An implementation for a fast public key cryptosystem", J. Cryptology, 3, 63-79, 1991. DSA is incorporated in the U.S. Government's Digital Signature Standard. These proposed schemes are discussed and compared with the present cyclotomic scheme as used in a digital signature scheme.

Terminology m a message to be signed, consisting of a binary string p a prime number q a prime factor of p-1 L the length in bits of p, practically, L determines the DL security level.

B the length in bits of q, practically, B determines the subgroup DL security level.

F(p) the field of p elements, represented by the set 1, least residues modulo p p-i} of least residues modulo p WO 98/36526 W09836526PCTIUS97IJ 7304 -8- F~p)-multiplicative group of F(p) 0 a collision resistant cryptographic hash function which maps binary strings to non-negative integers of at most a predetermined number of bits ElGamal System Setup Fig. IA shows the steps performed for each user during setup of an electronic signature system according to the ElGamal scheme. This process is performed by a processor of a general purpose digital computer.

Alternatively, this process may be performed by a special purpose printed circuit board used in conjunction with a general purpose computer, or by a "smart card", that is, a credit card sized portable device including a microprocessor.

At step 102, a prime number q having L-1 bits is chosen.

At step 104, a value p 2q 1 is computed.

At step 106, a test is made to determine whether p is prime.

Since q has L-1 bits, p has L bits.

If p is not prime, the process returns to step 102 and selects another prime number.

If p is prime, the process proceeds to step 108, and randomly selects an element g of the multiplicative group of the field F(p) of p elements.

At step 110, a test is made to determine whether g 2 1 and whether g" 1 in F(jp). If either of these tests fails, the process returns to step 108 and selects another element of as the element g.

If g *I1 and gq* 1 in then the element g is a generator for the multiplicative group F(p)y of the field Instead of the procedure depicted in steps 102-110, other procedures may be used to locate a field F(p) and a generator g.

After locating a generator g, the process proceeds to step 112, and randomly selects a value a in the range 2 :5 a :5 p-2.

At step 114, the process finds y in F(p).

The result of system setup is a public key and a private WO 98/36526 PCT/US97/17304 -9- 0 key The public key has length 3L bits. The private key has length L bits.

Finding the private key from the public key is a discrete logarithm (DL) problem in the field and is considered hard if p is big enough. Presently, suitable difficulty is presented when the length of p is L bits, L 1024, and when a prime factor of p-1 has at least 160 bits. As computational power becomes more available, these parameters will increase, to maintain the computational difficulty for an unauthorized user.

ElGamal Signature Generation Fig. 1B shows the steps performed by a party generating an electronic signature for a particular document according to the ElGamal scheme. The document is considered to be a bit string m. In practice, the generating party is a processor of a general purpose digital computer. In some 15 embodiments, the processor may be in a special purpose digital computer, such as a smart card.

At step 122, an integer k is randomly selected in the range 2 k p-2, with the greatest common divisor (GCD) of k and p-1 being 1, that is, k is selected so that it is co-prime with p-1.

20 0. At step 124, k' mod p-1 is found, that is, the value which satisfies 1 mod p-1.

At step 126, the value r g' in F(p) is obtained, with r in the range 1 r 5 p-1.

At step 128, the value s ar) mod p-1 is computed, with s in the range 0 s p-2. is a cryptographic hash function agreed upon by all users of the system. For example, a suitable standard cryptographic hash function is Secure Hash Algorithm SHA-1, as defined in FIPS 180-1, April 17, 1995, available from the National Technical Information Service, Springfield, Virginia.

The result of signature generation is a digital signature The signature has a length of 2L bits.

Only the possessor of the private key can properly sign messages. The secrecy of the private key is again protected by a DL problem: if k could be computed from r by computing the discrete logarithm WO 98/36526 PCT/US97/17304 0 log, r in then k' could be computed, which would make it possible to derive the private key from s, m and kI. Consequently, it is important that a particular value for k be kept private, and not be reused.

ElGamal Signature Verification Fig. 1C shows the steps performed by a party receiving a document which has been electronically signed according to the ElGamal scheme to determine whether the signature is valid.

The receiving party is assumed to have a message and a corresponding digital signature which was obtained as in Fig. 1B, and a public key which was obtained as in Fig. 1A and which was used to obtain the signature In practice, the receiving party is a processor of a general purpose digital computer. In some embodiments, the processor may be in a special purpose digital computer, such as a smart card.

15 At step 134, the process determines whether the value r is in the range 1 I r p-1. If not, at step 142, the signature is determined to be invalid.

If r is in the appropriate range, then at step 136, the value v, 20 gi y'r" in F(p) is computed. Next, at step 138, the value v gm"' in F(p) is computed.

At step 140, a test is made to determine whether v, v 2 If not, at step 142, the signature is determined to be invalid. If so, at step 144, the signature is determined to be valid.

Schnorr/DSA System Setup Fig. 2A shows the steps which must be performed for each user during setup of an electronic signature system according to the Schnorr scheme.

Schnorr's scheme is directed to using a small subgroup of the multiplicative group of a prime field of large characteristic, in order to make the signature shorter and the exponentiation faster since short exponents are used. If the subgroup order is prime and sufficiently large, then use of the subgroup does not affect the security of the scheme.

System setup for the DSA scheme is identical to system setup WO 98/36526 PCT/US97/17304 11 0 for Schnorr's scheme, except that the DSA scheme specifies the values for the length of certain parameters (B and as explained below.

At step 202, a prime number q of length B bits is chosen. In the DSA scheme, B is specified to be 160.

At step 204, an integer k is randomly chosen. Preferably, k has a length of 750-864 bits to provide sufficient security against an unauthorized user, but this will increase as processing power increases.

At step 206, a value p kq 1 of length L bits is computed.

In the DSA scheme, L is specified to be 512 i*64 for 0 S i s 8 with i being an integer.

At step 208, a test is made to determine whether p is prime.

If p is not prime, the process returns to step 204 and selects another integer k.

If p is prime, the process proceeds to step 210, and randomly selects an element h of At step 212, a value g in F(p) is obtained.

At step 214, a test is made to determine whether gol in F(p).

20 If the test fails, that is, g 1, the process returns to step 210 and selects another element of as the value h.

If gl 1 in then a generator g for a small subgroup G of the multiplicative group of a prime field F(p) of large characteristic has been located. The generator g is of the order q, since gol in gq=l. Instead of the procedure depicted in steps 202-214, other procedures may be used to locate a generator g.

After locating a generator g, the process proceeds to step 216, and randomly selects a value a in the range 2 5 a 5 q-1. It will be appreciated that this range is smaller than the corresponding range in step 112 of Fig. 1A for the ElGamal scheme.

At step 218, the process finds y g' in Finding the value a, given y and g, is a discrete logarithm (DL) problem, as discussed above.

The result of system setup is a public key and a private key The public key has length 3L B bits. The private key has length WO 98/36526 PCT/US97/17304 12- 0 B bits.

To find the private key from the public key it is necessary to solve either a DL problem in the field F(p) or a DL problem in the subgroup G of generated by g.

The DL problem in the field having cardinality p, where p is a prime number having length L bits and q is a prime factor of p-1 having length of at least B bits, is currently believed to be infeasible.

The DL problem in the subgroup G of the subgroup G having order q, with q having length of at least B bits, is currently believed to be infeasible.

9**9*9 Schnorr Signature Generation Fig. 2B shows the steps performed by a party generating an 1. electronic signature for a particular document according to the Schnorr scheme. In practice, the generating party is a processor of a general purpose digital computer. In some embodiments, the processor may be in a special purpose digital computer, such as a smart card.

At step 224, an integer k is randomly selected in the range 2 k q-1. The corresponding step for the ElGamal scheme, step 122 of Fig.

1B, uses an upper limit of p-2 for the range of k. Since p q, a public key cryptosystem according to Schnorr will have smaller elements than a public key cryptosystem according to ElGamal. For example, when q has a length of 160 bits, p has a length of about 1024 bits.

At step 226, the value r gk in F(p) is obtained, with r in the range 1 r S p-1. Since p q, the computation in step 226 is much faster than the corresponding computation in step 126 of Fig. 1B, that is, a public key cryptosystem according to ElGamal.

At step 228, the value e H(mlr), that is, the hash function applied to the concatenation of the message m and the signature element r, is obtained. The cryptographic hash function is assumed to yield values of length at most B bits.

At step 230, the value s (ae k) mod q is computed, with s in the range 0 5 s q-1.

WO 98/36526 WO 9/36526PCT/US97/I 7304 13 The result of signature generation is a digital signature The signature has a length of 2B bits.

Schnorr Signature Verification Fig. 2C shows the steps performed by a party receiving a document which has been electronically signed according to the Schnorr scheme to determine whether the signature is valid.

The receiving party is assumed to have a message (in) and a corresponding digital signature which was obtained as in Fig. 2B, and a public key which was obtained as in Fig. 2A and which was used to obtain the signature In practice, the receiving party is a processor of a a: general purpose digital computer. In some embodiments, the processor may in a special purpose digital computer, such as a smart card.

At step 236, the value v gay" in F(p) is computed. Next, at step 238, the value e' L1(mlkv) is computed.

At step 240, a test is made to determine whether e If not, at step 242, the signature is determined to be invalid. If so, at step 244, the *9::..signature is determined to be valid.

DSA Signature Generation Fig. 2D shows the steps performed by a party generating an electronic signature for a particular document according to the DSA scheme.

In practice, the generating party is a processor of a general purpose digital computer. In some embodiments, the processor may be in a special purpose digital computer, such as a smart card.

At step 324, an integer k is randomly selected in the range 2 k :5 q-l1.

At step 326, kc' mod q is found, that is, the value which satisfies 1 mod q.

At step 328, the value u g' in the subgroup G, generated by g, of F(p) is obtained, with u in the range 1 :s u p-1.

At step 330, the value r =u mod q is computed.

At step 332, the value s ar) mod q is computed, with s in the range 0 :5 s :5 q-1.

WO 98/36526 PCT/US97/17304 14- 0 At step 333, a test is made to determine whether s 0. If so, then processing returns to step 324 to select a new value for the integer k. If s 0, then the procedure proceeds to step 334 and is complete.

The result of signature generation is a digital signature The signature has a length of 2B bits.

DSA Signature Verification Fig. 2E shows the steps performed by a party receiving a document which has been electronically signed according to the DSA scheme to determine whether the signature is valid.

The receiving party is assumed to have a message and a corresponding digital signature which was obtained as in Fig. 2D, and a public key which was obtained as in Fig. 2A and which was used to generate the signature In practice, the receiving party is a processor of a general purpose digital computer. In some embodiments, the processor may be in a special purpose digital computer, such as a smart card.

At step 238, the process determines whether the value r is an integer in the range 1 I r 5 q-l. If not, at step 352, the signature is .20 determined to be invalid.

If r is in the appropriate range, then at step 340, the process determines whether the value s is an integer in the range 1 s s q-1. If not, at step 352, the signature is determined to be invalid.

If s is in the appropriate range, then at step 342, an integer w which is the inverse of s, that is, ws 1 mod q, is obtained.

At step 344, the value ul wH(m) mod q is computed, and the value u2 wr mod q is computed. At step 346, the value c in the subgroup G, generated by g, of is obtained, with c in the range 1 c p-1.

At step 348, the value v c mod q is computed.

At step 350, a test is made to determine whether v r. If not, at step 352, the signature is determined to be invalid. If so, at step 354, the signature is determined to be valid.

WO 98/36526 W09536526PCTIUS97/1 7304 15 ECDSA System Setup Fig. 3A shows the steps which must be performed during setup of an electronic signature system according to the ECDSA scheme. Steps 402- 416 are performed globally, that is, for all users, and thus need be performed only once. Steps 420-424 are performed by each user.

The ECDSA system is directed to using a large extension of the field of two elements. With an optimal normal basis representation of the extension field (see R.C. Mullin et al., "Optimal normal bases in GF(p)", Discrete Appi. Math., 22, 149-11, 1988/89), multiplication is very fast and squaring is performed by circular shifting, so exponentiation is efficiently 0 performed. However, a hardware implementation is required. Further, fields of characteristic two are thought to be more vulnerable to attacks than other :fields of comparable sizes.

At step 402, an integer t 2: 160 is selected with t e that is, t in the range 160 :5 t :5 250. The ECDSA scheme uses a system-wide finite field of 2' elements, where it is assumed that t B.

S At step 404, the coefficients a, p e for the curve E XY X1 aX 2 0 are selected. The ECDSA scheme assumes ueo an optimal normal basis to represent the elements of over a subfield of Using the curve E means using this optimal normal basis.

0 At step 406, 14 is computed. The value ti is one plus the number of different pairs y) satisfying E, where x, y c F(2T). That is, the group of E has order Another way of saying this is that g is the cardinality of the curve group.

At step 408, the factors of it are obtained.

At step 410, a test is made to determine whether there is a prime factor of g having at least 140 bits. If not, then processing returns to step 404 and picks a new elliptic curve E.

If there is a prime factor of Ah having at least 140 bits, then at step 412, q is set equal to this prime factor. It will be appreciated that q is a prime divisor of the order of the group of the elliptic curve E. Preferably, q has length of at least 140 bits.

-WO 98/36526 PCT/US97/17304 16- 0 At step 414, a point h on the curve E, that is, h(xo, yo), is selected such that e h I, where I is the identity element on the curve E. The symbol indicates scalar multiplication on the curve E. The group of E has order p, and q divides u.

At step 416 the point g on the curve E is chosen as g (g/q) h on the curve E. The point g on the curve E is of order q. The point g on the curve E generates the group G, which is a subgroup of the curve group.

The result of steps 402-416 is a global public key 1, t, q, The length of the global public key is not so important, as it is known to all, and does not vary by individual encrypting or decrypting party.

For each user, at step 420, a value a in the range 2 5 a 5 q-1 is randomly selected.

At step 422, a point P on the curve E, that is, is selected such that P a g on the curve E.

The result of steps 420-422 is a user specific public key of length B+1 bits, and a private key of length at most B bits. Although keys are small in the ECDSA scheme, this scheme is computationally expensive.

20 To find the private key from the user specific public key it is necessary to solve either a DL problem in the group associated with the curve E or a DL problem in the subgroup G of the group associated with the curve E.

The DL problem in a curve group over a field of cardinality 2', where t2160, is currently believed to be infeasible by supporters of elliptic curve cryptosystems.

The DL problem in the subgroup G of order q in the curve group, where q has at least 140 bits, is currently believed to be infeasible by supporters of elliptic curve cryptosystems.

ECDSA Signature Generation Fig. 3B shows the steps performed by a party generating an electronic signature for a particular document according to the ECDSA scheme. In practice, the generating party is a processor of a general purpose digital computer. In some embodiments, the processor may be in a special WO 98/36526 PCT/US97/17304 17purpose digital computer, such as a smart card.

At step 428, an integer k is randomly selected in the range 2 k q-2.

At step 430, kc' mod q is found, that is, the value which satisfies 1 mod q.

At step 432, the point u on the curve E, that is, is found such that u k g on the curve E.

At step 434, an integer r X(x) mod q is obtained, with r in the range 0 r S q-1. The function is a fixed and efficiently computable 0 bijection between the finite field and the set 1, of integers.

This bijection is known to all users of the system.

At step 435, a test is made to determine whether r 0. If so, then processing returns to step 428 to select a new value for the integer k. If r 0, then the procedure proceeds to step 436.

At step 436, the value s ar) mod q is computed, with s in the range 0 s S q-1.

At step 437, a test is made to determine whether s 0. If so, then processing returns to step 428 to select a new value for the integer k. If 20 s 0, then the procedure proceeds to step 438 and is complete.

The result of signature generation is a digital signature The signature has a length of at most 2B bits.

ECDSA Signature Verification Fig. 3C shows the steps performed by a party receiving a document which has been electronically signed according to the ECDSA scheme to determine whether the signature is valid.

The receiving party is assumed to have a message and a corresponding digital signature which was obtained as in Fig. 3B, and a public key comprising p, t, q, g) and which was obtained as in Fig. 3A and which was used to obtain the signature In practice, the receiving party is a processor of a general purpose digital computer. In some embodiments, the processor may be in a special purpose digital computer, such as a smart card.

WO 98/36526 PCT/US97/17304 18- At step 442, the process determines whether the value r is an 0 integer in the range 1 5 r q-1. If not, at step 456, the signature is determined to be invalid.

If r is in the appropriate range, then at step 444, the process determines whether the value s is an integer in the range 1 s q-1. If not, at step 456, the signature is determined to be invalid.

If s is in the appropriate range, then at step 446, an integer w which is the inverse of s, that is, ws 1 mod q, is obtained.

At step 448, the value ul wH(m) mod q is computed, and the value u2 wr mod q is computed. At step 450, the point c on the curve E, that is (ul o g) (u2 o P) on the curve E is obtained. The symbol indicates addition on the curve E.

At step 452, the value v mod q is computed, with v in the range 0 v q-1.

At step 454, a test is made to determine whether v r. If not, at step 456, the signature is determined to be invalid. If so, at step 458, the signature is determined to be valid.

Cyclotomic System Setup 20 Fig. 4A shows the steps which must be performed for each user during setup of an electronic signature system according to the present cyclotomic scheme. A purpose of the processing depicted in Fig. 4A is to find a generator g of a subgroup of the multiplicative group of the finite field such that g'satisfies both a desired Discrete Logarithm Security Level, which determines the choice of L, a desired Subgroup Discrete Logarithm Security Level, which determines the choice of B, and such that there is an optimal normal basis for over F(p).

The cyclotomic scheme uses subgroups, as does the Schnorr scheme, and also uses optimal normal bases, as does the ECDSA scheme. Use of subgroups results in short signatures and short exponents. Use of optimal normal bases results in efficient exponentiation. Consequently, a software implementation of the cyclotomic scheme is substantially faster than a software implementation of the Schnorr scheme.

WO 98/36526 PCTIUS97/17304 19- 0 Let R be a radix size. The value of R is machine dependent, and is chosen so that it is small enough for computation mod p to proceed rapidly, yet large enough so that exponentiation can proceed rapidly. Larger values of p result in smaller values of and since 2 operations per multiplication in the field are required, small values of t' are desirable.

Also, larger values of p expand the choices of cryptosystems that can be constructed. For presently available 32-bit architecture general purpose computers, R 32 is a suitable value. For newer 64-bit architectures, R 64 is a suitable value. In other embodiments, other values of R are suitable, and are not necessarily equal to the word length (in bits) of a computer which practices the present inventive technique.

At step 502, a control integer t' and integers t, s are chosen such that: s is at most R, and s is not much smaller than R, such as, .8R 5 s R. Larger s leads to higher efficiency. The integer s is used to restrict the size of a prime number p (see below). For example, 25 S s S 32; 20 (ii) t' 1, preferably such that t' has a factor t 1 for which t+1 is a prime number and t'/t is small, t'/t 5. Use of Sthe control integer t' allows a wider variety of choices for the number of bits in the prime number p as relates to the desired Discrete Logarithm Security Level which is reflected in L.

More particularly, the value of the prime number p depends on the value of R, which is machine dependent, as explained above. The control integer t' is approximately equal to L divided by the number of bits in the prime number p. As noted, t+1 must be a prime number. Ideally, t' t.

However, a value for t' can be used such that t' +1 is not a prime number, as long as t' has a factor t 1 for which t+ is a prime number and t'/t is small. That is, use of in addition to t, provides more flexibility.

(iii) t' s is close to L; and -20/1- (iv) s is at least B but not too much bigger than B so that the prime factor q (see step 510) is sufficiently easy to find, that is, s B. The function O(t) is Euler's phi function, namely, the number of positive integers s t that are co-prime to t.

At step 504, an odd prime number p is selected such that (log, p) 1 s, and p) 1) 2 L.

At step 506, a test is made to determine whether p is a primitive root mod t+1 of that is, whether p mod t+1 generates Specifically, this test can be performed by computing p' mod t+1 for each integer i, 1 i i s t, and checking that different numbers are obtained. If not, then processing returns to step 504 to select another prime number p.

.i If p is a primitive root mod t+1, then at step 508, the t'-th cyclotomic polynomial evaluated at p, is obtained.

15 The irreducible factorization of X'-1 in Z[X] is given by X'-1 Il(X) where is the d-th cyclotomic polynomial, as explained in H. Riesel, Prime Numbers and Computer Methods for Factorization, Birkhauser, 1985.

The factor is the only irreducible factor of X-1 that does not appear in the factorization of X'-1 for divisors s of t, with s t.

One way of obtaining is by using the factorization identity set forth above. Another way of obtaining is by looking up the coefficients c, 0 i t-1, of c,.

2 X c,X co in a table of cyclotomic polynomial coefficients, such as the table in Fig. 4D.

Using the table in Fig. 4D, it will be seen that, for example, X' X 1, and that X" X 9 1.

At step 510, a large prime factor q of is obtained.

Because the subgroup used in the cyclotomic scheme, and constructed as described below, has order q, and q is a divisor of the corresponding DL computation is difficult for an unauthorized person.

At step 512, a test is made to determine whether (log: q) 1 AMENDED

SHEET

IPEA/EP

WO 98/36526 PCT/US97/17304 -21 2 B. This condition ensures that it is possible to construct a large enough 0 subgroup of the multiplicative group of that cannot be embedded in a true subfield of In other words, to solve the DL problem, that is, to find a value a, given y and g, it is necessary to solve a DL problem in the entire field or in the subgroup generated by g; but, the DL problem cannot be reduced to a DL problem in a true subfield of so that computational difficulty is not reduced for an unauthorized party.

Also, combining the condition of step 502 that s B with the condition of step 512 that (log, q) 1 B, results in a situation in which (log q) 1 B.

It is an important aspect of the present invention that q is a divisor of and that (log, q) 1 2 B. If (log, q) 1 B, processing returns to step 504 to select another prime number p.

If (log, q) 1 2 B, then at step 514, an optimal normal basis i 2, over is obtained, with each a' being a zero of f,(X) 1) X' X 1.

Case 1: If t exists and t' t, then is represented by the integers mod p, and p-th powering in does not require any operations in 20 rather p-th powering is simply a permutation of the basis elements a', and thus is computationally inexpensive. Thus, multiplication and squaring in can be carried out very efficiently.

SCase 2: If t exists and t' t, then the elements of F(p" are represented using a convenient basis over In this case, p-th powering requires only a very small number of operations in and multiplication and squaring in can be carried out efficiently. If t'/t is small, then the difference in efficiency of arithmetic in in cryptosystem implementations is negligible as between case 1 and case 2.

Case 3: If t does not exist, then is represented in any convenient way, preferably using a sparse minimal polynomial to make multiplication and squaring in fast.

At step 516, an element b of is randomly selected.

At step 518, a value WO 98/36526 PCT/US97/17304 -22- (p -1) g =b q in F(p') is obtained. Computing g is very efficient since an optimal normal basis is used to represent At step 520, a test is made to determine whether g 1 in If not, that is, g 1, then processing returns to step 516 to select another element b.

If g 1, then g is the generator of the subgroup G of the 10 multiplicative group of the finite field The subgroup G is of order q. At step 522, a value a is randomly selected in the range 2 a q-2.

At step 524, a value y g' in is computed. Finding a value a, given y and g, is a DL problem.

The result of system setup is a public key and a private key The parameters g and y are represented using optimal normal bases. The public key has length 2L B 64 bits. The private key has length B bits.

20 Security is assured as in the Schnorr scheme, and by the fact that the subgroup G has order q, where q is a prime factor having a length of at least B bits of the t'-th cyclotomic polynomial evaluated at p, so that G cannot be effectively embedded in a proper subfield of As explained above, in the cyclotomic scheme, p-th powering is computationally easy, since it involves only rearranging the elements of the optimal normal basis. This is an important computational advantage of the cyclotomic scheme.

Computation in the cyclotomic scheme involves short rows of length (log 2 p) bits, which is suitable for a software implementation, whereas the ECDSA scheme involves long rows of length 1 bit, which is more suitable to a hardware implementation. That is, the cyclotomic ground field F(p) has elements of length (log, p) bits, whereas the ECDSA ground field F(2) has a elements of length 1 bit.

WO 98/36526 PCT/US97/17304 -23 0 Cyclotomic Signature Generation Fig. 4B shows the steps performed by a party generating an electronic signature for a particular document according to the cyclotomic scheme. In practice, the generating party is a processor of a general purpose digital computer. In some embodiments, the processor may be in a special purpose digital computer, such as a smart card.

At step 530, an integer k is randomly selected in the range 2 k q-2.

At step 532, k' mod q is found, that is, the value which satisfies 1 mod q.

At step 534, a value u g' in is found.

At step 536, an integer r x(u) mod q is obtained, with r in the range 0 s r S q-1. The function is a fixed and efficiently computable bijection between the finite field F(p t and the set 1, p-1} of integers.

This bijection is known to all users of the system. This is a different bijection than used in step 434 of Fig. 3B for the ECDSA scheme.

At step 537, a test is made to determine whether r 0. If so, 20 then processing returns to step 530 to select a new value for the integer k. If r 0, then the procedure proceeds to step 538.

At step 538, the value s ar) mod q is computed, with s in the range 0 5 s 5 q-1.

At step 539, a test is made to determine whether s 0. If so, then processing returns to step 530 to select a new value for the integer k. If s 0, then the procedure proceeds to step 540 and is complete.

The result of signature generation is a digital signature The signature has a length of 2B bits.

Cyclotomic Signature Verification Fig. 4C shows the steps performed by a party receiving a document which has been electronically signed according to the cyclotomic scheme to determine whether the signature is valid.

The receiving party is assumed to have a message and a corresponding digital signature which was obtained as in Fig. 4B, and a -24/1public key g, y, q, which was obtained as in Fig. 4A and which was used to generate the signature In practice, the receiving party is a processor of a general purpose digital computer. In some embodiments, the processor may be in a special purpose digital computer, such as a smart card.

At step 544, the process determines whether the value r is an integer in the range 1 5 r 5 q-1. If not, at step 558, the signature is determined to be invalid.

If r is in the appropriate range, then at step 546, the process determines whether the value s is an integer in the range 1 5 s 5 q-1. If not, at step 558, the signature is determined to be invalid.

10 If s is in the appropriate range, then at step 548, an integer w which is the inverse of s, that is, ws 1 mod q, is obtained.

:At step 550, the value ul wH(m) mod q is computed, and the value u2 wr mod q is computed. At step 552, the value v' gulya in F(p') 1 5 is computed. At step 554, the value v mod q is found.

At step 556, a test is made to determine whether v r. If not, at step 558, the signature is determined to be invalid. If so, at step 560, the signature is determined to be valid.

Cyclotomic Alternatives Applicability of the cyclotomic scheme is not limited to an electronic signature system, as described above. The cyclotomic scheme can be used in any cryptosystem that relies for its security on the difficulty of the DL problem, for example, a Diffie-Hellman key exchange scheme, an ElGamal public key encryption scheme, or digital signature generation and verification schemes as in the ElGamal, Schnorr and DSA schemes.

Proposed encryption schemes include the Data Encryption Standard (DES), as set forth in FIPS 46-2, 1993, available from the National Technical Information Service, Springfield, Virginia; and the ElGamal scheme, as set forth in T. EIGamal, "A public key cryptosystem and a signature scheme based on discrete logarithms", IEEE Trans. Info. Tech., 31, 469-472, 1985. Application of the cyclotomic scheme to these proposed AMENDED SHEET WO 98/36526 PCT/US97/17304 schemes is discussed below.

0 The users practicing the techniques described below are assumed to have a general purpose digital computer programmed to perform these techniques. Alternatively, these techniques may be performed by a special purpose printed circuit board used in conjunction with a general purpose computer, or by a "smart card", that is, a credit card sized portable device including a microprocessor.

Fig. 4E is a flowchart illustrating a DES system setup according to the present invention. In particular, Fig. 4E shows a Diffie-Hellman key exchange scheme modified by application of a cyclotomic scheme.

At step 600, all users are assumed to possess a globally shared public key g, q, obtained according to steps 500-520 of Fig. 4A. In contrast, in the ElGamal, Schnorr, DSA and cyclotomic digital signature schemes discussed above, each user is associated with a public key and a private key; that is, there is no globally shared public key.

When parties A and e wish to communicate, they must initially exchange information to establish a shared key. As illustrated in Fig. 4E, at step 602, party A randomly selects a value a, in the range 2 a, q-2, and 20 at step 604, finds y ga in F(p') At step 606, party A sends y, to party e. At step 608, party A receives y.

from party e. At step 610, party A computes Yo X(Ye in F(p') The function is the fixed and efficiently computable bijection between the finite field and the set 1, of integers which was used in step 536 of Fig. 4B. While it is not absolutely necessary to apply the function it is preferred, so as to convert a element represented in an optimal normal basis of a finite field to an ordinary integer.

Similarly, at step 603, party e randomly selects a value a. in the range 2 5 a, s q-2, and at step 605, finds WO 98/36526 PCTIUS97/17304 -26- Ye g in F(p') At step 607, party 8 sends ye to party A. At step 609, party e receives y, from party A. At step 611, party 9 computes yo X(Y in F(p') At step 612, parties A and 8 have established a shared key (yo).

10 Computations have taken place in the subgroup generated by g. It will be appreciated that an unauthorized party needs to solve a DL problem to decrypt a communication between parties A and 8.

Fig. 4F is a flowchart illustrating encryption for the DES system i setup according to the present invention. Essentially, at step 622, one of parties A and e uses their shared key (yo) to encrypt a message.

Fig. 4G is a flowchart illustrating decryption for the DES system setup according to the present invention. Essentially, at step 632, the other of parties A and a uses their shared key (yo) to decrypt the message that was encrypted in step 622.

For an ElGamal public key encryption scheme, which is different than the ElGamal digital signature scheme discussed above, let it be assumed that steps 500-526 shown in Fig. 4A have been performed, to obtain a public key g, y, q, and a private key for each user. It will be appreciated that an unauthorized user would need to determine the private key to decrypt an encrypted message, which requires solving a DL problem.

Fig. 4H is a flowchart illustrating encryption for the ElGamal system setup according to a cyclotomic scheme. At step 702, a party wishing to encrypt a message randomly selects an integer k in the range 2 k q-2.

At step 704, a value y g' in is obtained. At step 706, a value A y' in is obtained. The function is the inverse of the function which was used in step 536 of Fig. 4B. The outcome, at step 708, is an encrypted message 1).

WO 98/36526 PCT/US97/17304 -27- Fig. 4J is a flowchart illustrating decryption for the ElGamal 0 system setup according to a cyclotomic scheme. At step 722, a party wishing to decrypt the encrypted message X) finds a value C y" in and at step 724, finds a value r 1C in At step 726, a decrypted message m' is obtained as m' x(rY) in 1, All exponentiation takes place in the subgroup generated by g.

Performance Comparison Fig. 5A is a table of results for comparing signature generation performance of schemes for public key cryptosystems. The schemes 10 compared were ElGamal, Schnorr, DSA and cyclotomic using software implementations. The ECDSA scheme was not evaluated as it requires a hardware implementation to be effective.

Since the parameter B is not relevant for the ElGamal scheme, cases and are identical for the ElGamal scheme. As a practical matter, the DSA scheme allows only B 160 and L 512 i*64 for 0 i 58, which corresponds to only cases and In the examples, only integer messages were used, and were taken mod p-1 for ElGamal, and mod q for others. Hashing was not used.

20 Since the computation time for hashing is negligible, omission of hashing does not distort the performance results.

SIn particular, Fig. 5A shows runtimes in seconds on a Pentium 166 MHz processor, using a software implementation for each scheme.

A performance improvement due to the use of subgroups is seen when going from the ElGamal scheme to the Schnorr scheme. The performance of the DSA scheme is almost identical to the performance of the Schnorr scheme.

An additional performance improvement due to the use of optimal normal bases is seen when going from the Schnorr scheme to the cyclotomic scheme. In particular, for the examples in Fig. 5A, the cyclotomic scheme results in performance which is approximately three times faster than the performance of the Schnorr scheme.

Fig. 5B is a table of results for comparing signature verification WO 98/36526 WO98/3526PCT/US97/17304 28 0 performance of schemes for public key cryptosystems for the examples reported in the table of Fig. 5A. As with signature generation, the cyclotomic scheme results in performance which is approximately three times faster than the performance of the Schnorr scheme.

Fig. 6 is a chart showing the message signed and signature verified to obtain the performance results of Figs. 5A and Figs. 7A- I ID are charts showing the public key, private key, signature, and signature generation parameter k, for each of the public key cryptosystems in the examples used to obtain the performance results of Figs.

5A and 5B. The timing results were generated by averaging the outcomes for ten different representative messages. In practice, the signature generation parameter is not exchanged between parties; here, it is included so that results may be conveniently reproduced.

For the cyclotomic scheme, the values of g and y are given in the basis 1 :5i 5 tV, over For all other schemes, the values are in decimal representation.

By comparing the examples, it will be seen that the ElGamal scheme (Figs. 7A, 8A, 9A, 10A, I1IA) uses the longest values, while the cyclotomic scheme (Figs. 7D, 8D), 9D), 1lOD, 1 I D) uses the shortest values.

Further, as data security increases going from case to case in Figs.

and 5B, the length of the values in all schemes increases.

Fig. 12 is a block diagram of an environment in which the cyclotomic scheme may be implemented. General purpose computer includes cryptographic processor 11, communications interface 12, main processor 13, memory 14, communications bus 15 and communication line 16.

Memory 14 may include RAM, ROM, magnetic disk, optical disk or any other memory medium. Communication line 16 may be a wireline, an RF wireless line, an optical line, or any other communication medium. Smart card includes processor 21, memory 22, communications interface 23, communications bus 24 and communication line 25. General purpose ~computer 10 and smart card .20 are coupled to communication channel Central facility 40 is also coupled, via communication line 41, to -29/1communication channel 30. Central facility 40 includes appropriate processing hardware and software to practice the cyclotomic scheme, as will be appreciated with reference to general purpose computer 10 and smart card General purpose computer 10 executes software stored in memory 14 which includes calls that are directed by main processor 13 to cryptographic processor 11, which includes sufficient memory to operate in accordance with the cyclotomic scheme.

Smart card 20 executes software stored in memory 22 in Saccordance with the cyclotomic scheme.

Central facility 40 functions to generate global information and distribute it to all parties using the cyclotomic scheme. An example of global information is the global public key depicted in step 600 of Fig. 4E.

Although an illustrative embodiment of the present invention, and various modifications thereof, have been described in detail herein with reference to the accompanying drawings, it is to be understood that the invention is not limited to this precise embodiment and the described modifications, and that various changes and further modifications may be Seffected therein by one skilled in the art without departing from the scope of 20 the invention as defined in the appended claims.

rC ~L

Claims (23)

1. A method of determining a public key for a discrete logarithm cryptosystem. comprising the steps of: selecting a first prime number (504), obtaining a cyclotomic polynomial evaluated at the first prime number (508), obtaining a second prime number which is a factor of the cyclotomic polynomial evaluated at the first prime number (510), finding a generator of a subgroup of a multiplicative group of a finite 10 field, the order of the subgroup being the second prime number (518), and forming the public key to include the first and second prime numbers and the generator (526). S2. The method of claim 1, further comprising the step of representing the 15 finite field with an optimal normal basis.

3. The method of claim 1, wherein the second prime number q satisfies (log2 q) 1 B, where B is a predetermined number of bits. 20

4. The method of claim 1, further comprising the step of selecting a control integer and wherein the cyclotomic polynomial is the t'-th cyclotomic polynomial, and the public key includes the control integer t'. 448099 1 AMENDED SHEET IPEA/EP -31/1- A method of generating a digital signature for a message using the public key formed according to the method of claim 1, comprising the steps of: obtaining a public value based on the generator and a first integer (524), selecting a second integer (530), obtaining a first signature value based on the second integer and the generator (536), obtaining a second signature value based on the first signature value and the message (538), and 10 forming the digital signature to include the first and second signature values (540).

6. The method of claim 5, further comprising the step of representing the finite field with an optimal normal basis.

7. The method of claim 5, wherein the second prime number q satisfies (log, q) 1 B, where B is a predetermined number of bits.

8. The method of claim 5, further comprising the step of selecting a 20 control integer and wherein the cyclotomic polynomial is the cyclotomic polynomial. 448090 1 S T_ -32/1-

9. The method of claim 5, wherein the first signature value is based on a bijection of the generator raised to the power of the second integer. The method of claim 5, wherein the second signature value is based on combining the first signature value with a cryptographic hashing of the message.

11. A method of verifying a digital signature for a message, the digital signature being formed according to the method of claim 5, comprising the steps of: finding an inverse integer which is the inverse of the second 10 signature value (548), finding a first intermediate value based on the inverse integer and the message (550 ul), finding a second intermediate value based on the inverse integer and the first signature value (550 u2), 15 finding a third intermediate value based on the generator, the public value, and the first and second intermediate values (554), determining that the signature is valid when the third intermediate value is equal to the first signature value (556). 20

12. The method of claim 11, wherein the third intermediate value is a bijection of the generator raised to the power of the first intermediate value multiplied by the public value raised to the power of the second intermediate value.

13. A method of determining a shared key using the public key formed according to the method of claim 1, comprising the steps of: selecting an integer (602), receiving an intermediate value which is based on the generator (608). and forming the shared key as a function of the intermediate value and the integer (610). 448099 I -33/1-

14. The method of claim 13. further comprising the step of representing the finite field with an optimal normal basis. The method of claim 13, further comprising the steps of finding a second intermediate value which is based on the generator and the integer (604), and sending the second intermediate value to a party who is to share the shared key (606).

16. A method for secure communication of a message using the shared key formed according to the method of claim 13, comprising the step of: encrypting the message using the shared key.

17. The method of claim 16, further comprising the step of representing the finite field with an optimal normal basis. AMENDED SHEET IPEA/EP 9 -34/1-

18. A method for secure communication of a message, comprising the steps of: receiving an encrypted message which has been encrypted using the shared key formed according to claim 13, and decrypting the encrypted message using the shared key.

19. A method for secure communication of a message using the public key formed according to the method of claim 1, comprising the steps of: obtaining a public value based on the generator and a first integer (524), selecting a second integer (702), finding a first encrypted value based on the generator and the second integer (704), finding a second encrypted value based on the message, the public value and the second integer (706), and forming an encrypted message from the first and second encrypted values (708). The method of claim 19, further comprising the step of representing the finite field with an optimal normal basis. 448099 I 9* 99 0 9 AMaNZ HE -35/1-

21. A method for secure communication of a message, comprising the steps of: receiving an encrypted message which has been encrypted using the public key according to claim 1, finding a first intermediate value based on the first encrypted value and a private key, the private key being based on the generator (724), and decrypting the encrypted message based on the second encrypted value and the first intermediate value (726). 10

22. An apparatus for determining a public key for a discrete logarithm cryptosystem, comprising: means for selecting a first prime number, ;means for obtaining a cyclotomic polynomial evaluated at the first prime number, and 15 means for obtaining a second prime number which is a factor of the cyclotomic polynomial evaluated at the first prime number, means for finding a generator of a subgroup of a multiplicative group of a finite field, the order of the subgroup being the second prime number, and means for forming the public key to include the first and second 20 prime numbers and the generator.

23. The apparatus of claim 22. further comprising means for representing the finite field with an optimal normal basis.

24. The apparatus of claim 22, wherein the second prime number q satisfies (log2 q) 1 B, where B is a predetermined number of bits. The apparatus of claim 22, further comprising means for selecting a control integer and wherein the cyclotomic polynomial is the t'-th cyclotomic polynomial. and the public key includes the control integer t'. 448099 r 1. U' 0 *0* 0@ 00 0 -36/1-

26. The method of claim 1, further comprising the step of obtaining a public value based on the generator and a selected integer, and wherein the public key also includes the public value.

27. The method of claim 1, further comprising the step of forming a private key including the selected integer (526).

28. The apparatus of claim 22, further comprising means for obtaining a public value based on the generator and a selected integer, and wherein the public key also includes the public value.

29. The apparatus of claim 22, further comprising means for forming a private key including the selected integer. BDATED TmS iDAY OF /fJ A$'v AGNT THE APPCANT 448099 1 0000 0 0000 00 00 00 0 0 0 0 r 00*00 AMENDED SHEET IPEA/EP