Method, system, device for proving the authenticity of an entity and/or the integrity and/or the authenticity of a message using specific prime factors

Abstract

The proof is provided by means of the following parameters: a public module n formed by the product of f prime factors pi, f>2; a public superscript v; m base numbers gi, m>1. The base numbers gi are such that the two equations: x2≡gi mod n and x2≡−gi mod n cannot de solved in x in the ring of integers modulo n, and such that the equation xv≡gi2 mod n can be solved in x in the ring of integers modulo n in the case where the public superscript v is in the form v=2k, wherein k is a security parameter.

Description

BACKGROUND OF THE INVENTION

The present invention relates to the technical field of methods, systems and devices designed to prove the authenticity of an entity and/or the integrity and/or authenticity of a message.

The patent EP 0 311 470 B1, whose inventors are Louis Guillou and Jean-Jacques Quisquater, describes such a method. Hereinafter, reference shall be made to their work by the terms “GQ patent” or “GQ method”. Hereinafter, the expression “GQ2”, or “GQ2 invention” or “GQ2 technology” shall be used to describe the new developments of the GQ technology that are the object of pending applications filed on the same day as the present application by France Telecom, TDF and the firm Mathrizk, and having Louis Guillou and Jean-Jacques Quisquater as their inventors. The characteristic features of these pending applications are recalled whenever necessary in the following description.

According to the GQ method, an entity known as a “trusted authority” assigns an identity to each entity called a “witness” and computes its RSA signature. In a customizing process, the trusted authority gives the witness an identity and signature. Thereafter, the witness declares the following: “Here is my identity; I knew the RSA signature thereof”. The witness, without revealing the fact, proves that he knows the RSA signature of his identity. Through the RSA public identification key distributed by the trusted authority, an entity known as a “controller” ascertains, without obtaining knowledge thereof, that the RSA signature corresponds to the declared identity. The mechanism using the GQ method takes place “without transfer of knowledge”. According to the GQ method, the witness does not know the RSA private key with which the trusted authority signs a large number of identities.

The GQ technology described here above makes use of RSA technology. However, while the RSA technology truly depends on the factorization of the modulus n, this dependence is not an equivalence, indeed far from it, as can be seen in the so-called multiplicative attacks against various standards of digital signatures implementing the RSA technology.

The goal of the GQ2 technology is twofold: firstly to improve the performance characteristics of RSA technology and secondly to avert the problems inherent in RSA technology. Knowledge of the GQ2 private key is equivalent to knowledge of the factorization of the modulus n. Any attack on the triplets GQ2 leads to the factorization of the modulus n: this time there is equivalence. With the GQ2 technology, the work load is reduced for the signing or self-authenticating entity and for the controlling entity. Through a better use of the problem of factorizing in terms of both security and performance, the GQ2 technology averts the drawbacks of RSA technology.

The GQ method implements modulo computations of numbers comprising 512 bits or more. These computations relate to numbers having substantially the same size raised to powers of the order of 2¹⁶+1. Now, existing microelectronic infrastructures, especially in the field of bank cards, make use of monolithic self-programmable microprocessors without arithmetical coprocessors. The work load related to multiple arithmetical applications involved in methods such as the GQ method leads to computation times which, in certain cases, prove to be disadvantageous for consumers using bank cards to pay for their purchases. It may be recalled here that, in seeking to increase the security of payment cards, the banking authorities have raised a problem that is particularly difficult to resolve. Indeed, two apparently contradictory questions have to be resolved: on the one hand, increasing safety by using increasingly lengthy and distinct keys for each card while, on the other hand, preventing the work load from leading to excessive computation times for the user. This problem becomes especially acute inasmuch as it is also necessary to take account of the existing infrastructure and the existing microprocessor components.

SUMMARY OF THE INVENTION

The GQ2 technology provides a solution to this problem while boosting security.

The GQ2 technology implements prime factors having special properties. There are various existing techniques for producing these prime factors. An object of the present invention is a method for the systematic production of such prime factors. It also relates to the application that can be made of these factors especially in the implementation of the GQ2 technology. It must be emphasized right now that these special prime factors and the method used to obtain them can be applied beyond the field of GQ2 technology.

The invention can be applied to a method (GQ2 method) designed to prove the following to a controller entity:

• • the authenticity of an entity and/or
  • the integrity of a message M associated with this entity.

This proof is established by means of all or part of the following parameters or derivatives thereof:

• • a public modulus n constituted by the product of f prime factors p₁, p₂, . . . p_f (f being equal to or greater than 2),
  • a public exponent v;
  • m distinct integer base numbers g₁, g₂, . . . g_m (m being greater than or equal to 1).

The base numbers g_i are such that the two equations (1) and (2):x²≡g_i mod n and x²≡−g_i mod ncannot be resolved in x in a ring of integers modulo n,and such that the equation (3):x^v≡g_i² mod ncan be resolved in x in the ring of integers modulo n.

The method according to the invention is used to produce the f prime factors p₁, p₂, . . . p_f. in such a way that the equations (1), (2) and (3) are satisfied. The method according to the invention comprises the step of choosing firstly:

• • the m base numbers g₁, g₂, . . . g_m,
  • the size of the modulus n,
  • the size of the f prime factors p₁, p₂, . . . p_f.

The method relates to the case where the public exponent v has the form:v=2^k where k is a security parameter greater than 1. The security parameter k is also chosen as a prime number. This special value of the exponent v is one of the essential features of GQ2 technology.

Preferably, the m base numbers g₁, g₂, . . . g_m, are chosen at least partially among the first integers. Preferably again, the security parameter k is a small integer, especially below 100. Advantageously, the size of the modulus n is greater than several hundreds of bits. Advantageously again, the f prime factors p₁, p₂, . . . p_f have a size close to the size of the modulus n divided by the number f of factors.

According to a major characteristic of the method according to the invention, the f prime factors p₁, p₂, . . . p_f are not chosen in any unspecified way. Among the f prime factors p₁, p₂, . . . p_f, a certain number of them: e will be chosen to be congruent to 1 modulo 4. This number e of prime factors may be zero. Should e be zero, the modulus n will hereinafter be called a basic modulus. Should e>0, the modulus n will hereinafter be called a combined modulus. The f−e other prime factors are chosen to be congruent to 3 modulo 4. This number f−e of prime factors is at least equal to 2.

Choice of f−e Prime Factors Congruent to 3 Modulo 4

To produce the f−e prime factors p₁, p₂, . . . p_f−e congruent to 3 modulo 4, the following steps are implemented:

• • the first prime factor p₁ congruent to 3 modulo 4 is chosen and then,
  • the second prime factor p₂ is chosen such that p₂ is complementary to p₁ with respect to the base number g₁.

To choose the factor p_i+1, the following procedure is used in distinguishing two cases:

(1) the case where i>m

Should i>m, the factor p_i+1 congruent to 3 modulo 4 is chosen.

(2) Case where i≦m

Should i≦m, the Profile (Profile_i(g_i)) of g_i with respect to i first prime factors p_i is computed:

if the Profile_i(g_i) is flat, the factor p_i+1 is chosen such that p_i+1 is complementary to p₁ with respect to g_i,

• • else, among the i−1 base numbers g₁, g₂, . . . g_i−1 and all their multiplicative combinations, the number, hereinafter called g is chosen such that Profile_i(g)=Profile_i(g_i), and then p_i+1 is chosen such that Profile_i+1(g_i)≠Profile_i+1(g).

The terms “complementary”, “profile”, “flat profile” have the meanings defined in the description.

To choose the last prime factor p_f−e, the following procedure is used in distinguishing three cases:

(1) Case where f−e−1>m

Should f−e−1>m, p_f−e is chosen congruent to 3 modulo 4.

(2) Case where f−e−1=m

Should f−e−1=m, Profile_f−e−1(g_m) is computed with respect to f−e−1 first prime factors from p₁ to p_f−e−1,

• • if Profile_f−e−1(g_m) is flat, p_f−e−1 is chosen such that it is complementary to p₁ with respect to g_m,
  • else, the procedure stipulated here below is followed:

Among the m−1 base numbers from g₁ to g_m−1 and all their multiplicative combinations, the number hereinafter called g is chosen such that Profile_i(g)=Profile_i(g_i) and then p_f−e is chosen such that Profile_f−e(g)≠Profile_f−e(g_m).

(3) Case where f−e−1<m

If f−e−1<m, then p_f−e is chosen such that the following two conditions are met:

(3.1) First Condition

Profile_f−e−1(g_f−e−1) is computed with respect to the f−e−1 first prime factors from p₁ to p_f−e−1. Two cases are then to be considered. Depending on either of these two cases, the first condition will be different.

If Profile_f−e−1(g_f−e−1) is flat, p_f−e is chosen so that it meets the first condition of being complementary to p₁ with respect to g_f−e−1 (first condition according to the first case). Else, among the f−e−1 base numbers from g₁ to g_m−1 and all their multiplicative combinations, the number, hereinafter called g, is chosen such that Profile_i(g)=Profile_f−e−1(g_f−e−1) and then p_f−e is chosen so that it meets the condition of being such that Profile_f−e(g)≠Profile_f−e (g_m), (first condition according to the second case).

(3.2) Second Condition

Among all the last base numbers from g_f−e to g_m, those numbers whose Profile Profile_f−e−1(g_i) is flat are chosen and then p_f−e is chosen so that it meets the condition of being complementary to p₁ with respect to each of the base numbers thus selected (second condition).

Choice of e Prime Factors Congruent to 1 Modulo 4

To produce the e prime factors congruent to 1 modulo 4, each prime factor candidate p is evaluated, from p_f−e to p_f, in being subjected to the following two successive tests:

(1) First Test

The Legendre symbol is computed for each base number g_i, from g₁ to g_m, with respect to the candidate prime factor p,

• • if the Legendre symbol is equal to −1, the candidate p is rejected,
    • if the Legendre symbol is equal to +1, the evaluation of the candidate p is continued in passing to the following base number and then, when the last base number has been taken into account, there is a passage to the second test.

(2) Second Test

An integer number t is computed such that p−1 is divisible by 2^t, but not by 2^t+1, then an integer s is computed such that s=(p−1+2^t)/2^t+1.

The key (s, p) is applied to each public value G_i to obtain a result rr≡G_i^s mod p

If r is equal to g_i or −g_i, the second test is continued in passing to the following public value G_i+1.

If r is different from g_i or −g_i, a factor u is computed in applying the following algorithm specified for an index ii ranging from 1 to t−2. The algorithm implements two variables: w initialized by r and jj=2ⁱⁱ assuming values ranging from 2 to 2^t−2, as well a number b obtained by application of the key ((p−1)/2^t, p) to a non-quadratic residue of CG(p). The algorithm consists in repeating the following sequence as many times as is necessary:

• • Step 1: w²/G_i (mod p) is computed,
  • Step 2: the result is raised to the power of 2^t−ii−1. Two cases are to be considered.

First Case

If +1 is obtained, there is a passage to the following public value G_i+1 and the second test is performed for this public value.

Second Case

If −1 is obtained, jj=2ⁱⁱ is computed and then w is replaced by w.b^jj (mod p). Then, the algorithm is continued for the following value having an index ii.

At the end of the algorithm, the value in the variable jj is used to compute an integer u by the relation jj=2^t−u and then the expression t−u is computed. Two cases arise:

• • if t−u<k, the candidate p is rejected
  • if t−u>k, the evaluation of the candidate p is continued in passing to the following public value G_i+1 and then in continuing the second test.

The candidate p is accepted as a prime factor congruent to 1 modulo 4 if, at the end of the second test, for all the m public values G_i, it has not been rejected.

Application to the Public and Private Values of GQ2

The present invention also relates to a method (GQ2 method) applying the method that has just been described and making it possible, it may be recalled, to produce f prime factors p₁, p₂, . . . p_f having special properties; The method for the application of the method that has just been described is designed to prove the following to a controller entity,

• • the authenticity of an entity and/or
  • the integrity of a message M associated with this entity,

This proof is established by means of all or part of the following parameters or derivatives of these parameters:

• • m pairs of private values Q₁, Q₂, . . . Q_m and public values G₁, G₂, . . . G_m (m being greater than or equal to 1),

the public modulus n constituted by the product of said prime factors f p₁, p₂, . . . p_f (f being greater than or equal to 2),

• • the public exponent v.

Said modulus, said exponent and said values are linked by relations of the following type:G_i.Q_i^v≡1. mod n or G_i≡Q_i^v mod n.

Said exponent v is such thatV=2^k

where k is a security parameter greater than 1.

Said public value G_i is the square g_i² of the base number g_i smaller than the f prime factors p₁, p₂, . . . p_f. The base number g_i is such that the two equations:x²≡g_i mod n and x²≡−g_i² mod ncannot be resolved in x in the ring of integers modulo n and such that the equation:x^v≡g_i² mod ncan be resolved in x in the ring of the integers modulo n.

Said method implements an entity called a witness in the following steps. Said witness entity has f prime factors p_i and/or parameters of the Chinese remainders of the prime factors and/or of the public modulus n and/or the m private values Q_i and/or f.m components Q_{i, j}(Q_{i, j}≡Q_i mod p_j) of the private values Q_i and of the public exponent v.

The witness computes commitments R in the ring of integers modulo n. Each commitment is computed:

• • either by performing operations of the type:R≡r^v mod nwhere r is a random factor such that 0<r<n,
  • or by performing operations of the type:R_i≡r_i^v mod p_i where r_i is a random value associated with the prime number p_i such that 0<r_i<p_i, each r_i belonging to a collection of random factors {r₁, r₂, . . . r_f}, then by applying the Chinese remainder method.

The witness receives one or more challenges d. Each challenge d comprises m integers d_i hereinafter called elementary challenges. The witness, on the basis of each challenge d, computes a response D,

either by performing operations of the type:D≡r.Q₁^d1.Q₂^d2 . . . . Q_m^dm mod n

or by performing operations of the type:D_i≡r_i.Q_i,1^d1.Q_i,2^d2 . . . . Q_i,m^dm mod p_i and then by applying the Chinese remainder method.

The method is such that there are as many responses D as there are challenges d as there are commitments R, each group of numbers R, d, D forming a triplet referenced {R, d, D}.

Preferably, in order to implement the pairs of private values Q₁, Q₂, . . . Q_m and public values G₁, G₂, . . . G_m as just described, the method uses the prime factors p₁, p₂, . . . p_f and/or the parameters of the Chinese remainders, the base numbers g₁, g₂, . . . g_m and/or the public values G₁, G₂, . . . G_m to compute:

either the private values Q₁, Q₂, . . . Q_m by extracting a k-th square root modulo n of G_i, or by taking the inverse of a k-th square root modulo n of G_i,

• • or the f.m private components Q_{i, j} of the private values Q₁, Q₂, . . . Q_m such that Q_i,j≡Q_i(mod p_j).

More particularly, to compute the f.m private components Q_{i, j} of the private values Q₁, Q₂, . . . Q_m:

• • the key (s, p_j) is applied to compute z such that:z≡G_i^s(mod p_j)
  • and the values t and u are used.

The values t and u are computed as indicated here above when p_j is congruent to 1 modulo 4. The values t and u are taken to be respectively equal to 1 (t=1) and 0 (u=0) where p_j is congruent to 3 modulo 4.

If the value u is zero, we consider all the numbers zz such that:

• • zz is equal to z or such that
  • zz is equal to a product (mod p_j) of z by each of the 2^ii−t 2ⁱⁱ-th primitive roots of unity, ii ranging from 1 to min(k,t).

If u is positive, we can consider all the numbers zz such that zz is equal to the product (mod p_j) of za by each of the 2^k 2^k-th roots of unity, za designating the value of the variable w at the end of the algorithm described here above.

At least one value of the component Q_{i, j} is deduced therefrom. It is equal to zz when the equation G_i≡Q_i^v mod n is used or else it is equal to the inverse of zz modulo p_j of zz when the equation G_i.Q_i^v≡1.mod n is used.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1A shows the function “raise to the square in CG(p)” by an oriented graph in the case where p is congruent to 3 (mod 4).

FIG. 1B shows the function “raise to the square in CG(p)” by an oriented graph in the case where p is congruent to 5 (mod 8).

FIG. 1C shows the function “raise to the square in CG(p)” by an oriented graph in the case where p is congruent to 9 (mod 16).

FIG. 1D shows the function “raise to the square in CG(p)” by an oriented graph in the case where p is congruent to 17 (mod 32).

FIG. 2 shows the solutions to the equation (3.a) with k=6 and p congruent to 5 (mod 8), giving t=2.

FIG. 3 shows G_i=g_i² in a cycle with a prime factor p congruent to 9 (mod 16).

FIG. 4 shows G_i=g_i² on a branch with a prime factor p congruent to 65 (mod 128).

DETAILED DESCRIPTION OF THE INVENTION

The goal of GQ technology may be recalled: it is the dynamic authentication of entities and associated messages as well as the digital signature of messages.

The standard version of GQ technology makes use of RSA technology. However, although the RSA technology truly depends on factorizing, this dependence is not an equivalence, far from it, as can be shown from attacks, known as multiplicative attacks, against various digital signature standards implementing RSA technology.

In the context of GQ2 technology, the present part of the invention relates more specifically to the production of sets of GQ2 keys designed to provide for dynamic authentication and digital signature. The GQ2 technology does not use RSA technology. The goal is a twofold one: firstly to improve performance with respect to RSA technology and secondly to prevent problems inherent in RSA technology. The GQ2 private key is the factorization of the modulus n. Any attack on the GQ2 triplets amounts to the factorizing of the modulus n: this time there is equivalence. With the GQ2 technology, the work load is reduced both for the entity that signs or is authenticated and for the one that controls. Through an improved use of the problem of factorization, in terms of both security and performance, the GQ2 technology rivals the RSA technology.

The GQ2 technology uses one or more small integers greater than 1 for example m small integers (m≧1) called base numbers and referenced g_i. Then, a public verification key (v, n) is chosen as follows. The public verification exponent v is 2^k where k is a small integer greater than 1 (k≧2). The public modulus n is the product of at least two prime factors greater than the base numbers, for example f prime factors (f≧2) referenced by p_j, from p₁ . . . p_f. The f prime factors are chosen so that the public modulus n has the following properties with respect to each of the m base numbers from g₁ to g_m.

• • Firstly, the equations (1) and (2) cannot be resolved in x in the ring of the integers modulo n, that is to say that g_i and −g_i are two non-quadratic residues (mod n).x²≡g_i(mod n)  (1)x²≡−g_i(mod n)  (2)
  • Secondly, the equation (3) can be resolved in x in the ring of the integers modulo n.x^2k≡g_i²(mod n)  (3)Hereinafter, these properties are also called the GQ2 principles.

Since the public verification key (v, n) is fixed according to the base numbers from g₁ to g_m with m≧1, each base number g_i determines a pair of values GQ2 comprising a public value G_i and a private value Q_i: giving m pairs referenced G₁ Q₁ to G_m Q_m. The public value G_i is the square of the base number g_i: giving G_i=g_i². The private value Q_i is one of the solutions to the equation (3) or else the inverse (mod n) of such a solution.

Just as the modulus n is broken down into f prime factors, the ring of the integers modulo n are broken down into f Galois fields, from CG(p₁) to CG(p_f). Here are the projections of the equations (1), (2) and (3) in CG(p_j).x²≡g_i(mod p_j)  (1.a)x²≡−g_i(mod p_j)  (2.a)x^2k ≡g_i²(mod p_j)  (3.a)

Each private value Q_i can be represented uniquely by f private components, one per prime factor: Q_i,j≡Q_i (mod p_j). Each private component Q_i,j is a solution to the equation (3.a) or else the inverse (mod p_j) of such a solution. After all the possible solutions to each equation (3.a) have been computed, the Chinese remainder technique sets up all the possible values for each private value Q_i on the basis of f components of Q_i,1 to Q_i,f:Q_i=Chinese remainders (Q_i,1, Q_i,2, . . . Q_i,f) so as to obtain all the possible solutions to the equation (3).

The following is the Chinese remainder technique: let there be two positive integers that are mutually prime numbers a and b such that 0<a<b, and two components X_a from 0 to a−1 and X_b from 0 to b−1. It is required to determine X=Chinese remainders (X_a, X_b), namely the single number X of 0 to a.b−1 such that X_a≡X (mod a) and X_b≡X (mod b). The following is the Chinese remainder parameter: α≡{b (mod a)}⁻¹ (mod a). The following is the Chinese remainder operation: ε≡X_b (mod a); δ=X_a−ε, if δ is negative, replace δ by δ+a; γ≡α.δ (mod a); X=γ.b+X_b.

When the prime factors are arranged in increasing order, from the smallest p₁ to the greater p_f, the Chinese remainder parameters can be the following (there are f−1, namely at least one of the prime factors). The first Chinese remainder parameter is α≡{p₂ (mod p₁)}⁻¹ (mod p₁). The second Chinese remainder parameter is β≡{p₁.p₂ (mod p₃)}⁻¹ (mod p₃). The i-th Chinese remainder parameter is λ≡{p₁.p₂ . . . p_i—1 (mod p_i)}⁻¹ (mod p_i). And so on and so forth. Finally, in f−1 Chinese remainder operations, a first result (mod p₂ times p₁) is obtained with the first parameter and then a second result (mod p₁.p₂ times p₃) with the second parameter and so on and so forth until a result (mod p₁ . . . .. p_f—1 time p_f), namely (mod n).

The object of the invention is a method for the random production of any set of GQ2 keys among all the sets possible, namely:

• • the random production of any moduli among all the GQ2 moduli possible, namely the moduli ensuring that, for each of the m base numbers g_i, the equations (1) and (2) cannot be resolved in x in the ring of integers modulo n while the equation (3) has one of them,
  • computing all the possible solutions to each of the equations (3.a). The Chinese remainder technique enables the obtaining of a private value Q_i from each set of f components from Q_i,1 to Q_i,f, so as to obtain any solution in x for the equation (3) among all ht possible equations.Q_i=Chinese remainders (Q_i,1, Q_i,2, . . . . Q_i,f)

To grasp the problem, and then understand the solution to be given to the problem, namely the invention, we shall first of all analyze the applicability of the principles of GQ2 technology. Let us start by recalling the notion of rank in a Galois field CG(p) in order to study the functions “raised to the square in CG(p)” and “take a square root of a quadratic residue in CG(p)”. Then, we shall analyze the existence and number of solutions in x in CG(p) to the equations (1.a), (2.a) and (3.a).

Rank of the Elements in CG(p)

Let us take a odd prime number p and a positive prime number a smaller than p. Let us thereafter define {X}.{X}≡{x₁=a; puis, pour i≧1, x_i+1≡a.x_i(mod p)}Let us calculate the term for the index i+p and let us use Fermat's theorem:x_i+p≡a^p.x_i≡a.x_i≡x_i+1(mod p)

Consequently, the period of the sequence {X} is p−1 or a divider of p−1. This period depends on the value of a. By definition, this period is called “the rank of a (mod p)”. It is the index of appearance of unity in the sequence {X}.x_rank(a,p)≡1(mod p)

For example, when (p−1)/2 is an odd prime number p′, the Galois field CG(p) comprises a single element with a rank 1: it is 1, a single element with rank 2. It is −1, p′−1 elements of a rank p′, p′−1 elements of the rank 2.p′, namely of the rank p−1.

The elements of CG(p) whose rank is p−1 are called the primitive elements or again the generators of CG(p). The name is due to the fact that their successive powers in CG(p), namely the terms of the {X} sequence for the indices going from 1 to p−1, form a permutation of all the non-zero elements of CG(p).

According to a primitive element y of CG(p), let us evaluate the rank of the element yⁱ (mod p) as a function of i and p−1. When i is a prime number with p−1, it is p−1. When i divides p−1, it is (p−1)/i. In all cases, it is (p−1)/pgcd(p−1, i).

The Euler function is referenced by φ. By definition, since n is a positive integer, φ(n) is the number of positive integers smaller than n that are prime numbers with n. In the field CG(p), there are therefore φ(p−1) primitive elements.

By way of an illustration, here is the base of the RSA technology. The public modulus n is the product of f prime factors from p₁ to p_f with f≧2, such that for each prime factor p_j, the public exponent v is a prime number with p_j−1. The key (v, p_j) complies with the rank of the elements of CG(p_j): it permutates them. The inverse permutation is obtained with a key (s_j, p_j) such that p_j−1 divides v.s_j−1.

Squares and Square Roots in CG(p)

The elements x and p−x have the same square in CG(p). The key (2, p) do not permutate the elements of CG(p) because p−1 is an even value. For each prime number p, let us define an integer t as follows: p−1 is divisible by 2^t, but not by 2^t+1, namely p is congruent to 2^t+1 (mod 2^t+1). For example t=1 when p is congruent to 3 (mod 4); t=2 when p is congruent to 5 (mod 8); t=3 when p is congruent to 9 (mod 16); t=4 when p is congruent to 17 (mod 32); and so on and so forth. Each odd prime number is seen in one and only one category: p is seen in the t-th category. In practice, if we consider a fairly large number of successive prime numbers, about one in every two is found in the first category, one in four in the second, one in eight in the third, one in sixteen in the fourth, and so on and so forth. In short, one in 2^t on an average is found in the t-th category.

Let us consider the behavior of the function “raise to the square in CG(p)” according to the parity of the rank of the argument.

• • There is only one fixed element: it is 1. The square of any other element of an odd-parity rank is another element having the same rank. Consequently, the key (2, p) permutates all its (p−1)/2^t odd-parity rank elements. The number of permutation cycles depends on the factorization of (p−1)/2^t. For example, when (p−1)/2^t is a prime number p′, there is a big permutation cycle comprising p′−1 elements.
  • The square of any even-parity rank element is another element whose rank is divided by two. Consequently, the even-parity ranking elements are distributed over (p−1)/2^t branches. Each non-zero element with an odd-parity rank bears a branch with a length t comprising 2^t−1 elements, namely: an element of a rank divisible by two but not by four and then, if t≧2, two elements of a rank divisible by four but not by eight, and then if t≧3, four elements of a rank divisible by eight but not by sixteen, and then if t≧4, eight elements of a rank divisible by sixteen but not by 32 and so on and so forth. The 2^t−1 ends of each branch are non-quadratic residues; their rank is divisible by 2^t.

FIGS. 1A to 1D illustrate the function “raise to the square in CG(p)” by an oriented graph where each of the p−1 non-zero elements of the field finds its place: the non-quadratic residues 1 are in white and the quadratic residues 5 are in black; among the quadratic residues 5, the odd-parity ranking elements 3 are in circles.

These figures show respectively;

• • FIG. 1A: the case where p is congruent to 3 (mod 4);
  • FIG. 1B: the case where p is congruent to 5 (mod 8);
  • FIG. 1C: the case where p is congruent to 9 (mod 16);
  • FIG. 1D: the case where p is congruent to 17 (mod 32).

Let us now look at the way to calculate a solution in x to the equation x²≡a (mod p), it being known that a is a quadratic residue of CG(p), namely how “to take a square root in CG(p)”. There are of course several ways of obtaining the same result: the reader can advantageously consult Henri Cohen, “A Course in Computational Algebraic Number Theory”, published, Springer, Berlin, 1993, pp. 31–36 as well as “Graduate Texts in Mathematics”, vol. 138 (GTM 138).

Let us calculate an integer s=(p−1+2^t)/2^t+1 to establish a key (s, p). Let: ((p+1)/4, p) when p is congruent to 3 (mod 4), ((p+3)8, p) when p is congruent to 5 ((mod 8)), ((p+7)/16, p) when p is congruent to 9 (mod 16), ((p+15)/32, p) when p is congruent to 17 (mod 32), and so and so forth.

• • The key (s, p) gives the odd-parity ranking square root of any odd-parity ranking element. Indeed, in CG(p), r²/a is equal to a raised to the power (2.(p−1+2^t)/2^t+1)−1=(p−1)/2^t. Consequently, when a is in a cycle, the key (s, p) converts a into a solution that we shall call w. The other solution is p−w.

In general, the key (s, p) converts any quadratic residue a into a first approximation of a solution which shall be called r. The following are two key points followed by a rough sketch of a method for the step-by-step improvement of the approximation up to a square root of a.

• • Firstly, since a is a quadratic residue, the key (2^t−1, p) certainly converts r²/a into 1.
    • Secondly, it may be assumed that we know a non-quadratic residue of CG(p) that we name y; the key ((p−1)/2^t, p) converts y into an element that shall be called b: this is a root 2^t−1-th of −1. Indeed, y^(p−1)/2≡−1 (mod p). Consequently, in CG(p), the multiplicative group of the 2^t 2^t-th roots of unity is isomorphic to, the multiplicative group of the powers of b for the exponents from 1 to 2^t.
    • To approach a square root of a, let us raise r²/a to the power of 2^t−2 (mod p): the result is +1 or −1. The new approximation remains r if the result is +1 or else it becomes b.r (mod p) if the result is −1. Consequently, the key (2^t−2, p) certainly converts the new approximation into 1. It is possible to continue to approach the required value: at the next step, an adjustment will be made if necessary by multiplying by b² (mod p) and so on and so forth.

The following algorithm makes successive approximations to reach a square root of a from the integers r and b defined here above; it uses two integer variables: w initialized by r to represent the successive approximations and jj assuming values among the powers of 2, from 2 to 2^t−2.

For i ranging from 1 to t−2, repeat the following sequence:

Compute w²/a (mod p), then raise the result to the power 2^t−i−1 (mod p): +1 or −1 should be obtained. When −1 is obtained, compute jj=2ⁱ, then replace w by w.b^jj (mod p). When +1 is obtained, do nothing.

At the end of the computation, w and p−w are two square roots of a in CG(p). Furthermore, we learn that the rank of a in CG(p) is divisible by 2^t/jj but not by 2^t+1/jj. The relevance of this observation will be seen further below.

Analysis of the Principles of GQ2 Technology in CG(p)

Let us take two integers g and k greater than 1 and a prime number p greater than g. Let us analyze the existence and number of solutions in x in CG(p) in the equations (1.a), (2.a) and (3.a).

In the Galois field CG(p), let us distinguish different cases depending on the value of t, namely, according to the power of two which divides p−1. It may be recalled that p−1 is divisible by 2^t, but not by 2^t+1, namely that p is congruent to 2^t+1 (mod 2^t+1). The previous analysis gives us a fairly precise idea of the problem raised as well as a rough solution.

When t=1, p is congruent to 3 (mod 4) as shown in FIG. 1A. The Legendre symbols of g and −g with respect to p are different: any quadratic residue of CG(p) has two square roots in CG(p): one is a quadratic residue and the other is a non-quadratic residue. Firstly, one of the two equations (1.a) or (2.a) has two solutions in x in CG(p) and the other does not have any. Secondly, the equation (3.a) has two solutions in x in CG(p) whatever the value of k.

When t=2, p is congruent to 5 ((mod 8)) as shown in FIG. 1B. Two cases occur, depending on the Legendre symbol of g with respect top. When the symbol is equal to −1, g and −g are both non-quadratic residues of CG(p): the three equations (1.a), (2.a) and (3.a) have no solution in x in CG(p). When the symbol is equal to +1, g and −g are two quadratic residues of CG(p), each equation (1.a) and (2.a) has two solutions in x in CG(p). Furthermore, the rank of g² in CG(p) is an odd-parity value implying that whatever the value of k, the equation (3.a) has four solutions in x in CG(p) of which only one has an odd-parity rank.

FIG. 2 illustrates the solutions to the equation (3.a) with k=6 and p congruent to 5 (mod 8), giving t=2. It may be noted that, because the Legendre symbol of 2 with respect to p congruent to 5 (mod 8) is equal to −1.2^(p−1)/4 (mod p) is then a square root of −1. We therefore have:p≡5(mod 8); consequently (2|p)=−1(mod p); hence b²≡−1(mod p)

$p\equiv 2^{\frac{p-1}{4}}$

When t=3, p is congruent to 9 (mod 16) as shown in FIG. 1C. Let us consider the Legendre symbol of g with respect to p. When the symbol is equal to −1, g and −g are two non-quadratic residues of CG(p): the three equations (1.a), (2.a) and (3.a) have no solution in x in CG(p). When the symbol is equal to +1, g and −g are two quadratic residues of CG(p); each equation (1.a) and (2.a) has two solutions in x in CG(p). The existence of solutions in x to the equation (3.a) depends on the rank of g² in CG(p). This rank is an odd-parity value or is divisible by two but not by four. When the rank of g² in CG(p) is divisible by two but not by four, the equation (3.a) has four solutions in x in CG(p) for k=2; it cannot go above k≧3. When the rank of g² in CG(p) is an odd-parity value, the equation (3.a) has four solutions in x in CG(p) for k=2 and eight for k≧3. In both cases, only one value is an odd-parity value.

When t=4, p is congruent to 17 (mod 32) as shown in FIG. 1D. Let us consider the Legendre symbol of g with respect top. When the symbol is equal to −1, g and −g are two non-quadratic residues of CG(p): the three equations (1.a), (2.a) and (3.a) have no solution in x in CG(p). When the symbol is equal to +1, g and −g are two quadratic residues of CG(p); each equation (1.a) and (2.a) has two solutions in x in CG(p). The existence of solutions in x to the equation (3.a) depends on the rank of g² in CG(p). This rank is an odd-parity value or is divisible by two or four but not by eight. When the rank of g² in CG(p) is divisible by two but not by eight, the equation (3.a) has four solutions in x in CG(p) for k=2; it cannot go above k≧3. When the rank of g² in CG(p) is divisible by two but not by four, the equation (3.a) has four solutions in x in CG(p) for k=2 or eight for k≧3; it has no solutions for k≧4. When the rank of g² in CG(p) is an odd-parity value, the equation (3.a) has four solutions in x in CG(p) for k=2 and eight for k≧3 and sixteen for k≧4. In all three cases, only one value is an odd-parity value.

And so on and so forth so that the case where p is congruent to 1 (mod 4) can be summarized as follows.

When p is congruent to 1 (mod 4), let us consider the Legendre symbol of g with respect to p. When the symbol is equal to −1, g and −g are two non-quadratic residues of CG(p): the three equations (1.a), (2.a) and (3.a) have no solution in x in CG(p). When the symbol is equal to +1, g and −g are two quadratic residues of CG(p); each equation (1.a) and (2.a) has two solutions in x in CG(p). Let us define the integer u: the rank of g² in CG(p) is divisible by 2^u, but not by 2^u+1. The value of u is among the t−1 possible values, from 0 to t−2. The existence and the number of solutions in x in CG(p) to the equation (3.a) depend on the values of k, t and u. When u is positive and k is greater than t−u, the equation (3.a) does not have a solution in x in CG(p). When u is zero and k is greater than t, the equation (3.a) has 2^t solutions in x in CG(p). When k is smaller than or equal to t−u, the equation (3.a) has 2^k solutions in x in CG(p).

Applicability of the GQ2 Principles in the Rings of Integers Modulo

In order that the equation (1) and (2) respectively may have no solution in x in the ring of the integers modulo n, it is necessary and sufficient that, for at least one of the prime factors p, from p₁ to p_f, the equation (1.a) and (2.a) respectively will have no solution in x in CG(p).

In order that the equation (3) may have solutions in x in the ring of the integers modulo n, it is necessary and sufficient that, for each of the prime factors p, from p₁ to p_f, the equation (3.a) should have solutions in x in CG(p).

The equation (3) prohibits any prime factor p congruent with 1 (mod 4) as soon as, for one of the base numbers g, from g₁ to g_m: either the Legendre symbol of g with respect to p is equal to −1; or else the Legendre symbol of g with respect to p is equal to +1 with the condition: u positive and greater than t−k. In order that a prime factor p congruent to 1 (mod 4) may be possible, it is necessary to fulfill one of the following two conditions for each of the base numbers g, from g₁ to g_m, according to the two integers t and u defined here above. Either the rank of G=g² is an odd-parity rank in CG(p), namely u=0, whatever the value of k. Or else the rank of G=g² is an even-parity rank value in CG(p), namely u>0 and it meets the condition: u+k≦t.

A product of prime factors congruent to 1 (mod 4) cannot fulfill all the principles of GQ2 technology. Each GQ2 modulus must have at least two prime factors congruent to 3 (mod 4) such that, for each base number g, the Legendre symbol of g with respect to on of these factors differs from the Legendre symbol of g with respect to the other. When all the prime factors are congruent to 3 (mod 4), it will be said that the GQ2 modulus is basic. When, in addition to at least two prime factors congruent to 3 (mod 4), the modulus includes one or more prime factors congruent to 1 (mod 4), it will be said that the modulus GQ2 is combined.

Systematic Construction of Moduli GQ2

At the outset, it is necessary to fix the total constraints to be dictated on the modulus n: a size expressed in bits (for example, 512 or 1024 bits) as well as a number of most significant successive bits at 1 (at least one of course typically 16 or 32 bits), a numberf of prime factors and a number e (possibly zero) of prime factors having to be congruent to 1 (mod 4); the other prime factors, namely f−e factors, at least two, must be congruent to 3 (mod 4). The modulus n will be the product off prime factors of similar sizes. When e=0, a basic modulus GQ2 is obtained; when e>0, a combined modulus GQ2 is obtained. A basic modulus is the product of prime factors all congruent to 3 (mod 4). A combined modulus GQ2 appears therefore as the product of a basic modulus GQ2 multiplied by one or more other prime factors congruent to 1 (mod 4). First of all prime factors congruent to 3 (mod 4) are produced. Then, if e>0, prime factors congruent to 1 (mod 4) are produced.

For the efficacy of the construction of GQ2 moduli, it is definitely better to select each candidate before seeking to find out if it is a prime value.

Referenced by g₁ g₂ . . . , the base numbers are found typically among the first prime numbers: 2, 3, 5, 7, . . . If there are no indications to the contrary, the m base numbers are the m first prime numbers: g₁=2, g₂=3, g₃=5, g₄=7, . . . However, the following points must be noted: 2 must be avoided if a factor congruent with 5 (mod 8) is anticipated; 3 must be avoided if the public key (3, n) has to be used as the RSA public verification key.

Choice of f−e Prime Factors Congruent with 3 (mod 4)

On the basis of the second factor, the program requests and uses one base number per factor. For the choice of the last factor congruent with 3 (mod 4), the program finds out if there are other base numbers, namely if m is equal to or greater than f−e and then, if this is the case, requests and takes account of the last base numbers, from g_f−e to g_m. To formalize the choice of the prime factors congruent with 3 (mod 4), we have introduced a notion of the profile. The profile characterizes an integer g with respect to a set of prime factors greater than g and congruent with 3 (mod 4).

• • When an integer g has the same Legendre symbol with respect to two prime factors, it is said that the prime factors are equivalent with respect to g. Else, they are complementary with respect to g.

Referenced by Profile_f(g), the profile of an integer g with respect to f prime factors p₁ p₂ . . . p_f is a sequence of f bits, one bit per prime factor. The first bit is equal to 1; each following bit is equal to 1 or 0 depending on whether the next factor is equivalent or complementary to p₁ with respect to g.

• • When all the bits of a profile are equal to 1, it is said that the profile is flat. In such a case, all the Legendre symbols of g are equal to +1 or else to −1. When the profile of g is not flat, the equations (1) and (2) cannot be solved in x in the ring of the integers modulo n.
  • By definition, the profile of g with respect to a single prime number congruent to 3 (mod 4) is always flat. This extension is used to generalize the algorithm of choice of the prime factors congruent to 3 (mod 4).

When the profiles of two base numbers g₁ and g₂ are different, which implies at least three prime factors congruent to 3 (mod 4), the knowledge of the two private values Q₁ and Q₂ induces knowledge of two different decompositions of the modulus n. When the base numbers are small prime numbers, the program ensures that the profiles of 2^f−e−1−1 multiplicative combinations of f−e−1 basic prime numbers are all different: they take all the possible values. The notion of profile does not extend to the prime factors congruent to 1 (mod 4).

First prime factor p₁ congruent to 3 (mod 4): each candidate must be congruent to 3 (mod 4) without any other particular constraint.

Second prime factor p₂ congruent to 3 (mod 4) with the first base number g₁ being taken into account: each candidate must be complementary to p, with respect to g₁.

Third prime factor p₃ congruent to 3 (mod 4) with the second base number g₂ being taken into account: according to the profile of g₂ with respect to two first prime factors p₁ and p₂, two cases occur. When Profile₂(g₂) is flat, each candidate must be complementary to p, with respect to g₂. Else, we have Profile₂(g₁)=Profile₂(g₂); each candidate must then ensure that Profile₃(g₁)≠Profile₃(g₂).

Choice of i-th prime factor p_i+1 congruent to 3 (mod 4) with the base number g_i being taken into account: according to the profile of g_i with respect to i first prime factors p₁, p₂, . . . . p_i, two cases occur. When Profile_i(g_i) is flat, each candidate must be complementary to p₁ with respect to g_i. Else, among the i−1 base numbers g₁, g₂, . . . g_i−1 and all their multiplicative combinations g₁.g₂, . . . , g₁.g₂ . . . . g_i−1, namely 2ⁱ⁻¹−1 integers in all, there is one and only one integer g such that Profile_i(g_i)=Profile_i(g); each candidate must then ensure that Profile_i+1(g_i)≠Profile_i+1(g).

Last prime factor p_f−e congruent to 3 (mod 4) with the base number g_f−e−1 and the other base numbers from g_f−e to g_m being taken into account: the constraints due to the base number g_f−e−1 are taken into account as above. Furthermore, when m is equal to or greater than f−e, each candidate must provide for a non-flat profile for the last base numbers, from g_f−e, to g_m, with respect to the f−e prime factors. Each candidate must be complementary to p₁ with respect to all the values of g_i for which Profile_f−e−1(g_i) is flat.

In short, the prime factors congruent to 3 (mod 4) are chosen as a function of one another.

For i ranging from 0 to f−e−1, to choose the i+1-th prime factor congruent to 3 (mod 4), the candidate p_i+1 must successfully pass the following examination:

• • If i>m or if i=0, then the candidate p_i+1 has no other constraint; it is therefore accepted.
  • If 0<i≦m, then the candidate p_i+1 must take account of the i-th base number g_i. The profile Profile_i(g_i) of the base number g_i with respect to the i first prime factors from p₁ to p_i is computed. Depending on the result, one and only one of the two following cases may occur:
    • If the profile is flat, then the candidate p_i+1 must be complementary to p₁ with respect to g_i; else, it must be rejected.
    • Else, among the i−1 base numbers and all their multiplicative combinations there is one and only one number that we call g such that Profile_i(g)=Profile_i(g_i); then the candidate p_i+1 must be such that Profile_i+1(g)≠Profile_i+1(g_i); else, it must be rejected.
  • If i+1=f−e and i<m, namely to choose the last prime factor congruent to 3 (mod 4) when there remain base numbers, from g_f−e to g_m, which have not yet been taken into account, the candidate p_f−e must take them into account: among these base numbers, those numbers whose profile Profile_f−e−1(g_i) is flat are chosen; the candidate p_f−e must be complementary to p₁ with respect to each of the base numbers thus selected; else they must be rejected.

The candidate is accepted because it has successfully undergone the appropriate tests.

Choice of e Prime Factors Congruent to 1 (mod 4)

To be acceptable, each candidate p congruent to 1 (mod 4) must fulfill the following conditions with respect to each base number from g₁ to g_m.

• • Let us evaluate the Legendre symbol of each base number g_i with respect to p. If the symbol is equal to −1, let us reject the candidate p and go to another candidate. If the symbol is equal to +1, let us continue the evaluation of the candidate. It must be noted that if an integer 2 is used as the base number, then all the candidates congruent to 5 (mod 8) must be removed: the base number 2 is incompatible with a factor congruent to 5 (mod 8).
  • Let us calculate an integer s=(p−1+2^t)/2^t+1 to establish a key (s, p). Let us apply the key (s, p) to each public value G_i to obtain a result r. Two cases occur.
    • If r equals g_i or −g_i, then u=0. In this case, and in this case alone, G_i is in a cycle. A trivial case may be noted: G_i is in a cycle provided that p is congruent to 5 (mod 8) and that the Legendre symbol of g_i with respect to p is equal to +1. It may be recalled that G_i=4 is impossible in this case.
    • If r is equal to neither g_i nor −g_i, then u>0; it must be noted that the key ((p−1)/2^t, p) converts every non-quadratic residue y into an element b which is a primitive 2^t-th root of unity. The following algorithm computes u from r and b by using two integer variables: w initialized by r and jj taking values of 2 to 2^t−2.For i going from 1 to t−2, repeat the following sequence:

Compute w²/G_i (mod p_j) then raise the result to the power 2^t−i−1 (mod p_j): we must obtain +1 or −1. When −1 is obtained, compute jj=2ⁱ, then replace w by w.b^jj (mod p_j). When +1 is obtained, do nothing.

At the end of the computation, the variable w has the value g_i or −g_i. Furthermore, we know that the rank of G_i in CG(p_j) is divisible by 2^t/jj but not by 2^t+1/jj, namely that jj determines the value of u by jj=2^t−u. When v is greater than jj, namely k>t−u, reject the candidate and go to another candidate. When v is smaller than or equal to jj, namely k≦t−u, continue the evaluation of the candidate.

When the f prime factors have been produced, the public modulus n is the product of the f prime factors p₁, p₂, . . . p_f. The unsigned integer n can be represented by a binary sequence; this sequence complies with the constraints imposed at the beginning of a program for the size in bits and for the number of successive most significant bits at 1. The choice of the prime factors provides for the following properties of the modulus n with respect to each of the m base numbers g₁, g₂, . . . g_m. Furthermore, the equations (1) and (2) have no solution in x in the ring of the integers modulo n. Secondly, the equation (3) has solutions in x in the ring of the integers modulo n.

In short, the prime factors congruent to 1 (mod 4) are chosen independently of one another. While the factors congruent to 3 (mod 4) gradually take account of the base numbers, each prime factor congruent to 1 (mod 4) must take account of all the constraints dictated by each of the base numbers. Each prime factor congruent with 1 (mod 4), namely p, from p_f−e to p_f, should have successfully undergone the following examination in two steps.

1) The step (1) is executed successively for each of the m base numbers from g₁ to g_m.

The Legendre symbol of the current base number g with respect to the candidate p is computed. One and only of the following two cases arises: if the symbol is equal to −1, the candidate is rejected. Else (the symbol is equal to +1), the examination is continued in passing to the base number g following the step (1).

When the candidate is acceptable for all the m base numbers, the operation passes to the step (2).

2) The step (2) is executed successively for each of the m public values of G₁ to G_m.

An integer t is computed such that p−1 is divisible by 2^t but not by 2^t+1, then an integer s=(p−1+2^t)/2^t+1, so as to set up a key (s, p). The key (s, p) is applied to the current public value G=g² to obtain a result r, namely: r≡G^s (mod p). Depending on the result, one and only one of the following states arises:

• • a) If r is equal to g or to −g, then u=0; the examination of the candidate is continued in passing to the following public value G at the step (2).
  • b) Else, a positive number u is computed taking one of the values from 1 to t−2, in applying the following algorithm which implements two variables: jj taking values ranging from 2 to 2^t−2 and w initialized by r, as well as an integer b obtained by applying a key ((p−1)/2^t, p) to a non-quadratic residue of CG(p).
    • For an index ii ranging from 1 to t−2, the following operation is repeated:
      • w²/G (mod p) is computed and then a key (2^t−ii−1, p) is applied to the result to obtain +1 or −1 (else, there is proof that the candidate is not a prime factor). If −1 is obtained, then jj=2ⁱⁱ is computed and then c≡b^jj (mod p), and then w is replaced by w.c (mod p), then there is a passage to the next index ii. If +1 is obtained, there is a passage to the next index ii.
    • At the end of the algorithm, the value in the variable jj defines u by the relationship jj=2^t−u; the value in the variable w is a square root of G, namely g or −g (else, there is proof that the candidate is not a prime factor). Two cases occur:
      • If t−u<k then the candidate p is rejected because the branch where G occurs is not long enough.
        • If (t−u≧k), the evaluation of the candidate is continued in going to the next public value G following the step (2).

When the candidate is acceptable for all the m public values, it is accepted as a prime factor congruent with 1 (mod 4).

Computation of the Associated Values

To obtain the private components, let us first calculate all the solutions to the equation (3.a) in the two simplest and most current cases before taking up the general case.

For each prime factor p_j congruent to 3 (mod 4), the key ((p_j+1)4, p_j) gives the quadratic square root of any quadratic residue. From this, a method is deduced for computing a solution to the equation (3.a):s_j≡((p_j+1)/4)^k(mod(p_j−1)/2); then, Q_i,j≡G_i^sj(mod p_j)or else rather the inverse (mod p_j) of such a solution.s_j≡(p_j−1)/2−((p_j+1)/4)^k(mod(p_j−1)/2); then, Q_i,j≡G_i^sj(mod p_j)

In CG(p_j), there are then two and only two square roots of unity: +1 and −1; there are therefore two solutions in x to the equation (3.a): the two numbers Q_i,j and p_j−Q_i,j are the same square G_i(mod p_j).

For each prime factor p_j congruent to 5 (mod 8), the key ((p_j+1)/4, p_j) gives the odd-parity ranking square root of any odd-parity ranking element. From this, a solution to the equation (3.a) is deduced:s_j≡((p_j+3)/8)^k(mod(p_j−1)/4); then, Q_i,j≡G_i^sj(mod p_j)or else rather the inverse (mod p_j) of such a solution.s_j≡(p_j−1)/4−((p_j+3)/8)^k(mod(p_j−1)/4); then Q_i,j≡G_i^sj(mod p_j)In CG(p_j), there are then four and only four fourth roots of unity; there are therefore four solutions in x to the equation (3.a). Let us note that 2^(pj−1)/4 (mod p_j) is a square root of −1 because the Legendre symbol of 2 with respect to p congruent to 5 (mod 8) is equal to −1. If Q_i,j is a solution, then p_j−Q_i,j is another solution, as well as the product (mod p_j) of Q_i,j by a square root of −1.

For a prime factor p_j congruent to 2^t+1 (mod 2^t+1), the key ((p_j−1+2^t)/2^t+1, p_j) gives the odd parity square root of any odd-parity ranking element. It is therefore possible to compute a solution to the equation (3.a).

• • Let us first of all compute an integer s_j≡((p_j−1+2^t)/2^t+1)^k (mod (p_j−1)/2^t) to set up a key (s_j, p_j).
  • When the key ((p_j−1+2^t)/2^t+1, p_j) converts G_i into g_i or into −g_i the rank of G_i is an odd-parity value in CG(p_j) (u=0). Then, the key (s_j, p_j) converts G_i into a number z: this is the odd-parity ranking solution to the equation (3.a). According to the values of t and k, there is still min(2^k−1, 2^t−1) other solutions on one or more branches. The branch of z² carries another solution: this is p_j−z. When t≧2, the branch of z⁴ has two other solutions: it is the product of z by each of the two square roots of −1, namely each of the two primitive fourth roots of unity. Now, if y is a non-quadratic residue of CG(p_j), then y^(pj−1)/4 (mod p_j) is a square root of −1. In general, for i taking each value of 1 to min(k, t), the branch of the 2ⁱ-th power of z bears 2ⁱ⁻¹ solutions: these are the products (mod p_j) of z by each of the 2ⁱ⁻¹ primitive 2ⁱ-th roots of unity. Now if y is a non-quadratic residue of CG(p_j), then y to the power (p_j−1)/2ⁱ is a 2ⁱ-th primitive root of unity that we call c. The 2ⁱ⁻¹ to 2ⁱ-th primitive roots of unity are the odd parity powers of c: c, c³ (mod p_j), c⁵ (mod p_j), . . . c to the power 2ⁱ−1 (mod p_j).
  • When the key ((p_j−1+2^t)/2^t+1, p_j) converts G_i into an integer r that is neither g_i nor −g_i, the rank of G_i is an even-parity value in CG(p_j) (u>0). Then, provided that G_i is appropriately placed in a fairly lengthy branch, namely t≧k+u, there are 2^k solutions on the branch where G_i is located. To compute a 2^k-th root, it is enough to reiterate the above-stated square root computation algorithm k rank times, so as to compute the square root of the successive results up to a solution z. This computation may of course be optimized to directly approach a 2^kth root and then adjust the approximation of a 2^k-th root in a single operation to achieve a solution z. To obtain all the other solutions, it may be noted first of all that if y is a non-quadratic residue of CG(p_j), then y to the power (p_j−1)/2^k is a primitive 2^k-th root of unity which we shall call d. The 2^k 2^k-th roots of unity are successive powers of d: d, d² (mod p_j), d³ (mod p_j), . . . d to the power of 2^k−1 (mod p_j), d to the power 2^k (mod p_j) equal to 1. The 2^k solutions on the branch where G_i is located are the products (mod p_j) of z for each of these roots.

In short, to compute a component for the prime factor p and the base number g, with k, t and u being known, the following procedure is used:

• 1) An integer is computed: s≡((p−1+2^t)/2^t+1)^k (mod (p−1)/2^t) to set up a key (s, p). Then, the key (s, p) is applied to G to obtain z≡G^s (mod p). According to the value of u, there is a passage to the step (2) or (3).
• 2) If u=0, z is the odd-parity solution to the equation (3.a). There are still min(2^k−1, 2^t−1) other even-parity ranking solutions on one or more branches, very precisely on min(k, t) other branches. For i ranging from 1 to min(k, t), the branch of the 2ⁱ-th power of z has 2ⁱ⁻¹ solutions: these are the products (mod p) of z by each of 2ⁱ⁻¹ 2ⁱ-th primitive roots of unity. The generic solution to the equation (3.a) is shown by zz. The operation goes to the step (4).
• 3) If u>0, all the solutions to the equation (3.a) are even-parity solutions. There are 2^k of them and they are all in the branch on which G is, located; indeed: t−u≧k. To compute a solution, the following algorithm implements two variables: jj assuming values ranging from 2 to 2^t−2 and w initialized by z, as well as an integer b obtained by applying a key ((p−1)/2^t, p) to a non-quadratic residue of CG(p).
  • The following sequence is repeated k ranking times.
  • For an index ii ranging from 1 to t−2, the following operation is repeated: w²/G (mod p) is computed and then a key (2^t−ii−1, p) is applied to the result to obtain +1 or −1 (else there is proof that p is not a prime number). If −1 is obtained, then jj=2ⁱⁱ is computed, then c≡b^jj (mod p), then w is replaced by w.c (mod p), then there is a passage to the next index ii. If +1 is obtained, there is a passage to the next index ii.
  • At the end of the algorithm, the variable w has the value za. The 2^k solutions on the branch where G is located are the products (mod p) of za by each of the 2^k-th roots of unity. The generic solution to the equation (3.a) is represented by zz. The operation passes to the step (4).
• 4) With zz being known, a component value is deduced therefrom: it is the inverse of zz modulo p when the equation G.Q^v≡1 (mod n) is used and zz when the equation G≡Q^v (mod n) is used.Note. There are various methods to obtain the private components and the private values. If a collection of f components is known, namely the f components for a given base number, the Chinese remainder technique is used to compute the corresponding private value. It can be seen that for a given public value G and a modulus n, it is possible to have several possible private values Q. There are four of them when n is the product of two prime factors congruent to 3 (mod 4); there are eight of them with three prime factors congruent to 3 (mod 4); there are sixteen of them with two prime factors congruent to 3 (mod 4) and one congruent to 5 (mod 8). A judicious use of these multiple values may complicate the attacks by analysis of the electrical consumption of a chip card using GQ2.

Thus, as and when t increases, the program gets complicated for increasingly rare cases. Indeed, the prime numbers are distributed on an average as follows: t=1 for one in two, t=2 for one in four, t=3 for one in eight and so on and so forth. Furthermore, the constraints due to m base numbers make the candidacies increasingly unacceptable. Whatever the case may be, the combined moduli definitively form part of GQ2 technology; the type of GQ2 modulus in no way affects the dynamic authentication and digital signature protocols.

FIG. 3 illustrates G_i=g_i² in a cycle with a prime factor p congruent to 9 (mod 16), namely t=3, u=0, as well as k≧3. It may be noted that:

$b\equiv y^{\frac{p-1}{8}}$

• • (mod p)
  • b⁸≡1 (mod p)
  • b⁴≡−1 (mod p)

FIG. 4 illustrates G_i=g_i² on a branch with a prime factor p congruent to 65 (mod 128), namely t=6 as well as k=4 and u=2.

Here is a first set of keys GQ2 with k=6, giving v=64, m=3, giving three base: g₁=3, g₂=5 et g₃=7, and f=3, namely a modulus with three prime factors: two congruent to 3 (mod 4) and one to 5 (mod 8). It must be noted that g=2 is incompatible with a prime factor congruent to 5 (mod 8).

p₁=03CD2F4F21E0EAD60266D5CFCEBB6954683493E2E833

(2|p₁)=−1; (3|p₁)=+1; (5|p₁)=−1; (7|p₁)=+1

p₂=0583B097E8D8D777BAB3874F2E76659BB614F985EC1B

(2|p₁)=−1; (3|p₁)=−1; (5|p₁)=+1; (7|p₁)=−1

p₃=0C363CD93D6B3FEC78EE13D7BE9D84354B8FDD6DA1FD

(2|p₁)=−1; (3|p₁)=+1; (5|p₁)=+1; (7|p₁)=+1

n=p₁.p₂.p₃=FFFF81CEA149DCF2F72EB449C5724742FE2A3630D9 02CC00EAFEE1B957F3BDC49BE9CBD4D94467B72AF28CFBB26144CDF4BBDBA3C97578E29CC9BBEE8FB6DDDD

Q_1,1=0279C60D216696CD6F7526E23512DAE090CFF879FDDE

Q_2,1=7C977FC38F8413A284E9CE4EDEF4AEF35BF7793B89

Q_3,1=6FB3B9C05A03D7CADA9A3425571EF5ECC54D7A7B6F

Q_1,2=0388EC6AA1E87613D832E2B80E5AE8C1DF2E74BFF502

Q_2,2=04792CE70284D16E9A158C688A7B3FEAF9C40056469E

Q_3.2=FDC4A8E53E185A4BA793E93BEE5C636DA731BDCA4E

Q_1,3=07BC1AB048A2EAFDAB59BD40CCF2F657AD8A6B573BDE

Q_2,3=0AE8551E116A3AC089566DFDB3AE003CF174FC4E4877

Q_3,3=01682D490041913A4EA5B80D16B685E4A6DD88070501

Q₁=D7E1CAF28192CED6549FF457708D50A7481572DD5F2C335D8C69E22521B510B64454FB7A19AEC8D06985558E764C6991B05FC2A C74D9743435AB4D7CF0FF6557

Q₂=CB1ED6B1DD649B89B9638DC33876C98AC7AF689E9D1359E4DB17563B9B3DC582D5271949F3DBA5A70C108F561A274405A5CB 8 82288273ADE67353A5BC316C093

Q₃=09AA6F4930E51A70CCDFA77442B10770DD1CD77490E3398A AD9DC50249C34312915E55917A1ED4D83AA3D607E3EB5C8B197 697238537FE7A0195C5E8373EB74D

The following are other possible values for the components related to the p₃ which is congruent to 5 (mod 8).

The following is a square root of −1 in CG(p₃): c=2^(p3−1)/4 (mod p₃)=0C3000933A854E4CB309213F12CAD59FA7AD775AAC37

Q′_1,3=c.Q_1,3 (mod p₃)=050616671372B87DEC9AEEAC68A3948E9562F714D76C

Q′_2,3=c.Q_2,3 (mod p₃)=06F308B529C9CE8D037D01002E7C838439DACC9F8AA

Q′_3,3=c.Q_3,3 (mod p₃)=015BE9F4B92F1950A69766069F788E45439497463D58

Giving:

Q′₁=676DF1BA369FF306F4A1001602BCE5A008DB82882E87C148D0D820A711121961C9376CB45C355945C5F2A9E5AFAAD7861886284A 9B319F9E4665211252D74580

Q′₂=CAEC4F41752A228CF9B23B16B3921E47C059B9E0C68634C2C 64D6003156F30EF1BC02ADA25581C8FDE76AA 14AB5CC60A2DE1C 565560B27E8AA0E6F4BCA7FE966

Q′₃=2ACDF5161FE53B68CC7C18B6AFE495815B46599F44C51A6A1A4E858B470E8E5C7D2200EF135239AF0B7230388A6A5BDD8EE15B 0D094FC2BFA890BFDA669D9735

The following is a second set of keys GQ2, with k=9, that is v=512, m=2, that is two base numbers: g₁=2 and g₂=3, and f=3, giving a modulus with three prime factors congruent to 3 (mod 4).

p₁=03852103E40CD4F06FA7BAA9CC8D5BCE96E3984570CB

(2|p₁)=−1; (3|p₁)=−1; and we get: (6|p₁)=+1.

p₂=062AC9EC42AA3E688DC2BC871C8315CB939089B61DD7

(2|p₂)=+1; (3|p₂)=−1; and we get: (6|p₂)=−1.

p₃=0BCADEC219F1DFBB8AB5FE808A0FFCB53458284ED8E3

(2|p₃)=−1; (3|p₃)=+1; and we get: (6|p₃)=−1.

n=p₁.p₂.p₃=FFF5401ECD9E537F167A80C0A9111986F7A8EBA4D 6698AD68FF670DE5D9D77DFF00716DC7539F7CBBCF969E73A0C49 761B276A8E6B6977A21D51669D039F1D7

Q_1,1=0260BC7243C22450D566B5C6EF74AA29F2B927AF68E1

Q_2,1=0326C12FC7991ECDC9BB8D7C1C4501BE1BAE9485300E

Q_1,2=02D0B4CC95A2DD435D0E22BFBB29C59418306F6CD00A

Q_2,2=045ECB881387582E7C556887784D2671CA118E22FCF2

Q_1,3=B0C2B1F808D24F6376E3A534EB555EF54E6AEF5982

Q_2,3=0AB9F81DF462F58A52D937E6D81F48FFA4A87A9935AB

Q₁=27F7B9FC82C19ACAE47F3FE9560C3536A7E90F8C3C51E13C 35F32FD8C6823DF753685DD63555D2146FCDB9B28DA367327DD6EDDA092D0CF108D0AB708405DA46

Q₂=230D0B9595E5AD388F1F447A69918905EBFB05910582E5BA64 9C94B0B2661E49DF3C9B42FEF1F37A7909B1C2DD54113ACF87C6

The present application has described a method for the production of sets of GQ2 keys, namely moduli n and pairs of public and private values G and Q respectively, in which the exponent v is equal to 2^k. These sets of keys are used to implement a method designed to prove the authenticity of an entity and/or the integrity and/or the authenticity of a message as has been described.

In the pending applications filed on the same day by France Télécom, TDF and the firm Math RiZK, and whose inventors are Louis Guillou and Jean-Jacques Quisquater, the characteristic features of the methods, systems and devices designed to prove the authenticity of an entity and/or the integrity and/or the authenticity of a message have been claimed. These two applications are incorporated herein by reference.

Claims

1. A system for producing asymmetric cryptographic keys, said keys comprising m≧1 private values Q1, Q2, . . . , Qm and m respective public values G1, G2, . . . , Gm, the system comprising: a processor; anda memory unit coupled to the processor, the memory unit storing a set of instructions which when executed cause the processor to execute the following acts:selecting a security parameter k, wherein k is an integer greater than 1,determining a modulus n, wherein n is a public integer equal to the product of at least two prime factors p1, . . . , pf, at least two of these prime factors, say p1 and p2, being such that p1≡3 mod 4 and p2≡3 mod 4;selecting m base numbers g1, g2, . . . , gm, wherein each base number gi (for i=1, . . . , m) is an integer greater than 1 and is a non-quadratic residue of the ring of integers modulo n, and such that p2 is complementary to p1 with respect to one of the base numbers;calculating the public values Gi for i=1, . . . , m through Gi≡gi2 mod n; andcalculating the private values Qi for i=1, . . . , m by solving either the equation Gi.Qiv≡1 mod n or the equation Gi≡Qiv mod n, wherein the public exponent v is such that v=2k.

2. The system according to claim 1, wherein a number (f−e) (where e≧0) of prime factors of the modulus n which are congruent to 3 mod 4 is larger than 2, and those prime factors pj+1 for 2≦j≦m which are congruent to 3 mod 4 are determined iteratively as follows: the profile profilej(gj) of gj with respect to the prime factors p1, p2, . . . , pj is computed, andif profilej(gj) is flat, then the prime factor pj+1 is chosen such that pj+1 is complementary to pi with respect to gj; else, a number g is chosen among the (j−1) base numbers g1, g2, . . . , g1−1, and all of their multiplicative combinations, such that profilej(g)=profilej(gj), then pj+1 is chosen such that profilej+1(gj)≠profilej+1(g),wherein the last prime factor pf−e congruent to 3 mod 4 is, in the case that f−e≦m, chosen such that pf−e is complementary to pi with respect to all of the base numbers gi such that f−e≦i≦m and whose profile profilef−e−1(gi) is flat.

3. The system according to claim 1, wherein a number e of prime factors of the modulus n which arc congruent to 1 mod 4 is at least equal to 1, and each such prime factor is determined as follows: a candidate prime number p is chosen, such that the Legendre symbol of each base number gi (for i=1, . . . , m) with respect to p is equal to +1,the integer t is computed which is such that (p−1) is divisible by 2t, but not by 2i+1,the integer s=(p−1+2t)/2t+1 is computed,an integer b≡hp−1/2i mod p, where h is a non-quadratic residue of the body of integers modulo p, is computed,the m integers ri≡gi2s mod p for i=1, . . . , m are computed,an integer u is initialized to u=0,the following sequence of steps, where i is initialized to 1, is iteratively implemented: an integer w is initialized to w=ri,if ri≡±gi, the value of i is incremented and a sequence of steps with the new value of i is proceeded to if i>m, whereas the candidate prime number p is accepted as a factor of the modulus n if i=m,if ri≠±gi: an integer jj is initialized to 1,the following sequence of steps, where an integer ii is initialized to 1, is iteratively implemented;x≡w2/gi2 mod p is computed,y≡x2 mod p is computed, and if y=+1, the sequence is terminated at the current value of ii,if y=−1, jj is assigned the value jj=2ii, the number w is assigned a new value equal to the old value multiplied by bjj modulo p, and for ii<t−2, the value of ii is incremented and a new iteration is proceeded to with the new value of ii, for ii=t−2, the value of number u is updated through the relation jj=2t−ii, andif t−u<k, the candidate prime number p is rejected as a factor of the modulus n,if t−u>k, the value of i is incremented and a sequence of steps with the new value of i is proceeded to if i<n, whereas the candidate prime number p is accepted as a factor of the modulus n if i=m.

4. The system according to claim 3, wherein, to compute f.m private components Qi,j of the private values Q1, Q2, . . . , Qm, the following steps are implemented for each couple (i, j): an integer t is determined, which is equal to 1 if pj is congruent to 3 mod 4, and to the value obtained for t according to claim 3 if pj is congruent to 1 mod 4,an integer u is determined, which is equal to 0 if pj is congruent to 3 mod 4, and to the value obtained for u according to claim 3 if pj is congruent to 1 mod 4,the integer z≡Gis mod pj is computed, where s=(p−1+2t)/2t+1,all the numbers zz are being considered, which: if u=0, are such that zz=z or such that z is equal to the product modulo pj of z by each of the 2ii−1 2ii-th primitive roots of unity, for ii ranging from 1 to min(k,t),if u>0, are such that zz is equal to the product modulo pj of za by each of the 2k 2k-th roots of unity, where za is the value obtained for w according to claim 3, andfor each such number zz, a value for the component Qi,j is obtained by taking Qi,j equal to zz if the equation Gi≡Qiv mod n is used, or to the inverse of zz modulo pj if Gi.Qiv≡1 mod n is used for this value of i.

5. A computer-readable storage medium storing instructions for producing asymmetric cryptographic keys, said keys comprising m≧1 private values Q1, Q2, . . . , Qm and m respective public values G1, G2, . . . , Gm, the medium storing instructions which when executed cause a processor to execute the following acts: selecting a security parameter k, wherein k is an integer greater than 1;determining a modulus n, wherein n is a public integer equal to the product of at least two prime factors p1, . . . , pf, at least two of these prime factors, say p1 and p2, being such that p1≡3 mod 4 and p2≡3 mod 4;selecting m base number g1, g2, . . . , gm, wherein each base number gi (for i=1, . . . , m) is an integer greater than 1 and is a non-quadratic residue of the ring of integers modulo n, and such that p2 is complementary to p1 with respect to one of the base numbers;calculating the public values Gi for i=1, . . . , m through Gi≡gi2 mod n; andcalculating the private values Qi for i=1, . . . , m by solving either the equation Gi.Qiv≡1 mod n or the equation Gi≡Qiv mod n, wherein the public exponent v is such that v=2k.

6. The computer-readable storage medium storing instructions according to claim 5, wherein a number (f−e) (where e≧0) of prime factors of the modulus n which are congruent to 3 mod 4 is larger than 2, and those prime factors pj+1 for 2≦j≦m which are congruent to 3 mod 4 are determined iteratively as follows: the profile profilej(gj) of gj with respect to the prime factors p1, p2, . . . , pj is computed, andif profilej(gj) is flat, then the prime factor pj+1 is chosen such that pj+1 is complementary to p1 with respect to gj; else, a number g is chosen among the (j−1) base numbers g1, g2, . . . , gi−1 and all of their multiplicative combinations, such that profilej(g)=profilej(gj), then pj+1, is chosen such that profilej+1(gj)≠profilej+1(g),wherein the last prime factor pf−e congruent to 3 mod 4 is, in the case that f−e≦m, chosen such that pf−e is complementary to pi with respect to all of the base numbers gi such that f−e≦i≦m and whose profile profilef−e−1(gi) is flat.

7. The computer-readable storage medium storing instructions according to claim 5, wherein a number e of prime factors of the modulus n which are congruent to 1 mod 4 is at least equal to 1, and each such prime factor is determined as follows: a candidate prime number p is chosen, such that the Legendre symbol of each base number gi (for i=1 . . . , m) with respect to p is equal to +1,the integer t is computed which is such that (p−1) is divisible by 2t, but not by 2t+1,the integer s=(p−1+2t)/2t+1 is computed,an integer b≡hp−1/2′ mod p, where h is a non-quadratic residue of the body of integers modulo p, is computed,the m integers ri≡gt2s mod p for i=1 . . . , m are computed,an integer u is initialized to u=0,the following sequence of steps, where i is initialized to 1, is iteratively implemented: an integer w is initialized to w=ri,if ri=±gi, the value of i is incremented and a sequence of steps with the new value of i is proceeded to if i<m, whereas the candidate prime number p is accepted as a factor of the modulus n if i=m,if ri≠±gi: an integer jj is initialized to 1,the following sequence of steps, where an integer ii is initialized to 1, is iteratively implemented:x≡w2/gi2 mod p is computed,y≡x2 mod p is computed, and if y=+1, the sequence is terminated at the current value of ii,if y=−1, jj is assigned the value jj=2tt, the number w is assigned a new value equal to the old value multiplied by bjj modulo p, and for ii<−2, the value of ii is incremented and a new iteration is proceeded to with the new value of ii, for ii=t−2, the value of number u is updated through the relation jj=2t−ii, andif t−u<k, the candidate prime number p is rejected as a factor of the modulus n,if t−u>k, the value of i is incremented and a sequence of steps with the new value of i is proceeded to if i<m, whereas the candidate prime number p is accepted as a factor of the modulus n if i=m.

8. The computer-readable storage medium storing instructions according to claim 7, wherein, to compute f.m private components Qi,j of the private values Q1, Q2, . . . , Qm, the following steps are implemented for each couple (i, j): an integer t is determined, which is equal to 1 if pj is congruent to 3 mod 4, and to the value obtained for t according to claim 7 if pj is congruent to 1 mod 4,an integer u is determined, which is equal to 0 if pj is congruent to 3 mod 4, and to the value obtained for u according to claim 7 if pj is congruent to 1 mod 4,the integer z≡Gij mod pj is computed, where s=(p−1+2t)/2t+1,all the numbers zz are being considered, which: if u=0, are such that zz=z or such that zz is equal to the product modulo pj of z by each of the 2ii−1 2ii-th primitive roots of unity, for ii ranging from 1 to min(k,t),if u>0, arc such that zz is equal to the product modulo pj of za by each of the 2k 2k-th roots of unity, where za is the value obtained for w according to claim 7, andfor each such number zz, a value for the component Qi,j is obtained by taking Qi,j equal to zz if the equation Gi≡Qiv mod n is used, or to the inverse of zz modulo pj if Gi.Qiv≡1 mod n is used for this value of i.

9. A computer-implemented process for producing asymmetric cryptographic keys, said keys comprising m≧1 private values Q1, Q2, . . . , Qm, and m respective public values G1, G2, . . . ,Gm, the computer-implemented process comprising: selecting a security parameter k, wherein k is an integer greater than 1;determining a modulus n, wherein n is a public integer equal to the product of at least two prime factors p1, . . . , pf, at least two of these prime factors, say p1 and p2, being such that p1≡3 mod 4 and p2≡3 mod 4;selecting m base numbers g1, g2, . . . , gm, wherein each base number gi (for i=1, . . . , m) is an integer greater than 1 and is a non-quadratic residue of the ring of integers modulo n, and such that p2 is complementary to p1 with respect to one of the base numbers;calculating the public values Gi for i=1, . . . , m through Gi≡gi2 mod n; andcalculating the private values Qi for i=1, . . . , m by solving either the equation Gi.Qiv≡1 mod n or the equation Gi≡Qiv mod n, wherein the public exponent v is such that v=2k.

10. The computer-implemented process according to claim 9, wherein a number (f−e) (where e≧0) of prime factors of the modulus n which are congruent to 3 mod 4 is larger than 2, and those prime factors pj+1, for 2≦j≦m which are congruent to 3 mod 4 arc determined iteratively as follows. the profile profilej (gj) of gj with respect to the prime factors p1, p2, . . . , pj is computed, andif profilej(gj) is flat, then the prime factor pj+1 is chosen such that pj+1 is complementary to pi with respect to gj; else, a number g is chosen among the (j−1) base numbers g1, g2, . . . , gj−1 and all of their multiplicative combinations, such that profilej(g)=profilej(gj), then pj+1 is chosen such that profilej+1(gj)≠profilej+1(g),wherein the last prime factor pf−e congruent to 3 mod 4 is, in the case that f−e≦m, chosen such that pf−e is complementary to p1 with respect to all of the base numbers gj such that f−e≦i≦m and whose profile profilef−e−1(gi) is flat.

11. The computer-implemented process according to claim 9, wherein a number e of prime factors of the modulus n which are congruent to 1 mod 4 is at least equal to 1, and each such prime factor is determined as follows: a candidate prime number p is chosen, such that the Legendre symbol of each base number gi (for i=1, . . . , m) with respect to p is equal to +1,the integer t is computed which is such that (p−1) is divisible by 2t, but not by 2t+1,the integer s=(p−1+2t)/2t+1 is computed,an integer b≡hp−1/2t mod p, where h is a non-quadratic residue of the body of integers modulo p, is computed,the m integers ri≡gi2s mod p for i=1, . . . , m arc computed,an integer u is initialized to u=0,the following sequence of steps, where i is initialized to 1, is iteratively implemented: an integer w is initialized to w=ri,if ri=±gi, the value of i is incremented and a sequence of steps with the new value of i is proceeded to if i<m, whereas the candidate prime number p is accepted as a factor of the modulus n if i=m,if ri≠±gi: an integer jj is initialized to 1,the following sequence of steps, where an integer ii is initialized to 1, is iteratively implementedx≡w2/gi2 mod p is computed,y≡x2 mod p is computed, and if y=+1, the sequence is terminated at the current value of ii,if y=−1, jj is assigned the value jj=2ii, the number w is assigned a new value equal to the old value multiplied by bjj modulo p, and for ii<t−2, the value of ii is incremented and a new iteration is proceeded to with the new value of ii, for ii=t−2, the value of number u is updated through the relation jj=2t−tt, andif t−u<k, the candidate prime number p is rejected as a factor of the modulus n,if t−u>k, the value of i is incremented and a sequence of steps with the new value of i is proceeded to if i<m, whereas the candidate prime number p is accepted as a factor of the modulus n if i=m.

12. The computer-implemented process according to claim 11, wherein, to compute f.m private components Qi,j of the private values Q1, Q2, . . . , Qm, the following steps are implemented for each couple (i, j): an integer t is determined, which is equal to 1 if pj is congruent to 3 mod 4, and to the value obtained for t according to claim 11 if pj is congruent to 1 mod 4,an integer u is determined, which is equal to 0 if pj is congruent to 3 mod 4, and to the value obtained for u according to claim 11 if pj is congruent to 1 mod 4,the integer z≡Gis mod pj is computed, where s=(p−1+2t)/2i+1,all the numbers zz are being considered, which: if u=0, are such that zz=z or such that zz is equal to the product modulo pj ofz by each of the 2ii−t 2ii-th primitive roots of unity, for ii ranging from 1 to min(k,t),if u>0, are such that zz is equal to the product modulo pj of za by each of the2k 2k-th roots of unity, where za is the value obtained for w according to claim 11, andfor each such number zz, a value for the component Qi,j is obtained by taking Qi,j equal to zz if the equation Gi≡Qiv mod n is used, or to the inverse of zz modulo pj if Gi.Qtv=1 mod n is used for this value of i.