Encryption process employing chaotic maps and digital signature process

[0001] The present invention relates to data encryption, and in particular, to an encryption process that uses chaotic maps and to related digital signature processes.

[0002] The pioneering work on chaos synchronization [1] led to several applications in communications, in which chaotic systems with continuous-value signals were used to transmit information. Several schemes have been developed which allow transformation of the information signal into a chaotic waveform on the transmitter side and to extract the information signal from the transmitted waveform on the receiver side. The most important among them are chaotic masking, chaos shift keying and chaotic modulation. In early days (from 1992 to 1996) the main research goal was to develop schemes in which a single chaotic system is used for both modulation and encryption. This approach eventually evolved into two distinct research areas: chaos-based modulation [2, 3] and chaos-based cryptography [4, 5].

[0003] In chaotic modulation, the digital information is mapped to inherently wide-band chaotic signals. Thus, chaotic modulation offers a novel approach to spread-spectrum communication. Two most promising approaches in chaos-based modulation have recently emerged. In the first approach, the unmodulated chaotic waveform is transmitted along with the modulated signal (transmitted reference scheme) either using a separate channel or using time division. One instance of this approach, called frequency-modulated differential chaos shift keying, was studied in-depth by Kolumban and Kennedy [2, 6]. In another approach, a chaotic reference is regenerated at the receiver with the help of synchronization. In the references [3, 7] an example of such an approach is proposed, in which chaotic time pulsed sequences are used instead of continuous time waveform. Since the information about the state of the chaotic signal is contained entirely in the timing between pulses, the distortions that affect the pulse shape will not significantly influence the ability of the chaotic pulse generators to synchronize. This approach is known as chaotic pulse position modulation.

[0004] Cryptography is generally acknowledged as the best method of data protection against passive and active fraud [8]. An overview of recent developments in the design of conventional encryption algorithms is given in [8]. Three most common encryption objects are block-encryption algorithms (private-key algorithms), pseudo-random number generators (additive stream ciphers) and public-key algorithms.

[0005] Block ciphers transform a relatively short string (typically 64, 128 or 256 bits) to a string of the same length under control of a secret key. Several block encryption ciphers based on chaotic maps have been proposed, in which a discretization (process that describes the way a chaotic map is implemented in the computer) is not realized by rounding the chaotic map according to the computer arithmetic, but rather is constructed explicitly. Pichler and Scharinger [9] proposed cryptographic systems based on chaotic permutations constructed by explicitly discretizing the two dimensional bakers map. Ridrich [10] extended their ideas to chaotic permutations on any size of two dimensional lattices. The permutations benefit from the expanding property along one axis, technically avoiding the contracting property along the other axis. Masuda and Aihara [11] considered a discrete version of the skew-tent map, which exploits important chaotic properties such as the sensitive dependence on initial conditions and the exponential information decay. They discussed the difference between the discretized map and the original map, explaining the ergodic-like and chaotic-like properties of the discretized map.

[0006] A pseudo-random number generator is a deterministic method, usually described with a mapping, to produce from a small set of random numbers, called the seed, and a larger set of random-looking numbers called pseudo-random numbers. Chaotic systems may be used to generate pseudo-random numbers. For example, in a series of papers [12], the authors proposed a chaos derived pseudo-random number generator. They numerically observed that the average cycle and transient lengths grow exponentially with the precision of implementation, and from this fact deduced that using high-precision arithmetic one can obtain PRNGs which are still of cryptographic interest. Statistical properties of binary sequences generated by class of ergodic maps with some symmetrical properties are discussed in [13]. The authors derived a sufficient condition for this class of maps to produce a sequence if independent and identically distributed binary random variables. However, the authors did not discuss the implementation of these maps on finite-state machines and the consequence this implementation may have on the randomness of the generated sequences.

[0007] Certain applications in cryptography require the use of a truly random number generator (RNG), which is a device that outputs a sequence of statistically independent and unbiased numbers. It is widely accepted that the core of any RNG must be an intrinsically random physical process. Thus, it is no surprise that the proposals and implementations of RNGs range from tossing a coin, to measuring thermal noise from a resistor and shot noise from a Zener diode or a vacuum tube, measuring radioactive decay from a radioactive source, and sampling a stable high-frequency oscillator with an unstable low-frequency clock, to mention only a few proposals. For chaos-based generators of truly random numbers see for example references [14, 15].

[0008] Several applications [16, 17, 18] of chaos in cryptography have been proposed. The work presented in [16, 17, 18] ushers in the era of scientific chaos-based cryptography, which will trigger more research and real-world applications of chaos-based data protection.

[0009] In the work [16] a method for generating truly random numbers has been designed and implemented in CMOS technology. Random numbers are crucial importance in every encryption and data protection application. Block encryption algorithm based on chaotic maps are proposed in the patent application “Chaos-based data protection using time-discrete dynamical systems” by L. Kocarev, G. Jakimoski, G. G. Rizzotto, and P. Amato [17]. Lower bounds of number of active S-boxes as well as the upper bounds for differential and linear probabilities in the proposed algorithm have been derived analytically, and therefore, the resistance of this algorithm to differential and linear attacks has been proved. L. Kocarev, P. Amato, and G. G. Rizzotto in [18] have presented a class of pseudo-random-bit generators, for which security does not rely on a number-theoretical problem, and therefore, does not use modular multiplications. In contrast, its security relies on the large numbers of branches for an inverse of a function used in the algorithm. The generators use only binary operations and have been efficiently implemented in software.

[0010] The article Communication Theory of Secrecy Systems [22] by C. E. Shannon, published in 1949, ushered in the era of scientific secret-key cryptography. However, Shannon's article did not lead to an explosion of researches on cryptography comparable to that triggered by his earlier articles published in 1948 in information theory [23]. The real explosion of works on cryptography came with an article by W. Diffie and M. E. Hellman [24]. Diffie and Hellman showed for the first time that a secret communication was possible without any transfer of a secret key between sender and recipient, thus starting the era of public-key cryptography. Moreover, they suggested that computational complexity theory could be used for future research in cryptography.

[0011] Substantially, an encryption process is a process for transmitting data in a mode that ensures that the data remain private, by converting a message, referred to as a plain-text, into an encrypted format, referred to as a cipher-text. A sender encrypts the message by using an encryption key, while the recipient of the message recovers the plain-text from the received cipher-text by using a decryption key.

[0012] Public-key encryption algorithms, also called asymmetric algorithms, are designed so that (i) the encryption key is different from the decryption key; (ii) the encryption key can be made public; and (iii) the decryption key cannot, at least in a reasonable amount of time, be calculated from the encryption key.

[0013] There are many public-key algorithms. Only a few of them are both secure and practical, and only three of them work well for both encryption and digital signature: RSA, ElGamal, and Rabin [8]. In a public-key encryption system [8] each entity A has a public key “e” and a corresponding private key “d”. In secure systems, the task of calculating the private key “d” from the public key “e” is practically impossible.

[0014] The public key defines an encryption transformation E_e, while the private key defines the associated decryption transformation D_d. A sender B wishing to send a message M to a recipient A must obtain an authentic copy of the recipient public key “e”, use the encryption transformation to obtain the cypher-text c=E_e(M), and transmit the encrypted message “c” to the recipient A. The recipient A decrypts the cipher-text “c” using the decryption transformation and obtains the plain-text M=D_d(C).

[0015] Since 1976, numerous public-key algorithms have been proposed. Three most widely used public-key encryption processes are RSA, Rabin and ElGamal. The security of the RSA process is based on the intractability of the integer factorization problem. In the Rabin public-key encryption process, the problem faced by a passive adversary is computationally equivalent to factorizing a number. The security of the ElGamal public-key encryption process is based on the intractability of the discrete-logarithm problem.

[0016] Recall first the basic ElGamal algorithm. The ElGamal public-key algorithm can be viewed as Diffie-Hellman key agreement in key transfer-mode [8]. Consider a class of functions defined as π_p(x)=x^p(mod N) wherein N is a prime number, and x is a generator of the multiplicative group Z*_N, and 1≦p≦N−2.

[0017] Any two functions π_p and π_q, commute under composition:
π_p(π_q(x)=π_pq(x)

[0018] The Diffie-Hellman key agreement protocol describes how Alice and Bob agree on their common secret key. Alice generates a number p, computes y=π_p(x) and sends (x, y) to Bob. Bob creates a number q, computes z=π_q(x) and sends z to Alice. The secret key, which can be shared by both Alice and Bob, is computed as follows. Alice computes the secret key k as k=π_p(z), Bob computes the secret key k as k=π_q(y).

[0019] In the ElGamal public-key scheme, Alice generates a large random prime N and a generator x of the multiplicative group Z*_N of integers modulo N. She also generates a random integer s≦N−2 and computes A=x^s(mod N). Alice's public key is (x, N, A); Alice's private key is s. To encrypt a message m, Bob selects a random integer r≦N−2, computes B=x^r(mod N) and X=mA^r(mod N), and sends the cipher-text c=(B,X) to Alice. To recover the message m from c, Alice uses the private key s to recover m by computing m=B^−sX(mod N). The decryption allows recovery of the original message because B^−smA^r≡x^−rsmx^rs≡m(mod N).

[0020] Recall now the RSA algorithm. Let N=pq and φ=(p−1)·(q−1), where p and q are two large random (and distinct) primes p and q. Alice selects a random integer e, 1<e<φ, such that the greatest common divisor of “e” and “φ” is 1:
gcd(e,φ)=1
and computes the unique integer d, 1<d<φ, such that ed≡1(mod φ).

[0021] Alice's public key is (N, c); Alice's private key is d. To encrypt a message m, Bob computes c=m^e(mod N) and sends it to Alice. To recover the message m from c, Alice should use the private key d to recover m=c^e(mod N).

[0022] Let π_p(x)=x^p(mod N). The decryption in the RSA algorithm works for two reasons: the functions π_e and π_d commute under composition, and p is a periodic point of the function π_ed for every m such that m^ed≡m(mod N). The last follows from the following observation. Since ed≡1(mod φ), there is an integer k such that ed=1+kφ. If gcd(m,p)=1 then by Fermat's theorem m^p−1≡1(mod p). Raising both sides of this congruence to the power of k·(q−1) and then multiplying both sides by m yields m^ed≡m(mod p).

[0023] By the same argument it is possible to demonstrate that m^ed≡m(mod q). Finally, since p and q are distinct primes, thus m^ed≡m(mod N). Public-key encryption processes are much slower than symmetric-key encryption algorithms. For this reason, public-key encryption is most commonly used for encrypting short data and/or for transporting encryption keys, subsequently used for data encryption by symmetric-key algorithms.

[0024] In view of the foregoing background, an object of the present invention is to provide an encryption process, with or without public-keys, that may be implemented on common computers and when compared to the known public-key encryption processes can be executed in a reduced amount of time.

[0025] The encryption process of the present invention (i) is both secure and practical, (ii) is very fast, and (iii) may be used for realizing a digital signature process.

[0026] The invention essentially comprises encrypting/decrypting data by using permutable chaotic maps f(p(f⁻¹(.))), which are composite functions of a first function f(.), the inverse function of the first function f¹(.) and a second function p(.). These chaotic maps may be used for realizing encryption processes that may use a public key.

[0027] In general, each entity generates its own secret key of the process by using the second function p(.), while the first function is pre-established. For example, the function f(.) and the inverse thereof f⁻¹(.) may be implemented in an encrypting/decrypting device, and it is necessary to specify only the second function p(.). The first and second functions are defined on the same phase space, that is, they are defined in the same domain and their images are in this domain.

[0028] More precisely, the invention provides an encryption process that comprises the following steps: preliminarily choosing at least a secret key and a set of permutable functions defined on a certain phase space for encrypting/decrypting messages; choosing a code for encoding messages to be sent in the form of a number belonging to the phase space; and wherein the set of permutable functions is composed of chaotic maps generated by a composite function of a first function, the inverse function of the first function and a second function. The secret key is defined by using the second function.

[0029] Preferably the chaotic maps are Chebyshev maps, the degree of which is a secret key, or even chaotic maps, the second function of which is a trapdoor function (see [8]).

[0030] The invention provides further digital signature processes for a public-key encryption process according to the method of this invention that use chaotic maps, the second function of which is defined by a parameter (e.g., Chebyshev maps), and for a public-key encryption process that uses trapdoor functions.

[0031] The processes according to this invention may be easily implemented by a software computer program.

[0033] The encryption process of the invention is based on the use of chaotic maps that are composite functions f(p(f⁻¹(.))) generated by composing a first function f(.), a second function p(.) and the inverse of the first function f⁻¹(.). The functions f(.) and p(.) are defined on a phase space PS, that is, they are defined on a certain domain and their images are in the same domain, that is
f(.):PS→PS; p(.):PS→PS.

[0034] The phase space PS can be any domain and must be the same for functions f(.) and p(.). The invention essentially is based upon encrypting/decrypting messages by using permutable chaotic maps sharing the same first function f(.), and defining a secret key by using the second function p(.). Chaotic maps may be used both in public-key encryption processes and in processes that contemplate only a secret key.

[0035] Section 1: Symmetric-key encryption process. A symmetric key encryption process of the invention is implemented by using two permutable chaotic maps, a first f(p_S(f⁻¹(.))) and a second f(p_R(f⁻¹(.))), which means that p_S(p_R(.))=p_R(p_S(.)). The second functions of which p_S(.) and p_R(.) are invertible in the phase space.

[0036] A sender encrypts a message M, representing it in numerical form, to be sent to a recipient, by calculating the image thereof f(p_S(f⁻¹(M))) through the sender chaotic map. On his turn the recipient encrypts the encrypted message using his own chaotic map, sending back to the sender a twice encrypted message f(p_R(p_S(f⁻¹(M))))=f(p_S(p_R(f⁻¹(M)))).

[0037] The sender decrypts the twice encrypted message, sending to the recipient the quantity f(p_R(f⁻¹(M))) because
f(p_S⁻¹(f⁻¹(f(p_R(p_S(f⁻¹(M)))))))=f(p_S⁻¹(p_S(p_R(f⁻¹(M)))))=f(p_R(f⁻¹(M))).

[0038] Finally, the recipient recovers the message M by decrypting the quantity f(p_R(f⁻¹(M))).

[0039] Section 2: Public-key encryption process. Chaotic maps may be used also in public key encryption processes and in related digital signature processes. A first public-key encryption process of the invention that uses chaotic maps is based on the so-called trapdoor functions (see [8]), which are functions the inverse of which is extremely difficult to compute.

[0040] According to this encryption process, a recipient of a message chooses a second function p_R(.) as a public key, the inverse of which p_R⁻¹(.) is hardly computable constitutes his private key. A sender of a message encrypts an encoded message M by calculating the image thereof through the chaotic map using the public key of the recipient of the message M. Finally, the recipient decrypts the encrypted message by using his private key p_R⁻¹(.).

[0041] A digital signature for a message encrypted according to the above process is generated as follows. A sender encrypts a digital signature SIG, which is substantially a number representing a pre-established string or even a message M to be sent, by calculating the image thereof through the chaotic map f(p_S⁻¹(f⁻¹(.))) using his own private key p_S⁻¹(.), and encrypts a message M together with the encrypted digital signature appended thereto by calculating the image of the combination of the message and the encrypted digital signature through the chaotic map f(p_R(f⁻¹(.))) using the recipient public-key p_R(.).

[0042] A recipient decrypts the received encrypted message by calculating the image thereof through the chaotic map f(p_R⁻¹(f⁻¹(.))) using his own private key p_R⁻¹(.), and obtains the digital signature SIG decrypting the just decrypted message by calculating the image thereof through the chaotic map f(p_S(f⁻¹(.))) using the sender public key p_S(.).

[0043] Section 3: Chebyshev polynomials. For sake of simplicity, an embodiment of an encryption process of this invention is illustrated in the ensuing description by referring to Chebyshev polynomial maps, which are a particular kind of chaotic maps, though what will be similar for a chaotic map of any other kind.

[0044] A Chebyshev polynomial map T_p(.):[−1,1]→[−1,1] of degree p is

 T_p(x)=cos(p·cos⁻¹(x))  (1)

[0045] where p=2, 3, . . . This map has unique, absolutely continuous invariant measure: $\mu \left(x\right)\mathrm{dx}=\frac{\mathrm{dx}}{\pi \sqrt{1-x^2}}$

with a positive Lyapunov exponent λ=ln p. For p=2, Eq. (1) becomes the well-known logistic map [33]. One of the most remarkable properties of the Chebyshev polynomials is the so-called semi-group property [25]:
T_r(T_s(x))=T_rs(x)  (2)
An immediate consequence of this property is that Chebyshev polynomials are permutable, i.e.,
T_r(T_s(x))=T_s(T_r(x))  ( 3)
Let x_n be the trajectory of the Chebyshev map x_n+1=T_p(x_n) for n=0, 1, 2 . . . in the phase space [−1,1] with initial condition x₀. Therefore, for Eq. (2)
x_n=T_p(T_p( . . . T_p(x₀)))=T_pn(x₀).

[0046] Several low order Chebyshev polynomials are:

 T₂(x)=2x²−1
T₃(x)=4x³−3x²
T₄(x)=8x⁴−8x²+1
and are graphically shown in FIG. 1.

[0047] Section 4: A second public-key encryption process. A second public-key encryption process of the invention is in some measure similar to the ElGamal public-key process, but is different from it since it uses chaotic maps. A preferred embodiment thereof is now presented again using Chebyshev maps for an example, but that may be easily generalized for employing any other chaotic map.

[0048] A recipient A of a message M generates an integer “r” (preferably a large integer), selects a random number xε[−1,1] and computes T_r(x). The recipient public key is (x, T_r(x)), while his private key is the integer “r”.

[0049] A sender B encrypts a message M to be sent to the recipient A by performing the following steps: obtaining the recipient authentic public key (x, T_r(x)); encoding the message into a number Mε[−1,1]; generating an integer “s” (preferably a relatively large integer, as explained above); calculating T_s(x) and T_sr(x)=T_s(T_r(x)); calculating X=q_Ts(Tr(x))(M); and sending the cipher-text c=(T_s(x),X) to the recipient A. The function q_par(.) is a pre-established invertible parametric function defined almost everywhere on the phase space of the chaotic map, which in this case is the domain [−1,1].

[0050] The recipient A recovers the plain-text M from the cipher text “c” by using his private key “r” to calculate T_rs(x)=T_r(T_s(x)), which is equal to T_sr(x); and calculating M=q_Trs(x)⁻¹(X).

[0051] The function q_par(.) should be defined almost everywhere on [−1,1], that is, it should have at most isolated singular points for parε[−1,1]. This requisite is necessary because if this function has at most isolated points of discontinuity, the probability that the value X does not exist is theoretically null.

[0052] The parametric function q_par(.) may be defined as the product between its argument by the parameter “par”. It may be defined also as the ratio between the argument of the function by the parameter “par”, or the sum or even the arithmetic mean of them. Let us suppose that this function is q_par(M)=M·par.

[0053] Section 5: Implementation. There are two problems related to the software implementation of the algorithm. The first problem is that, encryption, decryption, signing and verifying signatures all involve multiplications by large numbers. This operation is preferably performed as follows. The number “s” is first factorized: $\begin{array}{cc}s=\underset{\underset{k_1}{︸}}{s_1\cdots s_1}\underset{\underset{k_2}{︸}}{s_2\cdots s_2}\cdots \underset{\underset{k_i}{︸}}{s_i\cdots s_i}=s_1^{k_1}{s}_2^{k_2}\cdots s_i^{k_i}& \left(4\right)\end{array}$
Then T_s(y) is calculated by iterating k₁+k₂ . . . +k_i times the Chebyshev map according to the following formula $\begin{array}{cc}{T}_s\left(y\right)=\underset{\underset{k_1}{︸}}{T_{s_1}(\cdots T_{s_1}}\cdots \underset{\underset{k_i}{︸}}{T_{s_i}\cdots T_{s_i}}\left(y\right)))& \left(5\right)\end{array}$
The fact that Chebyshev polynomials satisfy Eq. (5) makes them particularly suited for implementing the encryption process of the invention.

[0054] The second problem is related to the fact that the message M is recovered by using Eq. (3). However, due to the chaotic nature of the Chebyshev map, Eq. (3) holds only for s<s₀ and r<r₀, where s₀ and r₀ are numbers which depend on the arithmetic precision used for implementing the algorithm. The following TABLE 1 shows the values of s₀ and r₀ using different precision arithmetic and imposing that a 128 bits message be decrypted.

[0055]

	TABLE 1
	N-bit precision	r0, s0
	N = 256	˜270
	N = 512	˜2190
	N = 1024	˜2450
	N = 2048	˜29700

[0056] Preferably, the present encryption process is implemented by software that uses the GNU multiple precision arithmetic library [31]. GNU MP is a portable library written in C language for arbitrary precision arithmetic on integers, rational functions, and floating-point numbers. The library GNU MP does not have a routine for computing cosine functions with arbitrary precision. For this reason the value of T_s(x) is calculated by using the following recursive relation between Chebyshev polynomials:

 T_p+1(x)=2x·T_p(x)−T_p−1(x)  (6)

with T₀(x)=1 and T₁(x)=x.

[0057] Let's show a numerical example, using a 2048-bit precision arithmetic. The private and public keys of the entity A are
r=2⁶⁴·3³⁶·113²¹·23⁸¹·59⁴⁸≡2⁹¹⁰(x,T_r(x))=(0.25749480,−0.0176128306 . . . )
respectively.

[0058] The private key “s” of the entity B may be
s=3⁵⁴·7²²·41⁵⁵·133¹⁹·31⁷⁰≅2⁹²⁰
and a sample message to be sent may be $M=0.\underset{\underset{10}{︸}}{1\cdots 1}\underset{\underset{10}{︸}}{4\cdots 4}.$

[0059] The sender B computes
T_s(x)=0.9921943793 . . .
T_s(T_r(x))=0.6571609510 . . .
X=−0.7301788346 . . .
and sends the cipher-text c=(T_r(x),X) to the recipient A, which on his turn successfully recovers the plain-text message M.

[0060] Section 6: Considerations. According to another embodiment of this invention, the chaotic maps are Chebyshev polynomial maps defined according to the following formula:
cos(b^p·cos⁻¹(y))
wherein the order b^p of the map is a power of a pre-established integer base b raised to an integer exponent p. Images of these Chebyshev polynomial maps may be conveniently calculated using equations (4), (5) and (6) and can be used in the encryption processes of this invention wherein the integer exponent p is the secret key.

[0061] Advantages and peculiarities of the embodiment of an encryption process of this invention using Chebyshev polynomial maps are mentioned below.

• • [0062] 1. The main advantage over existing public-key encryption algorithms is that the private key can be any large integer. There is no need to look for large prime numbers. As can be seen from Eq. (4), generating a large number “s” and using it with Chebyshev polynomial maps as a secret key is easy.
  • [0063] 2. It is reasonably secure even for relatively small numbers.
  • [0064] 3. The algorithm is reasonably fast. On an Intel Pentium 550 MHz processor, the present algorithm encrypts and decrypts 128-bit messages using 2048 bit precision arithmetic in less than 1s. The total number of multiplications and additions required by the encryption process, computing T_s(x) by equations (5) and (6), is $d=2\left(\sum _{j=1}^{j=i}{s}_j{k}_j\right).$
    It is possible to optimize and/or speed up the algorithm by minimizing “d” while keeping “s” large or by using a procedure for computing cosine functions with arbitrary precision.
  • [0065] 4. An advantage of using Chebyshev maps is that x and T_s(x) (or T_s(x) and X) are independent random variables.
  • [0066] 5. It is possible to realize a relative digital signature process.
  • [0067] 6. The present encryption process, as well as the basic ElGamal algorithm, uses random numbers.
  • [0068] 7. A disadvantage of the algorithm of this invention, which comes from the chaotic property of the Chebyshev map, is that the length of the plain-text message should always be smaller than the precision used in the algorithm.
  • [0069] 8. Another disadvantage of the encryption process that derives from the ElGamal-type structure of the algorithm, is that the cipher-text “c” is in the form c=(T_s(x), X) so that the total length of the cipher-text, for 2048-bit arithmetic precision, is 4096 bits.
  • [0070] 9. Let us suppose that the message M is a plain English text written in the following form M=±0.a₁a₂ . . . a_k, where a_iε{0,1, . . . 9}. Given that T_r(x)ε[−1,1], an adversary could use this strategy to break the preferred embodiment of the encryption process in which X=M·T_rs(x): by dividing the interval [−1, 1] in a finite number of sub-intervals, choosing a number y from each sub-interval, computing M=X/y and guessing the meaning of a first digit a_i of the message M. In order to jeopardize this strategy, the sender B should choose the first 10 digits a_i (that is for i=1, 2, . . . , 10) in a random mode (which do not have to be known by the recipient A). The hypothetic attack would not work if all a_i are random numbers, as it happens when the message M is a key for a symmetric-key encryption process.

[0071] The just described algorithm, wherein X=M·T_rs(x), works with Chebyshev polynomials, but it can be generalized to work with any chaotic map T_r(.)=f(r·f⁻¹(.)), the second function of which is a linear function.

[0072] Now the following question is addressed: is there any other chaotic map with the semi-group property T_r(T_s(.))=T_rs(.)? Let us first consider polynomials. Two polynomials, P(.) and Q(.), are permutable if P(Q(x))=Q(P(x)) for all x. Let us write P∘Q to indicate the composite function P(Q(.)).

[0073] A set of polynomials of positive degree and such that any two polynomials are permutable, is called a chain. The Chebyshev polynomials T₁(x), T₂(x), . . . form a chain. The powers π_j(x)≡x^j, j=1, 2 . . . form a chain as well. Suppose that λ(x)=ax+b, a≠0 and λ⁻¹(x)=(x−b)/a. If P(.) and Q(.) are permutable, also λ⁻¹∘P∘λ and λ⁻¹∘Q∘λ are permutable. Let us say that P(.) and λ⁻¹∘P∘λ are similar.

[0074] The answer to the above question for polynomials is given by the following theorem [25]: if P(.) and Q(.) are permutable, either both are iterations of the same polynomial or both are similar, with respect to the same function λ, to either Chebyshev polynomials or powers. Thus, the sequences {T_j} and {π_j} are the only chains of polynomials, up to similarities.

[0075] Powers have been used in cryptography for a long time. For example, powers are used for encryption and decryption in the ElGamal public-key process. In this case the set of functions is defined as π_p=x^pmod n, where n is a prime number, x is a generator of the multiplicative group Z*_n, and p is an integer such that 1≦p≦n−2. Any two functions belonging to π_p, ∀p=1, 2 . . . n−2, are permutable.

[0076] Whether the chain {π_j}=x^j, j=1,2, . . . , xεR, can be used in a public-key encryption process is still an open question. Considering only the functions x^jmod 1, defined on the phase space [0, 1], which form a chain, then the dynamic is trivial: for j≧2 a trajectory of the map approaches zero. However, the following class of functions T_p(x)=((x+1)^p−1)mod 1, defined on the phase space [0, 1], which are chaotic, can be used in public-key encryption process. Another example of chaotic maps that can be used in encryption processes is
T_p(x)=cos └(2+cos⁻¹(x)^p−2┘
defined on [−1, 1].

[0077] At the present time only one class of rational maps with semi-group property is known: the so-called Jacobian elliptic Chebyshev rational maps introduced recently by Kohda and Fujisaki [32]. This embodiment of the second public-key encryption process of this invention works with these maps as well.

[0078] The ElGamal signature algorithm is a randomized signature process. It generates digital signatures appended on binary messages of arbitrary length, and requires a hash function h:{0,1}*→Z_p where p is a large prime number and Z_p is the set of integers modulo p. The DSA (Digital Signature Algorithm) is a variation of the ElGamal signature algorithm. For an introduction to hash functions, see [8].

[0079] Section 7: Another digital signature process. Yet another object of this present invention is a digital signature process for the above described second public-key encryption process, that contemplates the use of chaotic maps, the second function of which may be fully identified by a parameter, such as the Chebyshev polynomial maps.

[0080] Substantially, a sender performs the following steps: generates a public key (x, T_a(x)), wherein a first integer number “a” is the parameter of the chaotic map; generates a second integer number “k” and calculating a second image T_k(x) of the random number x through the chaotic map T_k(.), the parameter of which is the second integer number k; generates a verification number h; chooses a first test function V₁(.), a second test function V₂(.) and a third test function H(.), which in general depend on four independent variables, such that
V₁(T_a(x),T_k(x),H(h,T_k(x),k,a),x)≡V₂(h,T_k(x),H(h,T_k(x),k,a),x);
A second verification number s is generated using the third test function H(.) according to the following formula s=H(h,T_k(x),k,a). The pair (T_k(x), s) is sent as signature of a message M.

[0081] A receiver performs the following steps: obtains the authentic sender public key (x, T_a(x)); calculates a first test value v₁ using the first test function V₁(.) according to the following formula v₁=V₁(T_a(x),T_k(x),s,x); calculates a second test value v₂ using the second test function V₂(.) and the verification number h according to the following formula v₂=V₂(h,T_k(x),s,x); and verifies whether the test values first v₁ and second v₂ are identical.

[0082] Obviously, the test functions must be designed so that only the owner of the private key “a” is capable of generating a number “s” for which v₁=v₂. As it may be noticed, this digital signature process does not need any hash function, because the verification number “h” may be any pre-established number. Of course, it is possible to use hash functions in this digital signature process by calculating the verification number as the image through a hash h(.) function of the message M to be sent.

[0083] Two possible embodiments of this digital signature process for the case in which Chebyshev maps are used, are illustrated. According to a first embodiment, it is necessary to define a first g₁(.) and a second g₂(.) auxiliary function such that
g₁(h,T_k(x))·k=g₂(s)·a;  (7)
the first test value v₁ is the image of the random number x through the chaotic map according to the following formula v₁=T_g2(s)(T_a(x)); the second test value v₂ is the image of the random number x through the chaotic map according to the following formula v₂=T_g1(h,Tk(x))(T_k(x)).

[0084] Preferably, the first auxiliary function returns the value of the verification number h
g₁(h,T_k(x))=h
the second auxiliary function g₂(.) is the identity function
g₂(s)=s
and so, considering Eq. (7), $s=\frac{h·k}{a}.$

[0085] According to another embodiment of the digital signature process of this invention, the second verification number s is the ratio between the difference between the verification number h and the product between the integer a by the value T_k(x), and the second integer number k, $\frac{h-a·T_k\left(x\right)}{k}$
while the second test value v₂ is the image of the random number x through the Chebyshev map cos(h·cos⁻¹(.)) the order of which is the verification number h.

[0086] The receiver carries out the following operations: calculates an auxiliary value T_Tk(x)(T_a(x))=T_Tk(x)·a(x); calculates a second auxiliary value T_s(T_k(x))=T_sk(x); calculates the first test value v₁

 v₁=T_Tk(x)a+ks(x)

by combining the auxiliary values T_Tk(x)a(x) and T_sk(x) according to the following equation
v₁=T_Tk(x)a+ks(x)=T_Tk(x)a(x)·T_ks(x)−√{square root over ((1−T_Tk(x)a²(x))·(1−T_ks²(x)))}{square root over ((1−T_Tk(x)a²(x))·(1−T_ks²(x)))}{square root over ((1−T_Tk(x)a²(x))·(1−T_ks²(x)))}  (8)
Proof: given that
T_Tk(x)a+ks(x)=cos └(T_k(x)·a+ks)·cos⁻¹(x)┘
applying the following trigonometric identity
cos(y+z)=cos y·cos z−sin y·sin z
it is possible to state that
cos └(T_k(x)·a+ks)·cos⁻¹(x)┘=cos └T_k(x)·a·cos⁻¹(x)┘·cos └ks·cos⁻¹(x)┘+−sin └T_k(x)·a·cos⁻¹(x)┘·sin └ks·cos⁻¹(x)┘
and thus Eq. (8) is demonstrated.

[0087] Section 8: Further improvements. Public-key encryption processes that use Chebyshev maps are both secure and practical and can be used for both encryption and digital signature. In the previous works [19, 20, 21] the ElGamal public-key algorithm was extended to Chebyshev maps, defined on the set [−1, 1] and implemented using floating-point arithmetic.

[0088] According to another embodiment of this invention, it is possible to realize encryption processes using chaotic maps that always return integer numbers. Such encryption processes could be even faster than the previously disclosed processes because they contemplate only operations on integer numbers, that are performed faster than operations on floating point numbers.

[0089] ElGamal-like and RSA-like public-key algorithms using Chebyshev maps are proposed. Furthermore, the analysis of the periodic orbits in sequences of integers generated by Chebyshev maps is based on the arithmetic properties of torus automorphisms, another well-known class of chaotic maps.

[0090] The encryption process of the invention may be implemented also using torus automorphisms as chaotic maps. In order to explain more clearly how it is possible to use torus automorphisms in an encryption process of the invention, a brief introduction about torus automorphisms is given.

[0091] Section 9: Torus automorphisms. In this section some general properties of automorphisms of the two-dimensional torus are briefly discussed. An automorphism of the two-torus is implemented by 2×2 matrix M with integer entities and determinant ±1. The requirement that the matrix M has integer entities ensures that M maps torus into itself. The requirement that the determinant of the matrix M is ±1 guarantees invertibility. Here only strictly unimodular automorphisms, for which det M=1, are considered.

[0092] Let M be a 2-torus automorphism $\left[\begin{array}{c}{x}^{\prime }\\ y^{\prime }\end{array}\right]=M\left[\begin{array}{c}x\\ y\end{array}\right]\left(\mathrm{mod}1\right)$
where x, yε[0, 1]. Let k be a trace (which is an integer) of the automorphism M, f(z)=z²−kz+1 its characteristic polynomial, and λ one of its roots (for example the largest one): $\lambda =\frac{k+\sqrt{k^2-4}}{2}$
It is well-known that for k>2 (considering only positive k) that the automorphism M has strong chaotic properties, and in particular, it has a dense set of unstable periodic orbits. The detail structure of periodic orbits of the 2-torus automorphisms has been studied by Percival and Vivaldi [26].

[0093] Periodic orbits of a torus automorphism include those points having rational coordinates ξ=p₁/q₁, η=p₂/q₂, p_i, q_i integers. Let p_i, q_i be co-primes (their greatest common divisor is 1) and q be the least common multiple of q₁ and q₂. Clearing denominators, letting M act on Z², the lattice of integral vectors, and then the periodicity of the torus is taken into account by identifying points whose coordinates differ multiplies of q, i.e., the factor group Z²/gZ² is considered. Thus, the dynamics of periodic orbits is dynamics over a finite set of integers.

[0094] The article [26] illustrates the close link existing between arithmetic in algebraic number fields and strong chaotic dynamics. Main conclusions of Ref. [26] may be summarized as follows.

[0095] A 2-torus automorphism has three different types of (periodic) orbit structure, according to the classification of rational primes: inert, split and ramified primes [27].

[0096] The orbits which correspond to inert primes are almost without structure. The split primes have two distinct ideal factors, which correspond to orbits confined to invariant sublattices. For this reason, two ideal orbits which exist on split prime lattices are the “most ergodic” orbits, and thus, equilibrium averages computed with them minimize statistical fluctuations.

[0097] Both inert and split prime lattices are found infinitely often, and moreover with the same frequency in both cases. These are consequences of Dirichelet's theorem on the existence of infinity many primes in any arithmetic progression [28].

[0098] The ramified prime lattices support orbits which are exceptionally regular. However, there is only a finite number of ramified primes, so that this apparently contradictory phenomenon of regularity in chaos is in fact very rare.

[0099] Section 10: Floating-point arithmetic versus integer arithmetic. Chaotic systems are defined on real numbers. Any encryption algorithm which uses chaotic maps when implemented on a computer (finite-state machine) becomes a transformation from a finite set onto itself. Because of its wide dynamic range, the floating-point implementation seems to be the most appropriate for software realizations (implementation) of Chebyshev polynomials.

[0100] However, there are three reasons for not using floating-point arithmetic in public-key encryption. First, floating-point numbers are not uniformly distributed over any given interval of the real axis [29]. Furthermore, one may observe the existence of redundant number representations. Indeed, due to the normalized calculations in floating-point arithmetic, some floating-point numbers represent the same real signal value.

[0101] Second, non-invertibility of Chebyshev polynomials and their floating-point implementation imply a restriction on the length of the message. Indeed, the public-key encryption scheme proposed in [20, 21] and in this patent application can be viewed as a generalization of the ElGamal public-key process using Chebyshev polynomials.

[0102] The algorithm is summarized for sake of clarity. Alice generates a large integer s, selects a random number xε[−1,1] and computes T_s(x). Alice's public key is (x, T_s(x)), while her private key is s. Bob represents the message as a number Mε[−1,1], generates a large integer r and computes T_r(x), T_rs(x)=T_r(T_s(x)) and X=MT_rs. Bob sends the cipher-text c=(T_r(x),X) to Alice. To recover plain-text M from c, Alice uses the private key s to compute T_sr(x)=T_s(T_r(x)), and recovers M by computing $M=\frac{X}{T_{\mathrm{sr}}}.$
Let l_s, l_r, l_M be the lengths (in bits) of s, r, and M, respectively, and let N-bit precision arithmetic is used in a software implementation of the algorithm: then l_m≦N−l_s−l_r (see [20, 21]).

[0103] Third, the most important reason, there are no analytical tools for understanding the periodic structure of the periodic orbits in the floating-point implementation of chaotic maps (when implement on a computer all chaotic maps are periodic: all trajectories are eventually periodic). On the other hand, when using integers one may hope if a possible link between number theory and chaos theory has been established, as in the case of the torus automorphisms, to understand the structure of the orbits.

[0104] Section 11: Modified Chebyshev polynominals. In this section the following map is used
T_p: {0,1, . . . , N−1}→{0,1, . . . , N−1}
defined as:
y=T_p(x)(mod N)  (9)
where x and N are integers and T_p(x) is a Chebyshev map, to extend ElGamal and RSA public-key algorithms. The chaotic map (9) will be referred as a “modified Chebyshev polynomial”.

[0105] The modified Chebyshev polynomials can replace powers in ElGamal and/or RSA public-key algorithms only if they commute under composition and one can compute the period of their orbits.

[0106] The following two theorems show properties of modified Chebyshev polynomials. Theorem 1. Modified Chebyshev polynomials commute under composition, that is:
T_p(T_q(x)(mod N))(mod N)=T_pq(x)(mod N)
Theorem 2. Let N be an odd prime and xεZ such that 0≦x<N. Then the period of the sequence T_n(x)(mod N), for n=0, 1, 2 . . . is a divisor of N²−1.

[0107] The first theorem can be easily verified; the proof of the second theorem is given in Section 17. Sample trajectories of the map (9) for different values of the variable x when N=19 are given below:

• • [0108] x=0 1, 0, 18, 0, 1, 0, 18, 0, . . .
  • [0109] x=1 1, 1, 1, 1, . . .
  • [0110] x=2 1, 2, 7, 7, 2, 1, 2, 7, 7, 2, . . .
  • [0111] x=3 1, 3, 17, 4, 7, 0, 12, 15, 2, 16, 18, 16, 2, 15, 12, 0, 7, 4, 17, 3, . . .
  • [0112] x=4 1, 4, 12, 16, 2, 0, 17, 3, 7, 15, 18, 15, 7, 3, 17, 0, 2, 16, 12, 4, . . .
  • [0113] x=5 1, 5, 11, 10, 13, 6, 9, 8, 14, 18, 14, 8, 9, 6, 13, 10, 11, 5, . . .

[0114] The periods of all trajectories of the map (9) with N=19 are listed in the following Table

[0115]

TABLE 2
X	0	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18
Period	4	1	5	20	20	18	18	5	18	3	6	9	10	9	9	20	20	10	2

[0116] These periods always divide the number 18×20=2³3²5. One can easily show that for any odd prime N the periods of the trajectories starting from initial points x=0, x=1, and x=N−1 are always 4, 1, 2 respectively.

[0117] Section 12: Software implementation. In a public-key algorithm encryption, decryption, signing, and verifying signatures all involve multiplying with a large number. An algorithm for computing T_p(x)(mod N) when N and p are large numbers is now presented.

[0118] Equation (6) can be rewritten as $\begin{array}{cc}\left[\begin{array}{c}{T}_p\\ T_{p+1}\end{array}\right]=\left[\begin{array}{cc}0& 1\\ -1& 2x\end{array}\right]·\left[\begin{array}{c}{T}_{p-1}\\ T_p\end{array}\right]=A·\left[\begin{array}{c}{T}_{p-1}\\ T_p\end{array}\right]& \left(10\right)\end{array}$
or, after some algebra, as $\begin{array}{cc}\left[\begin{array}{c}{T}_p\\ T_{p+1}\end{array}\right]=A^p·\left[\begin{array}{c}{T}_0\\ T_1\end{array}\right]& \left(11\right)\end{array}$
Matrix exponentiation can be done effectively by the square and multiply algorithm. Using notation $I=\left[\begin{array}{cc}1& 0\\ 0& 1\end{array}\right]$
the algorithm in pseudo-C language for calculating the matrix exponent A^p is:
A^p=I; for (i=p.numBits( ); i>0; i—)
{A^p=(A^p)²; if (p.bitAt(i)==1) A^p=A^pA;}.

[0119] More in detail, an auxiliary matrix A^p is preliminarily set equal to the identity matrix. The quantity p.numBits( ) is the number of bits of the binary string representing the number p. An iterative cycle begins: a pointer “i” is initially equal to p.numBits( ) and is decreased at each iteration step until it nullifies. During each iteration step, the auxiliary matrix is powered to an exponent 2 and it is to be checked whether the i-th least (or most) significant bit of the binary string representing the number p equals 1 or not: if this condition is verified, the auxiliary array is multiplied by the matrix A and a successive iteration step is performed if i>0. When the iterative cycle ends, the auxiliary matrix is the exponent matrix to be calculated.

[0120] This algorithm is a version of the number exponentiation algorithm used in commercial asymmetric encryption algorithms, adapted to work on matrices.

[0121] The T_p(x)(mod N) calculation speed has been tested on Intel Pentium 1700 MHz processor with 512 MB RAM, using the equation (5). The test includes Java [30] and GNU Multiple Precision library [31] implementation. For N and p of order 1024 bits, calculating T_p(x)(mod N) takes
Java: ˜700 ms
GMP: ˜70 ms.

[0122] Section 13: ElGamal-like public-key encryption with modified Chebyshev polynomials. The ElGamal public-key encryption scheme can be viewed as a Diffie-Hellman key agreement in key transfer-mode [8]. Its security is based on the intractability of the discrete logarithm-problem and the Diffie-Hellman problem. The basic ElGamal and generalized ElGamal encryption schemes are described in [8].

[0123] Another encryption process of this invention is obtained by generalizing the ElGamal encryption scheme for using modified Chebyshev polynomials of Eq. (9). The ElGamal-like public-key encryption process of this invention comprises an algorithm for key generation and an algorithm for encrypting/decrypting a message.

[0124] Algorithm for key generation. A recipient should perform the following operations:

• • [0125] 1. Choosing a large random prime N and an integer x such that x<N.
  • [0126] 2. Choosing a random integer s<N and computing A=T_s(x)(mod N).
  • [0127] 3. The recipient's public key is (x, N, A), while his private key is s.

[0128] Algorithm for ElGamal-like public-key encryption. 1. Encryption. To encrypt a message m, a sender should perform the following operations:

• • [0129] (a) Obtaining the recipient's authentic public key (x, N, A).
  • [0130] (b) Encoding the message as an integer m in the range {0, 1, . . . , N−1}.
  • [0131] (c) Selecting a random integer r<N.
  • [0132] (d) Computing B=T_r(x)(mod N) and X=mT_r(A)(mod N).
  • [0133] (e) Sending the cipher-text c=(B, X) to Alice.
  • [0134] 2. Decryption. To recover the message m from c, the recipient should perform the following operations:
  • [0135] (f) Using the private key s to compute C=T_s(B)(mod N).
  • [0136] (g) Recovering m by computing m=XC⁻¹(mod N).

[0137] Proof that decryption works. It follows from the fact that

 T_s(B)=T_s(T_r(x))=T_r(T_s(x))=T_r(A)

[0138] The just described encryption process may be easily generalized by choosing preliminarily an invertible parametric function q_par(.) defined almost everywhere on the phase space (that is, the set of eventual singular points of this parametric function does not have any accumulation point) in function of the parameter par, then encrypting a message performing the step (d) calculating the variable X as follows:
X=q_Tr(A)(m)(mod N)
The recipient recovers the message m by performing the step (g) using the following formula:
m=q_Ts(B)⁻¹(X)(mod N)
wherein q_par⁻¹(.) is the inverse of the parametric function q_par(.). The parametric function may be the product between the argument of the function by its parameter as in the embodiment presented above, or alternatively it may be the sum or the ratio between the argument of the function by its parameter, or any other invertible parametric function.

[0139] An example of application of the described encryption process with artificially small parameters is presented below. Key generation. Alice chooses the prime N=1749940627, integers x=25749480 and s=7207480595, and computes A=173359085. Alice's public key is (N=1749940627, x=25749480, A=173359085), while her private key is s=7207480595.

[0140] Encryption. To encrypt a message m=11223344, Bob chooses an integer r=6431562606 and computes B=1399079193 and X=878048528. He sends the cipher text c=(B, X)=(1399079193, 878048528) to Alice. Decryption. To recover the message m from c, Alice computes C=1376040233 and m=11223344.

[0141] Security. If x>1, the Chebyshev polynomial T_n(x) can be written as
T_n(x)=cos h(n cos h⁻¹(x))
Thus, if y=T_n(x)(mod N), then, after some algebra, one finds
n=log_{x+√{square root over (x2 −1)}(}y+√{square root over (y²−1)})
In the case where both square roots √{square root over (x²−1)} and √{square root over (y²−1)} exist in GF(N), which is the Galois field of order N, one has a conventional discrete logarithm problem. On the other hand, if at least one of the square roots exists in the quadratic extension field GF(N²), this leads to a quadratic extension field generalization of the discrete logarithm problem. Thus, the modified ElGamal public-key algorithm of this invention is as secure as the original ElGamal algorithm.

[0142] Section 14: RSA-like public-key encryption with modified Chebyshev polynomials. The RSA-like encryption process, named after its inventors R. Rivest, A. Shamir, and L. Adleman, is the most widely used public-key encryption process. It may be used to provide both secrecy and digital signatures and its security is based on the intractability of the integer factorization problem.

[0143] This section describes a RSA-like encryption process according to this invention for Chebyshev polynomials. As for the RSA encryption process, the encryption process presented below can be used for both encryption and digital signature and its security is based on the intractability of the integer factorization problem.

[0144] The RSA-like public-key encryption process of this invention comprises an algorithm for key generation and an algorithm for encrypting/decrypting a message.

[0145] Algorithm for key generation. A recipient should perform the following operations:

• • [0146] 1. Choosing two large random (and distinct) primes p and q, roughly of the same size.
  • [0147] 2. Computing N=pq and φ=(p²−1)(q²−1).
  • [0148] 3. Selecting a random integer e, 1<e<φ, such that “e” and “φ” are coprime integers, that is gcd(e, φ)=1.
  • [0149] 4. Computing the unique integer d, 1<d<φ, such that ed≡1(mod φ).
  • [0150] 5. The recipient's public key is (N, e); the recipient's private key is d.

[0151] Algorithm for encryption. 1. Encryption. To encrypt a message m, a sender should perform the following operations:

• • [0152] (a) Obtaining the recipient's authentic public key (N, e).
  • [0153] (b) Encoding the message in an integer in the interval [1, N−1].
  • [0154] (c) Computing c=T_e(m)(mod N) according to Eq. (9) and sending it to the recipient.
  • [0155] 2. Decryption. To recover the message m from c, the recipient should perform the following operation:
  • [0156] (d) Using the private key d to recover m=T_d(c)(mod N).

[0157] The integers e and d in RSA key generation are called the encryption exponent and the decryption exponent, respectively, while N is called the modulus.

[0158] Proof that decryption works.—It was shown in Section 11 that if p is an odd prime number and 0≦g<p, then the period of the sequence T_n(g)(mod p), n=0, 1, . . . is a divisor of p²−1. Since ed≡1(mod φ), there is an integer k such that ed=1+kφ. Thus, one finds
T_d(T_e(x))≡T_de(x)≡T_1+kφ(x)≡T₁(x)≡x(mod p)
By the same arguments,
T_d(T_e(x))≡T_de(x)≡T_1+kφ(x)≡T₁(x)≡x(mod q)
Finally, since p and q are distinct primes, it is possible to use Chinese remainder theorem to show that
T_d(T_e(x))≡T_de(x)≡T_1+kφ(x)≡T₁(x)≡x(mod N).

[0159] An example of application of this algorithm with artificially small parameters is presented below.

[0160] Key generation. Alice chooses the primes p=21787 and q=3793 and computes N=82638091 and φ=6829053595064064. Alice chooses e=65537 and, using the extended Euclidean algorithm, finds d=2150406320724737. Alice's public key is the pair (N=82638091, e=65537), while her private key is d=2150406320724737.

[0161] Encryption. To encrypt a message m=11223344, Bob computes
c=T₆₅₅₃₇(11223344)(mod 82638091)=12355612
Decryption. To decrypt c, Alice computes
T_d(c)(mod N)=T_{2150406320724737}(12355612)(mod 82638091)=11223344
Like the well known RSA and ElGamal encryption processes, digital signature processes for both the RSA-like and the ElGamal-like encryption processes of this invention may be realized. For example, the digital signature process for public-key encryption processes already disclosed in Section 2 may be used.

[0162] Section 15: Ideal theory in quadratic fields. In this section the ideal theory in quadratic fields is briefly summarized following [27, 28]. Quadratic integers—The solutions of the linear equations with integral coefficient, ax+b=0, form the field of rational numbers. If the leading coefficient is equal to 1, a=1, the solutions are integers. Following Dedekind, quadratic irrationals are defined as the solutions of quadratic equations with integral coefficients, whereas quadratic equations whose leading coefficient is 1 yield quadratic integers. Thus (1+√{square root over (5)})/2 and 2i are quadratic integers, since they satisfy the equations x²−x−1=0 and x²+4=0, respectively. Quadratic integers coincide with the eigenvalues of 2 by 2 integral matrices. Sometimes, when possibility of confusion arises, ordinary integers will be called rational integers.

[0163] Norm, units and primes—By analogy with complex conjugates, the conjugate of a quadratic irrational z=(a+b√{square root over (D)})/c is defined as z′=(a−b√{square root over (D)})/c. The number zz′=N(z) is called norm of z. Then N(z)=N(z′) and N(zv)=N(z)N(v). In real fields, the norm of the number has nothing to do with its actual magnitude, and can be even negative. The norm of a quadratic integer is a rational integer.

[0164] The divisors of all rational integers are just 1 and −1, which are called units. The units of a quadratic field are precisely those quadratic integers of the field having unit norm. In real fields there is an infinity of units, forming a cyclic multiplicative group. So every unit can be expressed as a power of the generator of the group, which is called the fundamental unit. For instance, the golden mean (1+√{square root over (5)})/2 and 2143295+221064√{square root over (94)} are fundamental units in their respective fields.

[0165] A quadratic integer z that is not a unit is called a prime if a factorization z=uv is possible only when one of the two factors is a unit. For instance, z=2+√{square root over (7)} is a prime. Then one would hope that any integer can be factored in essentially only one way as product of primes. The richness and difficulty of the arithmetic of quadratic integers depends largely on the fact that unique factorization generally fails.

[0166] Quadratic residues—The values of a for which the congruence in x,
x²≡a(mod p)  (12)
is solvable are called quadratic residues of the odd prime p. The quadratic residue character is denoted by the Legendre symbol (a/p) [also written (a/p)], $\begin{array}{cc}\begin{array}{cc}\left(\frac{a}{p}\right)=1,& \mathrm{if}{x}^2\equiv a\left(\mathrm{mod}p\right)\mathrm{solvable}\mathrm{and}\gcd \left(a,p\right)=1\\ \left(\frac{a}{p}\right)=0,& \mathrm{if}\gcd \left(a,p\right)=p\\ \left(\frac{a}{p}\right)=-1,& \mathrm{if}{x}^2\equiv a\left(\mathrm{mod}p\right)\mathrm{unsolvable}\end{array}\}& \left(13\right)\end{array}$
Thus, [1+(a/p)] is the number of solutions to the equation (12) for any a.

[0167] Modules—A module is defined as a set of quantities closed under addition and subtraction. Thus, when a module contains an element ξ, it contains 0 (=ξ−ξ) as well as negatives −ξ (=0−ξ) and integral multiplies (ξ+ξ written as 2ξ, ξ+ξ+ξ written as 3ξ, etc). Capital letters M, N, D, etc., are used to denote modules. Combinations of a finite set of vectors V_i is considered
u=x₁V₁+x₂V₂+ . . . +x_sV_s  (14)
where the coefficients x_i range over all integers. The set of those u forms a module M and the vectors V_i are called a basis of the module, written
 M=[V₁, V₂, . . . , V_s]

[0168] Field—A field is a set of quantities taken from the complex numbers closed under the rational operations, namely addition, subtraction, multiplication, and division (excluding division by zero). In quadratic number theory, the considered field is taken to be the set of surds (a+b√{square root over (D)})/c for a, b, c integral, D fixed and not a perfect square, and c≠0. It can be seen that addition, subtraction, multiplication, and division of such quantities lead to quantities of the same form. This field is written symbolically as R(√{square root over (D)}) meaning that the set of surds is generated by adjoining √{square root over (D)} to the rationals. The field √{square root over (D)} is called a field over rationals. From D is extracted its (positive or negative) square-free kernel D₀, so that D=m²D₀. Note that √{square root over (D)} and √{square root over (D₀)} generate the same field. Let us define: $\begin{array}{cc}{\omega }_0=\{\begin{array}{cc}\sqrt{D_0},& \mathrm{if}{D}_0\ne \mathrm{mod}\left(4\right)\\ \frac{1+\sqrt{D_0}}{2},& \mathrm{if}{D}_0\equiv \mathrm{mod}\left(4\right)\end{array}& \left(15\right)\end{array}$
Thus, the basis of quadratic integers in R(√{square root over (D)}) is [1,ω₀].

[0169] This module is designated by the symbol D=[1,ω₀]. For example, the basis of R(√{square root over (2)}) is └1, √{square root over (2)}┘, the basis of R(√{square root over (5)}) is └1,(1+√{square root over (5)})/2┘, while R(√{square root over (8)}) has the same basis as R(√{square root over (2)}). In general, the field R(√{square root over (m²D₀)}) is independent of m, and so D and its basis.

[0170] The rational integer d, defined as $\begin{array}{cc}d=\{\begin{array}{cc}{D}_0,& \mathrm{if}{D}_0\equiv \mathrm{mod}\left(4\right)\\ 4D_0,& \mathrm{if}{D}_0\ne \mathrm{mod}\left(4\right)\end{array}& \left(16\right)\end{array}$
is called a field discriminant. All numbers sharing the same discriminant d form a field.

[0171] Integral domain—A set of quantities taken from complex numbers which is closed under addition, subtraction, and multiplication (ignoring division) is called a ring. If a ring contains the rational integers, it is called an integral domain. The quadratic integers of a fixed field R(√{square root over (D₀)}) form a domain which will be referred as D.

[0172] If the integral domain D of all quadratic integers of R(√{square root over (D)}) contains an integral domain D* which does not consist wholly of rationals, then D* is characterized by some fixed positive rational integer n as the set of integers of D which are congruent to a rational integer modulo n. The integral domain D* corresponding to n is written D_n. Thus D₁=D. Note also that D_n=[1,nω₀].

[0173] Ideals—Let us start with D_n, a quadratic integral domain. An ideal A in D_n is defined as a module in D_n with a special property that if α,βεA and ξεD_n, then α±βεA (property valid for modules), and αξεA (property distinguishing ideals). Starting with α, a fixed element of D_n, the principal ideal in D_n is defined as follows
A=(α)
as the set of αξ where ξεD_n. The ideal (1) is called the unit ideal. The sum of ideals is defined as the ideal A+B={α+β}, where αεA and βεB. The product of two ideals A and B is defined as the ideal C “generated by all products” αβ. Let us say ideal A divides ideal C in D_n (or A|C) if and only if an ideal B exists in D_n for which C=AB.

[0174] An indecomposable ideal in D_n is an ideal Q in D_n other than the unit ideal, which has no ideal in D_n as divisor other than Q and D_n. The integral domain D₁, has unique factorization into indecomposables if and only if all ideals are principals.

[0175] A prime ideal in D_n is an ideal P in D_n other than the unit ideal, with the property that for any two ideals in D_n, A and B, if P|AB, then P|A or P|B. Every prime ideal P belongs to a rational prime p determined uniquely by P|_(p).

[0176] The rational prime p factors in the quadratic field R(√{square root over (D)}), (D is a square-free integer), according to the following rules based on d, the discriminant of the field, and (d/p), the Kronecker symbol: $\begin{array}{cc}\begin{array}{ccc}\left(p\right)=\left(p\right)& \begin{array}{c}\mathrm{or}p\mathrm{is}\mathrm{inert}\left(\mathrm{does}\mathrm{not}\mathrm{factor}\right)\\ \mathrm{if}\mathrm{and}\mathrm{only}\mathrm{if}\end{array}& \left(\frac{d}{p}\right)=-1\\ \left(p\right)=P_1{P}_2& \begin{array}{c}\mathrm{or}p\mathrm{splits}\mathrm{into}\mathrm{two}\mathrm{different}\\ \mathrm{factors}\mathrm{if}\mathrm{and}\mathrm{only}\mathrm{if}\end{array}& \left(\frac{d}{p}\right)=1\\ \left(p\right)=P^2& \mathrm{or}p\mathrm{ramifies}\mathrm{if}\mathrm{and}\mathrm{only}\mathrm{if}& \left(\frac{d}{p}\right)=0\end{array}\}& \left(17\right)\end{array}$

[0177] Section 16: Dynamics and arithmetics. In this section arithmetic properties of torus automorphisms following [26] are briefly summarized. Consider the dynamics of the following map: $\begin{array}{cc}\left[\begin{array}{c}{x}_{n+1}\\ y_{n+1}\end{array}\right]=\left[\begin{array}{cc}0& 1\\ -1& k\end{array}\right]\left[\begin{array}{c}{x}_n\\ y_n\end{array}\right]\left(\mathrm{mod}N\right)& \left(18\right)\end{array}$
where x, y, k are integers and N is prime. Let us assume that 0<x₀, y₀<N and 2<k<N.

[0178] Let us consider a fixed value of the trace k, and let $\lambda =\frac{k+\sqrt{k^2-4}}{2}$
be the eigenvalue of the matrix in the equation (18). This determines a integral domain D₁ to which the eigenvalue belongs. Let d be the filed discriminant, i.e., d=D₀≡(mod 4) or d=4D₀≠(mod 4), where D₀ is the square-free kernel of k²−4.

[0179] Consider now the unit ideal in D₁ (i.e., D₁ itself): D₁=(1)=[1, ω₀], where ω₀ is given by equation (15). Multiplying this ideal by λ the same ideal is obtained, but with different basis. Its elements are integral linear combinations of the basis elements 1 and ω₀, given by the equation
λ[1,ω₀]=[m₁₁+ω₀m₂₁,m₁₂+ω₀m₂₂]
where the numbers m_ij are rearranged as a matrix: $\begin{array}{cc}{M}^{\prime }=\left[\begin{array}{cc}{m}_{11}& m_{21}\\ m_{12}& m_{22}\end{array}\right]& \left(19\right)\end{array}$

[0180] Since λ is a unit of norm +1, the matrix M′ is strictly unimodular (its determinant is equal to +1). Let us identify the point (x, y)εZ² with z=x+yω₀, i.e. z is a quadratic integer in the ideal (1). From Eq. (19) one obtains
λz=λx+λyω₀=x′+y′ω₀.

[0181] One can see that multiplication by λ corresponds to the action of the transpose M of M′ on Z²: M (x, y)=(x′, y′). In constructing the matrix M from Eq. (19), the largest solution λ of the equation λ²−kλ+1=0 has been used. This choice is not restrictive, since the smallest solution λ′, which is conjugate to λ, would just correspond to the inverse matrix M⁻¹, as easily verified. Note also that one can derive an explicit expression for M. Let k²−4=m²D₀, and D₀ be a square-free kernel. Thus for k odd, M reads: $M=\left[\begin{array}{cc}h& \left(h^2+mh-1\right)/m\\ m& h+m\end{array}\right]$
where h=(k−m)/2, while for k even, M reads: $M=\left[\begin{array}{cc}h& \left(h^2-1\right)/m\\ m& h\end{array}\right]$
where h=m/2.

[0182] The properties of the orbits generated by M are now determined for each value of k, a task which is greatly simplified by the choice of identifying Z² with the unit ideal in D₁. Then, one can determine the properties of the orbits generated by other 2 by 2 matrices with integer entries and determinant +1. It turns out, however, that the orbit structure depends to a great extent on the eigenvalue λ alone, which depends only on one parameter, the trace k.

[0183] In order to take into account the periodicity of the torus, a “two-dimensional” modular arithmetic is used, identifying quadratic integers which differ by elements of the ideal (N)=[N, Nω₀]. In other words, the points of square lattices with side N are identified. To do so a generalization of the concept of congruence is needed, since if z=x+yω₀ both x and y must be taken modulo N. Let us say that two quadratic integers v, z are congruent module an ideal A, and write v≡z(mod A), if v−z is contained in A.

[0184] The period of an orbit through the point (x, y) is given by the smallest integer T satisfying the congruence
λ^Tz≡(mod(N)), z=x+yω₀.

[0185] Note that since λ is a unit, (λ)A=A for any ideal A; thus, A is an invariant sublattice of Z². On the other hand, since one performs arithmetic modulo (N), the only invariant ideals on the torus are divisors of (N). To perform the ideal factorization of (N) (if N is an integer), its rational prime factors are first determined, N=p₁p₂ . . . p_n, where p_i are rational primes. This corresponds to the ideal factorization (N)=(p₁)(p₂) . . . (p_n). However, in the considered case N is a prime number. Below, an orbit which belongs to some ideal factor of (N) different from (1) is referred as an ideal orbit, otherwise it is referred as a free orbit. Below results proved in [26] are stated.

• • [0186] 1. If (d/N)=−1, (N) is inert. All orbits are free and have the same period T, which is a divisor of N+1. If T=(N+1)/m, then there are m(N−1) free orbits.
  • [0187] 2. If (d/N)=−1, (N) splits. All orbits have the same period T, which is divides N−1. If T=(N−1)/m, then there are m(N−1) free orbits and 2m ideal orbits.
  • [0188] 3. If (d/N)=−1, (N) ramifies. The periods of orbits are computed as follows. Let λ=(k+b√{square root over (D₀)}) (with k and b both even if D₀≠1(mod 4). Two cases are possible:
  • [0189] 3a. If k≡2(mod N), there are N−1 ideal fixed points, and N−1 free orbits of period N.
  • [0190] 3b. If k≡−2(mod N), there are (N−1)/2 ideal orbits of period 2, and (N−1)/2 free orbits of period 2N.

[0191] Section 17: Proof of the theorem 2. In this section a proof of the theorem 2 is given. Consider the following matrix: $C=\left[\begin{array}{cc}0& 1\\ -1& 2g\end{array}\right]$
Its largest eigenvalue is λ=g+√{square root over (g²−1)}. Let g²−1=m²D₀, where D₀ is square free kernel. Let us define an integer d as follows $d=\{\begin{array}{c}{D}_0,\mathrm{if}{D}_0\equiv \left(\mathrm{mod}4\right),\\ 4D_0,\mathrm{if}{D}_0\ne \left(\mathrm{mod}4\right).\end{array}$

[0192] The proof of the theorem 2 follows directly from the following theorem:

[0193] Let N be an odd prime and gεZ such that 0≦g<N. Let T be the period of the sequence T_n(g)(mod N), for n=0,1,2, . . . :

• • [0194] (i) if x²≡d(mod N) is solvable, then T is a divisor of N−1; otherwise,
  • [0195] (ii) if x²≡d(mod N) is unsolvable, then T is a divisor of N+1.

[0196] The proof of this theorem, however, follows from the results of the previous Section 16 if g≧2. Only the cases g=0 and g=1 must be considered. As mentioned in the Section 11 the periods of the trajectories starting from initial points g=0 and g=1 are always 4 and 1, respectively. Thus, for all odd primes N, the period of the sequence T_n(g)(mod N), is a divisor of N²−1.

• [0197] [1] L. M. Pecora and T. L. Carroll, “Synchronization in chaotic system”, Phys. Rev. Lett., Vol. 64, pp. 821-824, (1990).
• [0198] [2] G. Kolumban, M. P. Kennedy, G. Kis, Z. Jako, “FM-DCSK: a novel method for chaotic communications”, ISCAS '98. Proceedings of the 1998 IEEE International Symposium on Circuits and Systems, Volume: 4, Page(s): 477-480, (1998).
• [0199] [3] M. Sushchik, N. Rulkov, L. Larson, L. Tsimring, H. Abarbanel, K. Yao, A. Volkovskii, “Chaotic pulse position modulation: a robust method of communicating with chaos”, IEEE Communications Letters, Volume: 4 Issue: 4, Page(s): 128-130, (2000).
• [0200] [4] L. Kocarev, “Chaos-Based Cryptography: a Brief Overview,” IEEE Circuits and Systems Magazine, Volume 1, Issue 3, pp. 6-21, (2001).
• [0201] [5] F. Dachselt and W. Schwarz, “Chaos and cryptography”, lEEE Transactions on Circuits and Systems I: Fundamental Theory and Applications, Volume: 48 Issue: 12, pp: 1498-1509, (2001).
• [0202] [6] M. P. Kennedy, G. Kolumban, G. Kis, Z. Jako, “Performance evaluation of FM-DCSK modulation in multipath environments”, IEEE Transactions on Circuits and Systems I: Fundamental Theory and Applications, Volume: 47 Issue: 12, Page(s): 1702-1711, (2000).
• [0203] [7] N. F. Rulkov, M. M. Sushchik, L. S. Tsimring, and A. R. Volkovskii, “Digital Communication Uisng Chaotic Pulse Position Modulation”, IEEE Transactions on Circuits and Systems I: Fundamental Theory and Applications, Volume: 48 Issue: 12, Page(s): 1436-1444, (2001).
• [0204] [8] A. Menezes, P. van Oorschot, and S. Vanstone, Handbook of Applied Cryptography, CRC Press, 1997.
• [0205] [9] F. Pichler and J. Scharinger, “Finite dimensional generalized Baker dynamical systems for cryptographic applications”, Lect. Notes in Comput. Sci., vol. 1030, pp. 465-476, (1996).
• [0206] [10] J. Fridrich, “Symmetric ciphers based on two-dimensional chaotic maps”, Int. J. Bifurcation and Chaos, Volume 8, Issue 6, Pages: 1259-1284, (1998).
• [0207] [11] N. Masuda and K. Aihara, “Cryptosystems with discretized chaotic maps”, IEEE Transactions on Circuits and Systems I: Fundamental Theory and Applications, Volume: 49 Issue: 1, Pages: 28-40, (2002).
• [0208] [12] R. A. J. Matthews, “On the Derivation of a ‘Chaotic’ Encryption Algorithm”, Cryptologia, vol. 13, pp. 29-42, (1989); D. D. Wheeler, “Problems with Chaotic Cryptosystems”, Cryptologia, vol. 13, pp. 243-250, (1989); D. D. Wheeler and R. A. J. Matthews, “Supercomputer Investigations of a Chaotic Encryption Algorithm”, Cryptologia, vol. 15, no. 2, pp. 140-152, (1991).
• [0209] [13] T. Kohda and A. Tsuneda, “Statistics of Chaotic Binary Sequences”, IEEE Transactions on Information Theory, Volume: 43, Pages: 104-112, (1997).
• [0210] [14] C. S. Petrie and J. A. Connelly, “A noise-based IC random number generator for applications in cryptography”, IEEE Transactions on Circuits and Systems I: Fundamental Theory and Applications, Volume: 47 Issue: 5, Page(s): 615-621, (2000).
• [0211] [15] A. Gerosa, R. Bernardini, and S. Pietri, “A fully integrated chaotic system for the generation of truly random numbers”, IEEE Transactions on Circuits and Systems I: Fundamental Theory and Applications, Volume: 49 Issue: 7, Page(s): 993-1000, (2002).
• [0212] [16] L. Kocarev, T. Stojanovski, G. G. Rizzotto, F. Italia and D. Porto, “Method for generating a random number sequence and a relative random bit generator,” European Patent Application No. 01830764.5-1243 (EP 1 320 026), Date of filing: 13 Dec. 2001.
• [0213] [17] L. Kocarev, G. Jakimoski, G. G. Rizzotto, and P. Amato, “Chaos-based block encryption,” European Patent Application No. 01130846.7-1525 (EP 1 326 363), Date of filing: 27 Dec. 2001.
• [0214] [18] L. Kocarev, P. Amato, and G. G. Rizzotto, “Method of generating a chaos-based pseudo-random sequence and a hardware generator of chaos-based pseudo random bit sequences” European Patent Application No. 02425689.3; Date of filing: 12 Nov. 2002.
• [0215] [19] L. Kocarev, Z. Tasev, P. Amato, and G. G. Rizzotto, “Encryption process employing chaotic maps and digital signal process,” European Patent Application No. 03425219.7-1237; Date of filing: 7 Apr. 2003.
• [0216] [20] L. Kocarev and Z. Tasev, “Public-key encryption based on Chebyshev maps”, 2003 IEEE International Symposium on Circuits and Systems, May 25-May 28, 2003, Bangkok, Thailand, ISCAS 2003, accepted for publication.
• [0217] [21] L. Kocarev, Z. Tasev, and J. Makraduli, “Public-Key Encryption and Digital-Signature Schemes Using Chaotic Maps”, 16th European Conference on Circuits Theory and Design, Sep. 1-Sep. 4, 2003, Krakow, Poland, ECCTD 2003, accepted for publication. [22] C. E. Shannon, “Communication theory of secrecy systems”, Bell Systems Technical Journal, vol. 28, pp. 656-715, (1949).
• [0218] [23] C. E. Shannon, Bell Systems Technical Journal 27, 379 (1948); 27, 623 (1948).
• [0219] [24] W. Diffie and M. E. Hellman, “New directions in cryptography”, IEEE Trans. of Information Theory, Volume 22, pp. 644-654, (1976).
• [0220] [25] T. J. Rivlin, Chebyshev polynomials, John Wiley and Sons, Inc., New York, 1990.
• [0221] [26] I. Percival and F. Vivaldi, “Arithmetical properties of strongly chaotic motions”, Physica D, vol. 25, no. 1-3, pp. 105-130, (1987).
• [0222] [27] H. Cohn, A second course in number theory, John Wiley & Sons, Inc., New York, 1962.
• [0223] [28] H. Hasse, Number theory, Springer-Verlag, Berlin, 2002.
• [0224] [29] D. E. Knuth, The Art of Computer Programming, Reading, Mass.: Addison Wesley, 1998, vol. 2.
• [0225] [30] http://java.sun.com.
• [0226] [31] www.swox.com/gmp/.
• [0227] [32] T. Kohda and H. Fujisaki, “Jacobian elliptic Chebyshev rational maps”, Physica D 148, 242 (2001).
• [0228] [33] R. L. Devaney, “An Introduction to Chaotic Dynamical Systems”, Benjamin/Cummings, New York, 1984.