Digital signatures on a smartcard

0001The present invention relates to methods and apparatus for generating digital signatures.

0002It has become widely accepted to conduct transactions, such as financial transactions or exchange of documents, electronically. In order to verify the transaction, it is also well-known to "sign" the transaction digitally so that the authenticity of the transaction can be verified. The signature is performed according to a protocol that utilizes the message, ie. the transaction, and a secret key associated with the party. The recipient can verify the signature using a public key of the signing party to recover the message and compare it with the transmitted message. Any attempt to tamper with the message or to use a key other than that of the signing party will result in an incompatibility between the sent message and that recovered from the signature or will fail to identify the party correctly and thereby lead to rejection of the transaction.

0003The signature must be performed such that the signing party's secret key cannot be determined. To avoid the complexity of distributing secret keys, it is convenient to utilize a public key encryption scheme in the generation of the signature. Such capabilities are available where the transaction is conducted between parties having access to relatively large computing resources but it is equally important to facilitate such transactions at an individual level where more limited computing resources are available.

0004<nplcit id="ncit0001" npl-type="b"><text>Advances In Cryptology. Santa Barabara, Aug. 11-15, 1991, Proceedings Of The Conference On Theory And Applications Of Cryptographic Techniques (cryto), Berlin, Springer, De (1991), 279-287</text></nplcit> discloses Frobenius theorem generally. Automated teller machines (ATMs) and credit cards are widely used for personal transactions and as their use expands, so the need to verify such transactions increases. Transaction cards, i.e. credit/debit cards or pass cards are now available with limited computing capacity (so-called "Smart Cards") but these do not have sufficient computing capacity to implement existing digital signature protocols in a commercially viable manner.

0005As noted above, in order to generate a digital signature, it is necessary to utilize a public key encryption scheme. Most public key schemes are based on the Diffie Hellman Public key protocol and a particularly popular implementation is that known as DSS. The DSS scheme utilizes the set of integers Zp where p is a large prime. For adequate security, p must be in the order of 512 bits although the resultant signature may be reduced mod q, where q divides p-1, and may be in the order of 160 bits.

0006The DSS protocol provides a signature composed of two components r, s. The protocol requires the selection of a secret random integer k referred to as the session key from the set of integers (0, 1, 2,...q-1), i.e k ∈ { 0, 1, 2, ... q-1} .

0007The component r is then computed such that <maths id="math0001" num=""><math display="block"><mi mathvariant="normal">r</mi><mo>=</mo><mfenced open="{" close="}" separators=""><msup><mi mathvariant="normal">β</mi><mi mathvariant="normal">k</mi></msup><mspace width="1em" /><mi>mod p</mi></mfenced><mspace width="1em" /><mi>mod q</mi></math><img file="EP0807908B1_D0001.tif" /></maths> where β is a generator of q.

0008The component s is computed as <maths id="math0002" num=""><math display="block"><mi mathvariant="normal">s</mi><mo>=</mo><mfenced open="[" close="]" separators=""><msup><mi mathvariant="normal">k</mi><mrow><mo>-</mo><mn mathvariant="normal">1</mn></mrow></msup><mfenced separators=""><mi mathvariant="normal">h</mi><mfenced><mi mathvariant="normal">m</mi></mfenced></mfenced><mo>+</mo><mi>ar</mi></mfenced><mspace width="1em" /><mi>mod q</mi></math><img file="EP0807908B1_D0002.tif" /></maths><ul id="ul0001" list-style="none" compact="compact"><li>where m is the message to be transmitted,</li><li>h(m) is a hash of that message, and</li><li>a is the private key of the user.</li></ul>

0009The signature associated with the message is then s,r which may be used to verify the origin of the message from the public key of the user.

0010The value β^k is computationally difficult for the DSS implementation as the exponentiation requires multiple multiplications mod p. This is beyond the capabilities of a "Smart Card" in a commercially acceptable time. Although the computation could be completed on the associated ATM, this would require the disclosure of the session key k to the ATM and therefore render the private key, a, vulnerable.

0011It has been proposed to precompute β^k and store sets of values of r and k on the card. The generation of the signature then only requires two 160 bit multiplications and signing can be completed within ½ second for typical applications. However, the number of sets of values stored limits the number of uses of the card before either reloading or replacement is required. A problem that exists therefore is how to generate sufficient sets of values within the storage and/or computing capacity of the card.

0012One possibility is to use a smaller value of p but with the DSS scheme this will jeopardize the security of the transaction.

0013An alternative encryption scheme that provides enhanced security at relatively small modulus is that utilizing elliptic curves in the finite field 2^m. A value of m in the order of 155 provides security comparable to a 512 bit modulus for DSS and therefore offers significant benefits in implementation.

0014Diffie Hellman Public Key encryption utilizes the properties of discrete logs so that even if a generator β and the exponentiation β^k is known, the value of k cannot be determined. A similar property exists with elliptic curves where the addition of two points on a curve produces a third point on the curve. Similarly, multiplying any point on the curve by an integer k produces a further point on the curve. However, knowing the starting point and the end point does not reveal the value of the integer 'k' which may then be used as a session key for encryption. The value kP, where P is an initial known point, is therefore equivalent to the exponentiation β^k.

0015In order to perform a digital signature on an elliptic curve, it is necessary to have available the session key k and a value of kP referred to as a "session pair". Each signature utilizes a different session pair k and kP and although the representation of k and kP is relatively small compared with DSS implementations, the practical limits for "Smart Cards" are in the order of 32 signatures. This is not sufficient for commercial purposes.

0016One solution for both DSS and elliptic curve implementations is to store pairs of signing elements k, kP and combine stored pairs to produce a new session pair. For an elliptic curve application, this would yield a possible 500 session pairs from an initial group of 32 stored signing elements. The possibilities would be more limited when using DSS because of the smaller group of signing elements that could be stored.

0017In order to compute a new session pair, k and kP, from a pair of stored signing elements, it is necessary to add the values of k, e.g. k, + k₂ → k and the values of k₁P and k₂P to give a new value kP. In an elliptic curve, the addition of two points to provide a third point is performed according to set formula such that the addition of a point k1P having coordinates (x,y) and a point k1P having coordinates (x₂y₂) provides a point k3P whose x coordinate x3 is given by: <maths id="math0003" num=""><math display="block"><msub><mi mathvariant="normal">x</mi><mn mathvariant="normal">3</mn></msub><mo>=</mo><mfrac><mrow><msub><mi mathvariant="normal">y</mi><mn mathvariant="normal">1</mn></msub><mo>⊕</mo><msup><msub><mi mathvariant="normal">y</mi><mn mathvariant="normal">2</mn></msub><mn mathvariant="normal">2</mn></msup></mrow><mrow><msub><mi mathvariant="normal">x</mi><mn mathvariant="normal">1</mn></msub><mo>⊕</mo><msub><mi mathvariant="normal">x</mi><mn mathvariant="normal">2</mn></msub></mrow></mfrac><mspace width="1em" /><mo>⊕</mo><mspace width="1em" /><mfrac><mrow><msub><mi mathvariant="normal">y</mi><mn mathvariant="normal">1</mn></msub><mo>⊕</mo><msub><mi mathvariant="normal">y</mi><mn mathvariant="normal">2</mn></msub></mrow><mrow><msub><mi mathvariant="normal">x</mi><mn mathvariant="normal">1</mn></msub><mo>⊕</mo><msub><mi mathvariant="normal">x</mi><mn mathvariant="normal">2</mn></msub></mrow></mfrac><mo>⊕</mo><mspace width="1em" /><msub><mi mathvariant="normal">x</mi><mn mathvariant="normal">1</mn></msub><mo>⊕</mo><msub><mi mathvariant="normal">x</mi><mn mathvariant="normal">2</mn></msub><mn mathvariant="normal">.</mn></math><img file="EP0807908B1_D0003.tif" /></maths>

0018This computation may be significantly simplified using the normal basis representation in a field F2^m, as set out more fully in our PCT Application Serial No. <patcit id="pcit0001" dnum="CA1995CA00452W"><text>PCT/1995CA/00452</text></patcit>. However, even using such advantageous techniques, it is still necessary to utilize a finite field multiplier and provide sufficient space for code to perform the computation. This is not feasible within the practical limits of available "Smart" cards.

0019As noted above, the ATM used in association with the card has sufficient computing power to perform the computation but the transfer of the coordinates of k₁P and k₂P from the card to the terminal would jeopardize the integrity of subsequent digital signatures as two of the stored signing elements would be known.

0020It is therefore an object of the present invention to obviate or mitigate the above disadvantages and facilitate the preparation of additional pairs of values from a previously stored set.

0021In general terms, one aspect of the present invention proposes to compute on one computing device an initial step in the computation of a coordinate of a point derived from a pair of points to inhibit recognition of the individual components, transfer such information to another computing device remote from said one device, perform at least such additional steps in said derivation at such other device to permit the completion of the derivation at said one device and transfer the result thereof to said one computing device.

0022Preferably, the initial step involves a simple field operation on the two sets of coordinates which provides information required in the subsequent steps of the derivation.

0023Preferably also the additional steps performed at the other device complete the derivation.

0024In a preferred embodiment, the initial step involves the addition of the x coordinates and the addition y coordinates to provide the terms (x₁⊕x₂) and (y₁⊕y₂).

0025The addition of the coordinates is an XOR operation that can readily be performed on the card and the results provided to the terminal.

0026In this manner, the coordinates (x,y) representing kP in a stored signing element are not disclosed as insufficient information is provided even with subsequent uses of the card. Accordingly, the x coordinate of up to 500 signatures can be generated from an initial set of 32 stored signing elements.

0027The new value of k can be computed on the card and to avoid computing the inverse k^-1, alternative known masking techniques can be utilized.

0028A further aspect of the present invention provides a method of generating additional sets of points from the initial set that may be used individually as a new value of kP or in combination to generate still further values of kP.

0029According to this aspect of the invention, the curve is an anomalous curve and the Frobenius Operator is applied to at least one of the coordinates representing a point in the initial set to provide a coordinate of a further point on the elliptic curve. The Frobenius Operator ∅ provides that for a point (x₁,y₁) on an anomalous curve, then ∅ (x₁,y₁) is a point (x₁²,y₁²) that also lies on the curve. In general, ∅¹(x₁y₁) is a point x^{2<sup2>1</sup2>}, y^{2<sup2>1</sup2>} that also lies on the curve. For a curve over the field 2^m, there are m Frobenius Operators so for each value of kP stored in the initial set, m values of kP may be generated, referred to as "derived" values. The new value of k associated with each point can be derived from the initial relationship between P and ∅P and the initial value of k.

0030For a practical implementation where 32 pairs of signing elements are initially retained on the card and the curve is over the field 2¹⁵⁵, utilizing the Frobenius Operator provides in the order of 4960 possible derived values and by combining pairs of such derived values as above in the order of 10⁷ values of kP can be obtained from the initial 32 stored signing elements and the corresponding values of k obtained to provide 10⁷ session pairs.

0031Preferably, the stored values of kP are in a normal basis representation. The application Frobenius Operator then simply requires an "i" fold cyclic shift to obtain the value for an ∅' operation.

0032According to a further aspect of the invention, there is provided a method of generating signature components for use in a digital signature scheme, said signature components including private information and a public key derived from said private information, said method comprising the steps of storing private information and related public key as an element in a set of such information, cycling in a deterministic but unpredictable fashion through said set to select at least one element of said set without repetition and utilizing said one element to derive a signature component in said digital signature scheme.

0033Embodiments of the invention will now be described by way of example only with reference to the accompanying drawings, in which <ul id="ul0002" list-style="none" compact="compact"><li><figref idref="f0001">Figure 1</figref> is a schematic representation of a programmable credit card;</li><li><figref idref="f0001">Figure 2</figref> is a schematic representation of a transaction performed between the card and network;</li><li><figref idref="f0002">Figure 3</figref> is a schematic representation of the derivation of a session pair from a pair of stored signing elements;</li><li><figref idref="f0003">Figure 4</figref> is a schematic representation of one step in the transmission of information shown in <figref idref="f0001">Figure 2</figref>;</li><li><figref idref="f0004">Figure 5</figref> is a schematic representation of a preferred implementation of the derivation of a session pair from two pairs of stored values; and</li><li><figref idref="f0005">Figure 6</figref> is a schematic representation of a selection unit shown in <figref idref="f0001">Figure 1</figref>.</li></ul>

The System

0034Referring therefore to <figref idref="f0001">Figure 1</figref>, a programmable credit card 10 (referred to as a 'SMART' card) has an integrated circuit 12 embedded within the body of card 10.

0035The integrated circuit includes a logic array 14, an addressable memory 16 and a communication bus 18. The memory 16 includes a RAM section 20 to store information, a pair of cyclic shift registers 22 for temporary storage of information and programming code 24 for control of the logic array 14 and communication bus 18. The array 14 includes an arithmetic unit 26 to provide modular arithmetic operation, e.g. additional and multiplication, and a selection unit 28 controlled by the programming code 24. It will be appreciated that the description of the card 10 is a schematic and restricted to that necessary for explanation of the preferred embodiment of the invention.

0036The card 10 is used in conjunction with a terminal 30, for example an automated teller machine (ATM), that is connected to a network to allow financial transactions to be conducted. The terminal 30 includes a keypad 32 to select options and tasks and has computing capabilities to perform the necessary functions in conjunction with the card 10.

0037Access to the terminal 30 is obtained by inserting card 10 into a reader 34 and entering a pass code in a conventional manner. The pass code is verified with the card 10 through communication bus 18 and the terminal 30 activated. The keypad 32 is used to select a transaction, for example a transfer of funds, between accounts and generate a message through the network to give effect to the transactions, and card 10 is used to sign that transaction to indicate its authenticity. The signature and message are transmitted over the network to the intended recipient and upon receipt and verification, the transaction is completed.

The Card

0038The RAM section 20 of memory 16 includes digital data string representing a private key, a, which remains secret with the owner of the card and a corresponding public key Q = aP where P is the publicly known initial point on the selected curve. The RAM section 20 also includes a predetermined set of coordinates of points, kP, on an elliptic curve that has been preselected for use in a public key encryption scheme. It is preferred that the curve is over a finite field 2^m, conveniently, and by way of example only, 2¹⁵⁵, and that the points kP are represented in normal basis representation. The selected curve should be an anomalous curve, e.g. a curve that satisfies y² + xy = x³ + 1, and has an order, e. Each point kP has an x coordinate and a y coordinate and is thus represented as two 155 digital data strings that are stored in the RAM 20. By way of example, it will be assumed that the RAM 20 contains 32 such points identified generically as kP and individually as k₀P, k₁P...k₃₁P. Similarly, their coordinates (x,y) will be individually designated x₀y₀...X₃₁Y₃₁.

0039The points kP are precomputed from the chosen parameters of the curve and the coordinates of an originating point P. The k-fold addition of point P will provide a further point kP on the curve, represented by its coordinates (x,y) and the value of k cannot be determined even if the coordinates of points P and kP are known.

0040RAM 20 therefore contains the values of k associated with the respective points kP so that a set of stored signing elements k,kP is available for use in the signing of the transaction.

Signing

0041To sign a message m generated by the transaction, one session pair k_j; k_jP is required and may be obtained from RAM 20 as set out more fully below. Assuming that values k_j, k_jP have been obtained, the signing protocol requires a signature r,s) where <dl id="dl0001" compact="compact"><dt>r</dt><dd>is the data string representing the x-coordinate, x_j reduced mod q (q is a preselected publicly known divisor of e, the order of the curve, i.e. q/e_x);and</dd><dt>s</dt><dd>= [k^-1(h(m)) + ar] mod q where h(m) is a q-bit hash of the message m generated by the transaction.</dd></dl>

0042In this signature, even though r is known, s contains the secret k and the private key, a, and so inhibits the extraction of either.

0043The generation of s requires the inversion of the value k and since k is itself to be derived from the stored set of values of k, it is impractical to store corresponding inverted values of possible k's. Accordingly, a known masking technique is used to generate components r, s¹ and u of a signature. This is done by selecting an integer, c, and computing a value u = ck. The value s¹ = c(h(m) + ar) mod q.

0044The signature value s can then be obtained by the recipient computing s¹u^-1 = k^-1 [h(m) + ar].

0045The signature (r,s¹,u) can be computed on the card 10 and forwarded by bus 18 to the terminal 30 for attachment to the message m.

Generation of Session Pair

0046As noted above, in order to generate the signature (r,s), it is necessary to have for session pair k and kP. Security dictates that each session pair is only used once and it is assumed that the number of signing elements stored in RAM 20 is insufficient for commercial application.

0047In the preferred embodiment, two techniques are used to generate additional session pairs to the stored signing elements. It will be appreciated that each technique may be used individually although the combination of the two is preferred.

(i) Frobenius Operator

0048The first technique involves the use of the Frobenius Operator to derive additional session pairs from the stored signing elements and is shown in <figref idref="f0002">Figure 3</figref>. The Frobenius Operator denoted Ø operates on a point P having coordinates (x,y) on an anomalous elliptic curve in the finite field 2^m such that ∅ⁱP = (x^{2<sup2>1</sup2>},y^{2<sup2>1</sup2>}). Moreover, the point ∅ⁱP is also on the curve. In the field 2¹⁵⁵, there are 155 Frobenius Operators so each point kP stored in memory 20 may generate 155 points on the curve by application of the Frobenius Operators. Thus, for the 32 values of kP stored, there are 4960 possible values of kP available by application of the Frobenius Operator.

0049To derive the value of ∅¹P, it is simply necessary to load the x and y coordinates of a point kP into respective shift registers 22 and perform an i-fold cyclic shift. Because the coordinates (x,y) have a normal basis representation, a cyclic shift in the register 22 will perform a squaring operation, and an i-fold cyclic shift will raise the value to the power 2ⁱ. Therefore, after the application of i clock cycles, the registers 22 contain the coordinates of ∅¹(kP) which is a point on the curve and may be used in the signing protocol. The 155 possible values of the coordinates (x,y) of ∅¹(kP) may be obtained by simple cyclic shifting. The representations in the registers 22 may then be used to obtain r.

0050Where the use of Frobenius Operator provides sufficient values for commercial use, only one coordinate is needed to compute the value of r and so only a single shift register is needed. However, as will be described below, further session pairs can be derived if both the coordinates are known and so a pair of registers is provided.

0051For each value of ∅¹(kP), it is necessary to obtain the corresponding value of k ∅(P) = λP. λ is a constant that may be evaluated ahead of time and the values of its first m powers, λⁱ computed. The m values are stored in RAM 20.

0052In general, ∅ⁱ(kP) → λⁱkP so the value of k associated with Øⁱ(kP) is λⁱk. Since k is stored for each value of kP in RAM 20 and λⁱ is also stored, the new value of k, i.e. λⁱk, can be computed using the arithmetic unit 26.

0053As an alternative, to facilitate efficient computation of λⁱ and avoid excessive storage, it is possible to precompute specific powers of λ and store them in RAM 20. Because m is 155 in the specific example, the possible values of i can be represented as an 8-bit binary word. The values of λ² → λ^{2<sup2>7</sup2>} are thus stored in RAM 20 and the value of λ represented in binary. The presto red values of λ^{2<sup2>i</sup2>} are then retrieved as necessary and multiplied mod e by arithmetic unit 26 to provide the value of λⁱ. This is then multiplied by k to obtain the new value associated with ∅¹(kP).

0054It will be seen therefore that new session pairs k, kP may be derived simply and efficiently from the stored signing elements of the initial set. These session pairs may be computed in real time, thereby obviating the need to increase storage capacity and their computation utilizes simple arithmetic operations that may be implemented in arithmetic unit 26.

(ii) Combining Pairs

0055A further technique, illustrated schematically in <figref idref="f0003">Figure 4</figref>, to increase the number of session pairs of k and kP available, and thereby increase the number of signatures available from a card, is to combine pairs of stored signing elements to produce a new derived value. The addition of two points k₁P and k₂P will produce a third point k₃P that also lies on the curve and may therefore be used for signatures.

0056The addition of two points having coordinates (x₁,y₁)(x₂y₂) respectively on a curve produces a new point having an x coordinate x₃ where <maths id="math0004" num=""><math display="block"><msub><mi mathvariant="normal">x</mi><mn mathvariant="normal">3</mn></msub><mo>=</mo><mfrac><mrow><msub><mi mathvariant="normal">y</mi><mn mathvariant="normal">1</mn></msub><mo>⊕</mo><msup><msub><mi mathvariant="normal">y</mi><mn mathvariant="normal">2</mn></msub><mn mathvariant="normal">2</mn></msup></mrow><mrow><msub><mi mathvariant="normal">x</mi><mn mathvariant="normal">1</mn></msub><mo>⊕</mo><msub><mi mathvariant="normal">x</mi><mn mathvariant="normal">2</mn></msub></mrow></mfrac><mo>⊕</mo><mfrac><mrow><msub><mi mathvariant="normal">y</mi><mn mathvariant="normal">1</mn></msub><mo>⊕</mo><msub><mi mathvariant="normal">y</mi><mn mathvariant="normal">2</mn></msub></mrow><mrow><msub><mi mathvariant="normal">x</mi><mn mathvariant="normal">1</mn></msub><mo>⊕</mo><msub><mi mathvariant="normal">x</mi><mn mathvariant="normal">2</mn></msub></mrow></mfrac><mo>⊕</mo><msub><mi mathvariant="normal">x</mi><mn mathvariant="normal">2</mn></msub></math><img file="EP0807908B1_D0004.tif" /></maths>

0057In the finite field 2m, y1⊕y2 and x1⊕x2 is an XOR field operation that may be performed simply in logic array 16. Thus the respective values of x₁ ,x₂ and y₁ ,y₂ are placed in respective ones of registers 22 and XOR'd. The resultant data string is then passed over communication bus 16 to the terminal 30. The terminal 30 has sufficient computing capacity to perform the inversion, multiplication and summation to produce the value of x₃. This is then returned to register 22 for signature. The potential disclosure of x₃ does not jeopardize the security of the signature as the relevant portion is disclosed in the transmission of r.

0058The value of k₁+k₂ is obtained from the arithmetic unit 26 within logic array 16 to provide a value of k₃ and hence a new session pair k₃, k₃P is available for signature.

0059It will be appreciated that the value for y₃ has not been computed as the signing value r is derived from x₃ rather than both coordinates.

0060It will be noted that the values of x₁ and x₂ or y₁ and y₂ are not transmitted to terminal 30 and provided a different pair of points is used for each signature, then the values of the coordinates remains undisclosed.

0061At the same time, the arithmetic functions performed on the card are relatively simple and those computationally more difficult are performed on the terminal 30.

Preferred Implementation of Generating Session Pairs

0062The above technique may of course be used with pairs selected directly from the stored signing elements or with the derived values obtained using the Frobenius Operator as described above. Alternatively, the Frobenius Operator could be applied to the value of kP obtained from combining pairs of the stored signing elements to provide m possible values of each derived value. This may be accomplished by cyclically shifting the coordinates of the derived value in respective registers 22 as described above.

0063To ensure security and avoid duplication of session pairs, it is preferred that only one of the stored signing elements should have the Frobenius Operator applied, as in the preferred embodiment illustrated in <figref idref="f0004">Figure 5</figref>.

0064In this arrangement, the coordinates x₁,y₁ of one of the stored signing elements is applied to the registers 22 (step I in <figref idref="f0004">Figure 5</figref>) and cyclically shifted i times to provide ∅¹ k₁P (step II).

0065The respective coordinates, x_∅1,y_∅1, are XOR'd with the coordinates from another of the stored values k₂P (step III) and the summed coordinates transmitted to ATM 30 (step IV) for computation of the coordinate x₃ (step V). This is retransmitted to the card 10 (step VI) for computation of the value r.

0066The value of k₁ is retrieved and processed by arithmetic unit 26 (step VII) to provide λⁱk and added to k₂ to provide the new value k₃ for generation of signature component s as described above. In this embodiment, from an original set of 32 stored signing elements stored on card 10, it is possible to generate in the order of 10⁷ session pairs. In practice, a limit of 10⁶ is realistic.

Selection of Pairs Stored Signing Elements

0067The above procedure requires a pair of stored signing elements to be used to generate each session pair. In order to preserve the integrity of the system, the same set cannot be used more than once and the pairs of stored values constituting the set must not be selected in a predictable manner.

0068This selection function is performed by the selection unit 28 whose operation is shown schematically in <figref idref="f0005">Figure 6</figref>.

0069Selection unit 28 includes a set of counters 40,42,44 whose outputs address respective look up tables 46,48,50. The look up tables 46,48,50 map the successive outputs of the counters to pseudo random output values to provide unpredictability for the selection stored signing elements.

0070The 32 stored values of k and kP are assigned nominal designations as elements in a set 52 ranging from -15 to +15 with one designated ∞. To ensure that all available combinations of stored values are used without repetition, the nominal designations are grouped in 16 pairs in an ordered array 54 such that the difference (mod 31) in the assigned values of a pair uses all the numbers from 1 to 30. ∞ is grouped with 0. This array provides a first row of a notional matrix.

0071Successive rows 54a,b,c,etc. of the notional matrix are developed by adding 1 to each assigned designation of the preceding row until 15 rows are developed. In this way a matrix is developed without repetition of the designations in each cell. By convention ∞ + 1 = ∞.

0072Counter 42 will have a full count after 15 increments and counter 40 will have a full count after 14 increments. Provided the full count values of counters 40,42 are relatively prime and the possible values of the counter 44 to select Frobenius Operator are relatively large, the output of counters 40,42,44 are mapped through the tables 46,48,50 respectively to provide values for row and column of the notional matrix and the order i of the Frobenius Operator to be applied. The mapping provided a non-sequential selection from the array 54.

0073The output of table 48 selects a column of the array 54 from which a designation associated with a starting pair can be ascertained. In the example of <figref idref="f0005">Figure 6</figref>, the output of counter 42 is mapped by table 48 to provide an output of 3, indicating that column 3 of array 54 should be selected. Similarly, the output of counter 40 is mapped through table 46 to provide a count of 3 indicating that values in row 3 of the matrix should be used.

0074The assigned designations for a particular row are then obtained by adding the row value to the values of the starting pair. This gives a new pair of assigned designations that indicate the locations of elements in set 52. The signing elements are then retrieved from the set 52.

0075One of those pairs of signing elements is then output to a shift register 22 and operated upon by the designated Frobenius Operator ∅. The value of the Frobenius Operation is obtained from the output of table 50 which maps counter 44. The value obtained from table 5 sets the shift clock associated with register 22 so that the contents of the register 22 are cyclically shifted to the Frobenius value ∅ indicated by the output of table 50.

0076Accordingly, a new value for kP is obtained. The associated value of k can be computed as described above with the arithmetic unit utilizing the output of table 50 to determine the new value of λ. Accordingly, a derived value is obtained.

0077The derived value and signing element are then combined as described at (ii) above with respect to <figref idref="f0003">Figures 4</figref> and <figref idref="f0004">5</figref> to provide a new session pair k, kP for use in the signing process.

0078The use of the counters 40,42 provides input values for the respective tables so that the array 54 is accessed in a deterministic but unpredictable fashion. The grouping of the pairs in the array 54 ensures there is no repetition in the selected elements to maintain the integrity of the signature scheme.

0079Counter 44 operates upon one of the selected pairs to modify it so that a different pair of values is presented for combination on each use, even though multiple access may be made to the array 54.

0080The counters 40,42,44 may also be utilized to limit the use of the Smart Card if desired so that a forced expiry will occur after a certain number of uses. Given the large number of possible signatures, this facility may be desirable.

0081Alternative structures to the look up tables 46,48,50 may be utilized, such as a linear feedback shift register, to achieve a mapped output if preferred.

0082In summary, therefore, pairs of signing elements from an initial set of stored values can be selected in a deterministic and unpredictable manner and one of those elements operated upon by the Frobenius Operator to provide additional values for the elements. The elements may then be combined to obtain a new session pair with a portion of the computation being performed off card but without disclosing the value of the elements. Accordingly, an extended group of session pairs is available for signing from a relatively small group of stored values.