DUAL HASH ALGORITHM IMPLEMENTATION
Background of the invention The present invention relates to an efficient hardware implementation of the Secure Hash Algorithm (SHA-1) and Message Digest Algorithm (MD5).
Description of the Prior Art
Hash algorithms and message digests are frequently used in applications such as digital signatures, where it is desirable to verify the authenticity of a document or file.
Techniques for producing message digests are beneficial as they reduce the amount of data processing needed to a manageable and consistent level.
The Secure Hash algorithm (SHA-1 ) is specified in Secure Hash Standard (FIPS PUB 180-1), and is an algorithm which operates on an input data file to produce a condensed representation of that file. Specifically, an message of arbitrary length is processed to produce a message digest consisting of exactly 160 bits.
The Message Digest Algorithm (MD5), developed by Professor Ronald Rivest, has a similar function. It accepts inputs of arbitrary length and produces an output message digest consisting of exactly 128 bits.
Both algorithms may be used as a constituent part of a digital signature application. Both algorithms are computationally intensive and, when implemented in software, as is the norm in prior art systems, can take a great number of processor clock cycles to complete.
The present invention therefore aims to overcome problems with the prior art implementation of these systems, particularly in relation to speed of operation and power consumption.
Summary of the Present Invention
In a first broad form the present invention provides apparatus arranged to accept digital data as an input, and to process said data according to one of either the
Secure Hash Algorithm (SHA-1 ) or Message Digest (MD5) algorithm to produce a fixed length output word, said apparatus including: a plurality of rotational registers for storing data, one of said registers being arranged to receive the input data; and data stores for initialisation of some of said plurality of registers according to whether the SHA-1 or MD5 algorithm is used, said data stores including fixed data relating to SHA-1 and MD5 operation; and a plurality of dedicated combinatorial logic circuits arranged to perform logic operations on data stored in selected ones of said plurality of registers.
Preferably, the register arranged to receive the input data is arranged to receive said input data serially.
Preferably, the registers and combinatorial logic circuits are interconnected for communication via a pair of data busses. It is particularly preferable if the registers and combinatorial logic circuits are connected to write to a respective bus via respective tristate buffers.
Preferably, the apparatus includes a control circuit arranged to generate individually gated clock signals for each register. This results in lower power consumption as only active registers need to be clocked.
Preferably, the control circuit is further arranged to generate individual enabling signals to control the tristate buffers. The control circuit may be implemented as a dedicated state machine, or by another means such as a microcontroller.
Preferably, the rotational registers are arranged to be multiplexed prior to connection to a tristate buffer. This results in a lower bus loading.
Preferably, the combinatorial logic circuits include a copy circuit, a shift left circuit, a NOT circuit, an ADD circuit, an OR circuit, an AND circuit and an XOR circuit. Each circuit is dedicated to its particular task, avoiding redundancy.
Preferably, the apparatus is implemented as an integrated circuit, typically of the ASIC type. The apparatus may be incorporated with other apparatus, typically digital signature apparatus.
Embodiments of the present invention utilise the fact that both algorithms may be broken down into a series of individual steps. Prior art approaches to implementing the algorithms in software do not utilise any specialised hardware components, which results in a relatively slow process. However, embodiments of the invention identify, where possible, the common elements between the MD5 and SHA-1 algorithms and provide specialised hardware components to achieve the required functionality. Hardware is selected which allows for the maximum sharing of components and hence the minimum overall component count.
Embodiments of the present invention allow a relatively small number of dedicated components to be used in a circuit to efficiently calculate either MD5 or SHA-1 message digests. Since the operations involved in both algorithms are similar, the circuit can be optimised to allow components which are common to both algorithms to be provided only once. Allowing either of the algorithms to be used in calculating a message digest is advantageous as there are several digital signature systems operational which make use of one or other of the SHA-1 or MD5 algorithms. Systems utilising an embodiment of the invention benefit from increased flexibility and speed.
Brief Description of the Drawings
For a better understanding of the present invention and to understand how the same may be brought into effect, the invention will now be described by way of example only, with reference to the appended drawings in which:
Figure 1 shows a view of the architecture of the combined SHA-1 and MD5 processor; and
Figure 2 shows a view of the control circuit used to control the architecture shown in Figurel .
Detailed Description of the Preferred Embodiments
Figure 1 shows a customised architecture which is arranged to receive a data input 150, process it using the shown elements, and produce a data output 155. The hardware shown is able to perform either SHA-1 or MD5 processing on the input data, and has been optimised in order to minimise the amount of hardware needed to perform either one of the algorithms.
The circuit includes a plurality of registers for storing data. There are ten registers provided in two banks 110, 115 for storing part of the data being processed. In addition, two temporary registers 120, 135 are provided for intermediate processing and temporary storage. Also provided are two banks 125, 130 of circular shift registers W15[31 :0] - W0[31 :0]. Register W15 of bank 125 is arranged to receive the input data 150. Any data held in W15 at that time is shifted to W14; the data in W14 is shifted to W13 and so on, until the data held in W0 is lost. The outputs of banks 125 and 130 are multiplexed before being attached to the read bus 1.40 by a tristate buffer in order to reduce bus loading.
The registers are mutually interconnected for communication via a read bus 140 and a write bus 145.
The read bus 140 is connected to a range of logic circuits which provide combinatorial functions. These functions are: Copy (CP) 200, Shift Left multiple positions (SL*) 205, NOT 210, ADD 215, OR 220, AND 225, XOR 230 and Shift Left one position (SL1 ) 235. Functions 200, 205, 210 require only a single input variable and receive it directly from the read bus 140. The other functions 215, 220, 225 and 230 require two input variables and receive one from the read bus 140 and the other from a temporary register (ACCU[31 :0]) 135. Register 135 also provides the input for shift register 235.
Also connected to the read bus via a multiplexer and a tristate logic gate is a bank 160 of registers including fixed constants used in the initialisation of the circuit for either SHA-1 or MD5 mode calculations. K[t] is provided for initialisation of SHA-1 , and T[i] is provided for initialisation of MD5. In total, approximately seventy five
constants each having a length of 32 bits are required, and grouping them together in this fashion allows them to be conveniently accessed. The synthesis tool which places the gates in the finished custom device is then able to optimise the logic, resulting in a smaller gate count, and thus a smaller area of silicon os required.
Calculation of either SHA-1 or MD5 requires the use of selected ones of the provided registers and combinatorial functions. In particular, calculation of the SHA-1 algorithm uses all the registers of bank 110 and of bank 115. Calculation of MD5 requires only the use of four registers (HO - H3) of bank 110 and four registers (A-D) of bank 1 15. This allows the unused registers to be used for temporary storage if required. However, when the result of the calculation 155 is unloaded from register HO of bank 1 10, all five registers are read since they are implemented as shift registers, and this ensures that their contents are unchanged.
All devices which can output data to the read bus 140 are connected to the bus via a tristate buffer. Each buffer is individually enabled via a control signal created by the control circuit shown in Figure 2. Likewise, the combinatorial functions 200-235 which can write data onto the write bus 145 are connected to the write bus via individually controllable tristate buffers.
The group of clock signals 345 to individual registers are created from a master clock signal 340. The master clock signal is ANDed with a control signal to create a gated clock signal for the appropriate register. In this way, the energy consumption of the complete circuit is reduced as only active registers need to be clocked.
Figure 2 shows the a top level view of the control circuit 400 which generates the various control signals for the circuit of Figure 1. In particular, it generates, from a master clock signal 340, a series 345 of gated individual clock signals which are used to clock the various registers of Figure 1. It also generates individual enable signals for each of the tristate buffers shown in Figure 1. The control circuit mat take the form of a finite state machine including associated controlling circuits.
The following pseudo-code represents the steps performed in calculating a message digest according to the SHA-1 algorithm on an input data word of arbitrary length.
The high level algorithm details in broad terms the steps taken in performing a calculation according to the SHA-1 algorithm. The following more detailed code provides step by step instructions on performing the individual instructions needed to calculate the message digest. 5
SHA-1 Algorithm
// High Level Algorithm
initialize SHA-1 internal registers (H0,H1,H2 ,H3 ,H4) 10 foreach Mi, block of 512 bits of M do load Mi into data registers W[0] to [15] , start core SHA-1 end unload H0,H1,H2,H3 ,H4 15
//Detailed steps
- SHA-1 initialization 20 HO = 67452301
HI = EFCDAB89
H2 = 98BADCFE > • H3 = 10325476
H4 = C3D2E1F0 25
Core SHA-1
A=H0, B=H1, C=H2, D=H3 , E=H4
MASK=OO00OOOF for t=0 to 79 do 30 s = t and MASK; if (t>=16) [s] = SL1 ( [(s + 13) and MASK] xor [(s + 8) and MASK] xor [(s + 2) and MASK] xor [s]); end if
TEMP = SL5(A)+ ft(B,C,D)+E +W[s]+K[t] E=D, D=C, C=5L30(B) , B=A, A=TEMP end for
HO =H0+A, HI =H1+B, H2 =H2+C, H3=H3+D, H4=H4+E
// The functions SLl, SL5 and SL30 are circular left rotation
// of the 32 bit operand by 1 bit, 5 bits and 30 bit
// respectively.
// The constants Kt are defined by the following:
Kt = 5A82 7999 ( 0 <= t <= 19)
Kt = 6ED9 EBA1 (20 <= t <= 39)
Kt = 8F1B BCDC (40 <= t <= 59)
Kt = CA62 C1D6 (60 <= t <= 79)
// The functions ft(B,C,D) is defined by the following: ft (B,C,D) = (B and C) or ((not B) and D) (0 <= t <= 19) ft (B,C,D) = B xor C xor D (20 <= t <= 39) ft (B,C,D) = (B and C) or (B and'ϋ) or (C and D) (40 <= t <= 59) ft (B,C,D) = B xor C xor D (60 <= t <=79) .
The following pseudo-code represents the steps performed in calculating a message digest according to the MD5 algorithm on an input data word of arbitrary length.
MD5 Pseudo Algorithm
// Here, the four auxiliary functions that each take as input // three 32 -bit words and produce as output one 32 -bit word // are defined:
F(X,Y,Z) = (X and Y) or (not(X) and Z) G(X,Y,Z) = (X and Z) or (Y and not(Z)) H(X,Y,Z) = X xor Y xor Z
I (X , Y , Z) = Y xor (X or not ( Z ) )
// A 64-element table T [1 ... 64] constructed from the sine // function is defined. Let T[i] denote the i-th element of 5 // the table, which is equal to the integer part of 4294967296 // times abs(sin(i)), where i is in radians.
High Level Algorithm ' initialize MD5 internal registers (H0,H1,H2,H3) 10 foreach Mi, block of 5.12 bit of M do load Mi into data registers [0] to [15] • ; start core MD5 end unload HO, HI, H2, H3 15
MD5 initialization ■
HO = 67 45 23 01
HI = ef cd ab 89
. H2 = 98 ba dc fe
20 H3 = 10 32 54 76
Core MD5 ■ A=H0, B=H1, C=H2, D=H3
25 // Round 1.
// Let [abed k s i] denote the operation
// a = b + ((a + F(b,c,d) + W [k] + T [i] ) <<< s) .
// Do the following 16 operations.
30 [ABCD 0 7 1] [DABC 1 12 2] [CDAB 2 17 3] [BCDA 3 22 4]
[ABCD 4 7 5] [DABC 5 12 6] [CDAB 6 17 7] [BCDA 7 22 8]
[ABCD 8 7 9] [DABC 9 12 10] [CDAB 10 17 11] [BCDA 11 22 12]
[ABCD 12 7 13] [DABC 13 12 14] [CDAB 14 17 15] [BCDA 15 22 16]
/ / Round 2.
// Let [abed k s i] denote the operation // a = b + ((a + G(b,c,d) + [k] + T[i]) <<< s) . // Do the following 16 operations.
[ABCD 1 5 17] [DABC 6 9 18] [CDAB 11 14 19] [BCDA 0 20 20] [ABCD 5 5 21] [DABC 10 9 22] [CDAB 15 14 23] [BCDA 4 20 24] [ABCD 9 5 25] [DABC 14 9 26] [CDAB 3 14 27] [BCDA 8 20 28] [ABCD 13 5 29] [DABC 2 9 30] [CDAB 7 14 31] [BCDA 12 20 32]
// Round 3.
// Let [abed k s i] denote the operation
// a = b + ( (a + H(b,c,d) + W [k] + T[i]) <<< s) .
// Do the following 16 operations.
[ABCD 5 4 33] [DABC 8 11 34] [CDAB 11 16 35] [BCDA 14 23 36] [ABCD 1 4 37] [DABC 4 11 38] [CDAB 7 16 39] [BCDA 10 23 40] [ABCD 13 4 41] [DABC 0 11 42] [CDAB 3 16 43] [BCDA 6 23 44] [ABCD 9 4 45] [DABC 12 11 46] [CDAB 15 16 47] [BCDA 2 23 48]
// Round 4.
// Let [abed k s i] denote the operation
// a = b + ( (a + I(b,c,d) + W [k] + T[i] ) <<< s) .
// Do the following 16 operations.
[ABCD 0 6 49] [DABC 7 10 50] [CDAB 14 15 51] [BCDA 5 21 52] [ABCD 12 6 53] [DABC 3 10 54] [CDAB 10 15 55] [BCDA 1 21 56] [ABCD 8 6 57] [DABC 15 10 58] [CDAB 6 15 59] [BCDA 13 21 60] [ABCD 4 6 61] [DABC 11 10 62] [CDAB 2 15 63] [BCDA 9 21 64]
HO =H0+A, HI =H1+B, H2 =H2+C, H3=H3+D
The information below sets out the so-called atomic operations which are required to perform the different algorithm calculations. The following steps indicate the
operation number, the operation performed, and the status of the read 140 and write 145 busses. Each operation listed below takes exactly one clock cycle.
SHA-1 ALGORITHM initialization
## operation Readbus ritebus
01. A := HO HO (copy)
02. B := HI HI (copy)
03. C := H2 H2 (copy)
04. D := H3 H3 (copy)
05. E := H4 H4 (copy)
0<=t<=15
## operation Readbus Writebus
01. ACCU := B B (copy)
02. TMP := ACCU and C C (and)
03.' ACCU := NOT B B (not)
04. ACCU := ACCU and D D (and)
05. ACCU := ACCU or TMP TMP (or)
06. ACCU := ACCU + [0] [0] (+)
07. ACCU := ACCU + E E (+)
08. TMP := SL5 (A) A (SL5)
09. ACCU := ACCU + TMP TMP (+)
10. TMP := ACCU + K[t] K[t] (+)
11. E := D D (copy)
12. D := C C (copy)
13. C := SL30 (B) B (SL30)
14. B := A A (copy)
15. A := TMP TMP (copy)
16. ROTATE [i]
6<=t<=19
## operation Readbus Writebus
01. ACCU := B B (copy)
02. TMP := ACCU and C C (and)
03. ACCU := NOT B B (not)
04. ACCU := ACCU and D D (and)
05. TMP := ACCU or TMP TMP (or)
06. ACCU := W[13] W[13] (copy)
07. ACCU := ACCU xor W[8] W[8] (xor)
08. ACCU := ACCU xor W[2] W[2] (xor)
09. ACCU := ACCU xor W[0] W[0] (xor)
10. W[0] := SLl (ACCU) (SLl)
11. ACCU := W[0] W[0] (copy)
12. ACCU := ACCU + TMP TMP (+)
13. ACCU := ACCU + E E (+)
14. TMP : = SL5 (A) A (SL5)
15. ACCU := ACCU + TMP TMP (+)
16. TMP := ACCU + K[t] K[t] (+)
17. ■ Ξ := D D (copy)
18. D := C C (copy)
19. C := SL30 (B) B (SL30)
20. B := A A (copy)
21. A := TMP TMP (copy)
22. ROTATE W[i] 0<=t<=39 and 60<=t<=79
inal round
## operation Readbus Writebus
01. ACCU := A A (copy)
02. HO := ACCU + HO HO (+)
03. ACCU := B B (copy)
04. HI := ACCU + HI HI (+)
05. ACCU := C C (copy)
06. H2 := ACCU + H2 H2 ( +)
07. ACCU := D D (copy)
08. H3 := ACCU + H3 H3 (+)
09. ACCU := E E (copy)
10. H4 ACCU + H4 H4 (+)
MD5 ALGORITHM initialization
## operation Readbus Writebus
01. A := HO HO (copy)
02. B := HI HI (copy)
03. C := H2 H2 (copy)
04. D := H3 H3 (copy)
Round 1 (16 iterations): 0<=i<=15; k=0; s=7, 12 , 17, 22, 7, 12, 17, 22. ## operation Readbus Writebus ' 01. ACCU := B B (copy)
02. TMP := ACCU and C C (and) • ' 03.' ACCU := NOT B B (not)
04. ACCU := ACCU and D D (and)
05. TMP := ACCU or TMP TMP (or)
06. ACCU := W[k] W[k] (copy)
07. ACCU := ACCU + A A (+)
08. ACCU := ACCU + T[i] T[i] (+ )
09. TMP = ACCU + TMP TMP ( +)
10. ACCU := SL[s] (TMP) TMP (SL[s])
11. TMP = ACCU + B B (+)
12. A : = D D (copy)
13. D : = C C (copy)
14. C : = B B (copy)
15. B : = TMP TMP (copy)
16. ROTATE W[k]
Preparation for Round 2 01. ROTATE W[k]
Round 2(16 iterations) : 16<=i<=31; k=l; s=5 ,9, 14, 20 , 5, 9, 14, 20 , 5.
## operation Readbus Writebus
01. ACCU := B B (copy)
02. TMP := ACCU and D D (and)
03. ACCU := NOT D D (not)
04. ACCU := ACCU and C C (and)
05. TMP := ACCU or TMP TMP (or)
06. ACCU := W[k] W[k] (copy)
07. ACCU := ACCU + A A ( + )
08. ACCU := ACCU + T[i] T[i] ( + )
09. TMP := ACCU + TMP TMP ( + )
10. ACCU := SL[s] (TMP) TMP (SL[s])
11. TMP := ACCU +• B B ( + )
12. A = D D (copy)
1133.. DD := C C (copy) •
14. C = B • B (copy)
15. B = TMP TMP (copy)
16. ROTATE W[k]
17. ROTATE W[k] 18. ROTATE W[k]
19. ROTATE W[k]
20. ROTATE W[k] Preparation for Round 3
01. ROTATE W[k] 02. ROTATE W[k]
03. ROTATE W[k]
04. ROTATE W [k]
Round 3(16 iterations) :32<=i<=47 ; k=5; s=4, 11, 16, 23, , 11,16
As an example of how the information above should be interpreted, step number 2 of the SHA-1 initialisation section relates to the operation B:= H1, meaning that the register B is set to the value stored in H1. To achieve this, the tristate buffer 321 of register H1 and the tristate buffer 301 of the copy logic are enabled together. At the same time, the clock to register B is enabled, resulting in the data in H1 being written
into B. The tristate buffer control and clock signals are generated by the control circuit 400.
Similarly, step number 10 in the SHA-1 0<= t <= 15 stage relates to the operation TMP := ACCU + K[t]. The multiplexer and tristate buffer 332 is enabled for K[10]. The tristate buffer 304 is enabled for the ADD logic 215 and a gated clock signal is created and applied to the TMP register 120. In this way, the rising clock signal causes the sum of the data in K[10] and ACCU to be written into the TMP register.
The last instruction in the 0<= t <= 15 stage for SHA-1 (and the 0<= i <= 15 stage for MD5) causes the entire Wi chain to be rotated, so that W14 is loaded with the data previously in W15, W13 receives the data previously in W14, and W15 receives the data previously in WO. Advantageously, this instruction may be implemented in parallel with the instruction above it (Step 15) as the rotate instruction does not involve placing data onto the data bus. In this way, one clock cycle per iteration is saved, leading to a total saving of 80 cycles for SHA-1 and 64 cycles for MD5.
The embodiment presented has a bus width of 32 bits. However, it is possible to reduce the bus width to reduce the silicon area of the design at the expense of operational speed. If the bus width is reduced to 16 bits, each 32 bit XOR operation, for example, will take two cycles rather than one cycle if a 32 bit bus was used.
In the light of the foregoing description, it will be clear to the skilled man that various modifications may be mode within the scope of the invention.
The present invention includes and novel feature or combination of features disclosed herein either explicitly or any generalisation thereof irrespective of whether or not it relates to the claimed invention or mitigates any or all of the problems addressed.