Decryption method

Abstract

The invention relates to a method of determining a plaintext M on the basis of a cipher C and using a secret key d, wherein the secret key d is used in binary form, wherein the plaintext M is determined in each iteration step i for the corresponding bit di and a security variable Mn is determined in parallel therewith, and then a verification variable x is determined by means of a bit-compatible exponent of the secret key d.

Description

FIELD OF THE INVENTION

The invention relates to a method for determining a plaintext on the basis of a cipher.

BACKGROUND OF THE INVENTION

Such methods are known for example by the RSA method. In the RSA method, a plaintext is encrypted by means of a public key, wherein this cipher can be decrypted again by means of an associated secret key. Since the encrypted data are usually highly confidential and nevertheless are publicly accessible, the data are more and more frequently being exposed to attacks in order to spy out the secret key so that the encrypted data can be decrypted and thus undesirably determined in order to misuse the decrypted data.

Such attacks have become known as timing attacks or differential fault analysis (DFA) attacks, in which the computation time or running time of a calculation or a fault behavior during manipulations is observed in order to determine the secret key that is used during such processes.

Therefore, methods have been created which, using considerable computational effort through an inverse RSA function or a second RSA calculation, attempt to ascertain such manipulations and make them ineffective.

OBJECT AND SUMMARY OF THE INVENTION

The object of the invention is to provide a method for determining a plaintext on the basis of a cipher, which is not susceptible to timing attacks and differential fault analysis attacks and nevertheless is associated with a relatively low amount of additional effort.

This is achieved according to the invention by a method of determining a plaintext M on the basis of a cipher C and using a secret key d, wherein the secret key d is used in binary form, wherein the plaintext M is determined in each iteration step i for the corresponding bit d_i of the secret key and a security variable M_n is determined in parallel therewith, and then a verification variable x is determined by means of a bit-compatible exponent of the secret key d.

Advantageous further developments are described in the dependent claims.

BRIEF DESCRIPTION OF THE DRAWINGS

The invention will be further described with reference to an example of embodiment shown in the drawings to which, however, the invention is not restricted.

FIG. 1 shows a schematic illustration of the RSA method.

FIG. 2 shows a block diagram of the RSA method.

FIG. 3 shows an illustration of a timing attack.

FIG. 4 shows a block diagram of the “always multiplication and squaring” method.

FIG. 5 shows a block diagram of the verification method according to the invention.

FIG. 6 shows a block diagram of the verification method according to the invention.

FIG. 7 shows block diagram of the verification method according to the invention.

DESCRIPTION OF EMBODIMENTS

Encryption and decryption methods are very widespread today, since confidential information is used very frequently and is also transmitted in a publicly accessible manner. An implementation of such an encryption and the associated decryption according to the prior art will be described below with reference to the so-called RSA method according to Rivest, Shamir and Adleman. In the RSA method, firstly a plaintext M is encrypted using a public key g to form a cipher C, that is to say a secret text. This encrypted cipher C can then also be made public or transmitted, since the cipher C cannot be decrypted without the secret key d. The calculation of the plaintext M is carried out by a modular exponentiation (mod N) of the cipher C using the secret key d. FIG. 1 shows a schematic diagram 1 in order to illustrate the decryption according to the RSA method of M=C^d mod N. For this, FIG. 1 shows a block 2 which represents the RSA decryption. The input variables used are the cipher C and the secret key d, so that the plaintext M is obtained as the result.

The implementation of this equation generally takes place by means of the so-called “multiplication and squaring” algorithm. Here, the key d is used in its binary form with the length L:

$\begin{array}{c}d=d_0+2d_1+4d_2+\dots 2^{L-1}{d}_{L-1}\\ =\sum _{i=0}^{L-1}{2}^i·d_i\end{array}$ $d_i\in \left\{0,1\right\}$ If this form is used, the result is a product chain as follows:

$M=C^{d_0}·C^{2d_1}·C^{4d_1}·C^{8d_1}\dots ·C^{2^{L-1}{d}_{L-1}}\mathrm{mod}N$ $M=\prod _{i=0}^{L-1}{C}^{2^i{d}_i}\mathrm{mod}N$ If x_i=C²ⁱ, then in

$M=\prod _{i=0}^{L-1}{x}_i^{d_i}\mathrm{mod}N$ $\mathrm{where}{x}_i^{d_i}=\{\begin{array}{c}{x}_i\mathrm{for}{d}_i=1\\ 1\mathrm{for}{d}_i=0\end{array}$ the variable x_i can be calculated iteratively:x_i+1=C²ⁱ⁺¹=(C²ⁱ)²=x_i² The “multiplication and squaring” algorithm is thus obtained as a pseudo-code:

M=1;x=C;
for i= 0 to L-1
if d == 1
M = M * x mod N
end if
x=square(x) mod N
endfor

FIG. 2 shows the associated procedure 10 of modular exponentiation as a block diagram.

The method starts in block 11, and in block 12 the method is initialized with the values M=1, x=C and i=0. In block 13 an interrogation takes place as to whether the bit d; of the secret key d is equal to 1. If this is the case, the method continues with block 14; if not, the method continues with block 15. In block 14, M=x*M mod N is calculated. The method then also continues with block 15, wherein x=x² mod N is determined. Thereafter, in block 16, an interrogation takes place as to whether i=L−1. If this is the case, the method is terminated in block 18; if not, i=i+1 is set in block 17 and the method continues again with block 13. L cycles are carried out, in which in each case one bit d; of the secret key d is processed.

Timing attacks on the RSA method were introduced in 1998. In these attacks, the secret key d is derived from the different running time or computing time in the respective cycles. If d_i=1, the multiplication in block 14 is carried out, i.e. there is a long running time. If d_i=0, the multiplication in block 14 is not carried out and the result is therefore a short running time. Detection of the running time or of the computing times for each cycle takes place for example by evaluating the current consumption, by recording the cache activity in PC applications or by measuring the electromagnetic radiation of components.

Such a current consumption of a chip card microcontroller as a function of time is shown by way of example in FIG. 3 and illustrates the mode of operation of these timing attacks in a simple manner. FIG. 3 shows regions of different current consumption as a function of time, wherein the regions of low current consumption have two typical widths, i.e. durations. The first region 20 represents a region of squaring, in which x=x² mod N is determined, while the region 21 represents a region of multiplication, in which M=x*mod N is calculated. Since the last calculation according to the method of FIG. 2 is carried out only if the bit d_i=1, then for the present case d_i must be equal to 1. This is then followed by regions 22, 23 and 24, in which the multiplication is not carried out and thus d_i must be equal to 0. It is thus possible to detect in a relatively simple manner whether d_i=0 or d_i=1. The corresponding value of d; is shown in the bottom line of FIG. 3. It is thus possible to detect the respective key bit d_i based on the current curve by means of the different running times for “multiplication” and “squaring”. In order to prevent these attacks, use is made of the so-called “always multiplication and squaring” method which, for the case where d_i=0, always carries out an identical but ineffective multiplication which leads to a constant cycle time for d_i=1 or d_i=0. The associated pseudo-code is accordingly:

M=1;x=C;
for i= 0 to L-1
if d == 1
M = M * x mod N
else
M * x mod N
end if
x=square(x) mod N
endfor

FIG. 4 shows a block diagram 30 for illustrating this improved RSA method. The method starts in block 31, and in block 32 the method is initialized with corresponding start values. In block 33 an interrogation takes place as to whether d_i=1. If this is the case, the method continues with block 34; if not, the method continues with block 35. In block 34, M=x*M mod N is calculated. In block 35, x*M mod N is carried out as a so-called ineffective multiplication. The method then continues with block 36, in which x=x² mod N is determined. Thereafter, in block 37, an interrogation takes place as to whether i=L−1. If this is the case, the method is terminated in block 39; if not, i=i+1 is set in block 38 and the method continues again with block 33. L cycles are again carried out, in which in each case one bit d; of the secret key d is processed.

Following the implementation of the improved RSA method, another method of attack, the so-called differential fault analysis (DFA) attack, on this algorithm became known, according to which the multiplication in the individual cycles is disrupted for example by physical influences such as light, electromagnetic pulses, power supply pulses or the like. If a disruption of the multiplication does not have any effect on the end result, the associated cycle carries out an ineffective multiplication as described above. The corresponding key bit d_i is then d_i=0. However, if the disruption alters the end result, the key bit is thus d_i=1. If the method is attacked in this way on a cycle-by-cycle basis, the entire secret key d can be determined.

In order to prevent such an attack, the calculation of the cipher is usually verified by the inverse RSA function using the public key e through C=M^e or by a second RSA calculation. In the first case, the public key e must be known in the system. In the second case, the time taken for the calculation is doubled.

The method according to the invention for protection against the above-described DFA attacks provides for verification of the calculation of the exponential equation M=C^d by means of a checksum. In this case, a method is carried out which makes use of the ineffective multiplication shown in the method of FIG. 4. To this end, the cipher M_n of the binary complementary exponent of d is calculated during the ineffective multiplication, see FIG. 5. The method according to the invention as shown in FIG. 5 provides for verification of the “always multiplication and squaring” method by calculating M_n.

FIG. 5 shows a block diagram 40 for illustrating this method which has been improved with regard to DFA attacks. The method starts in block 41, and in block 42 the method is initialized with start values. In block 43 an interrogation takes place as to whether d_i=1. If this is the case, the method continues with block 44; if not, the method continues with block 45. In block 44, M=x*M mod N is calculated. In block 45, M_n=x*M mod N is calculated. The method then also continues with block 46, in which x=x² mod N is determined. Thereafter, in block 47, an interrogation takes place as to whether i=L−1. If this is the case, the method is terminated in block 49; if not, i=i+1 is set in block 48 and the method continues again with block 43. In block 49, the calculation C*M*M_n mod N=x is queried as part of the verification block 53. If the equation is satisfied, a non-disrupted calculation is recognized in block 51 and a corresponding signal is returned. However, if the equation is not satisfied, a disrupted calculation is recognized in block 50 and a corresponding error signal is returned. The method is terminated in block 52. L cycles are again carried out, in which in each case one bit d; of the secret key d is processed.

If, according to FIG. 5, the calculation M_n=x*M_n mod N is carried out during the ineffective multiplication, the following is obtained at the end of the last cycle:M_n=C^d mod N wherein the complement d of the key d has to be replaced by the equationd=2^L−1−d. This gives:M_n=C^2L−1−d mod N. If the producty=C·M·M_n mod N y=C·C^d·C^2L−1−d mod N y=C^2L mod N is calculated, the result y can be compared directly with the auxiliary variable x, which after L cycles assumes the same value x_L=C^2L mod N. Any disruption due to a so-called DFA attack thus means that x is not equal to the product y.

With just two multiplications and one comparison with a typical 1024-bit RSA (1024 multiplications+ 1024 squaring calculations), the effort for this verification is low.

The calculation is even more advantageous if M_n is initialized with C. There is thus no need for the multiplication by C after the last cycle. Moreover, the memory requirement is reduced since there is no need to store C after the initialization. Such a method is shown in FIG. 6, wherein optimization is carried out by initializing M_n=C prior to the exponentiation.

FIG. 6 shows a block diagram 60 for illustrating this method which has been improved with regard to DFA attacks and optimized. The method starts in block 61, and in block 62 the method is initialized with start values including M_n=C. In block 63 an interrogation takes place as to whether d_i=1. If this is the case, the method continues with block 64; if not, the method continues with block 65. In block 64, M=x*M mod N is calculated. In block 65, M_n=x*M_n mod N is calculated. The method then also continues with block 66, in which x=x² mod N is determined. Thereafter, in block 67, an interrogation takes place as to whether i=L−1. If this is the case, the method is terminated in block 69; if not, i=i+1 is set in block 68 and the method continues again with block 63. In block 69, the calculation M*M_n mod N=x is queried as part of the verification block 73. If the equation is satisfied, a non-disrupted calculation is recognized in block 71 and a corresponding signal is returned. However, if the equation is not satisfied, a disrupted calculation is recognized in block 70 and a corresponding error signal is returned. The method is terminated in block 72.

However, according to the invention, the above-described method can also be applied to other methods or to general mathematical structures, such as to processes of the “always addition and doubling” method. FIG. 7 shows a block diagram 80 for illustrating a corresponding “always addition and doubling” method which has been improved with regard to DFA attacks and optimized, such as an ECC or HECC method, wherein the ECC method is the method of elliptical curve cryptography and the HECC method is the method of hyperelliptical curve cryptography. The method starts in block 81, and in block 82 the method is initialized with start values. In block 83 an interrogation takes place as to whether d_i=1. If this is the case, the method continues with block 84; if not, the method continues with block 85. In block 84, M=x+M is calculated. In block 85, M_n=x+M_n is calculated. The method then also continues with block 86, in which x=2*x is determined. Thereafter, in block 87, an interrogation takes place as to whether i=L−1. If this is the case, the method continues in block 89; if not, i=i+1 is set in block 88 and the method continues again with block 83. In block 89, the calculation M+M_n=x is queried as part of the verification block 93. If the equation is satisfied, a non-disrupted calculation is recognized in block 91 and a corresponding signal is returned. However, if the equation is not satisfied, a disrupted calculation is recognized in block 90 and a corresponding error signal is returned. The method is terminated in block 92.

This verification method can also be used for general mathematical groups. Let (G,+,O) be a group containing elements of G, a neutral element O and a group linker “+”. The n-fold summing of a group element P is denoted n*P, in particular 0*P=O and (−n)*P=n*(−P), wherein “−P” is the inverse element of P. In order to protect the implementation of the operation d*P using an optionally also secret scalar factor d≧0 with a bit length L against timing attacks, an “always addition and doubling” algorithm can also be implemented in the same way as the “always multiplication and squaring” algorithm. The above-described protection against DFA attacks can also be transferred in an analogous manner; the auxiliary variable y is calculated at the end:y=M+M_n=(d*P)+((2^L−1−d)*P)+P y=2^L*P A DFA attack has then taken place when, and only when, for the auxiliary variable x, x≠y.

LIST OF REFERENCES

• 1 diagram
• 2 block of diagram 1
• 10 block diagram showing the procedure of modular exponentiation
• 11 block of block diagram 10
• 12 block of block diagram 10
• 13 block of block diagram 10
• 14 block of block diagram 10
• 15 block of block diagram 10
• 16 block of block diagram 10
• 17 block of block diagram 10
• 18 block of block diagram 10
• 20 region
• 21 region
• 22 region
• 23 region
• 24 region
• 30 block diagram
• 31 block of block diagram 30
• 32 block of block diagram 30
• 33 block of block diagram 30
• 34 block of block diagram 30
• 35 block of block diagram 30
• 36 block of block diagram 30
• 37 block of block diagram 30
• 38 block of block diagram 30
• 40 block diagram
• 41 block of block diagram 40
• 42 block of block diagram 40
• 43 block of block diagram 40
• 44 block of block diagram 40
• 45 block of block diagram 40
• 46 block of block diagram 40
• 47 block of block diagram 40
• 48 block of block diagram 40
• 49 block of block diagram 40
• 50 block of block diagram 40
• 51 block of block diagram 40
• 52 block of block diagram 40
• 53 verification block of block diagram 40
• 60 block diagram
• 61 block of block diagram 60
• 62 block of block diagram 60
• 63 block of block diagram 60
• 64 block of block diagram 60
• 65 block of block diagram 60
• 66 block of block diagram 60
• 67 block of block diagram 60
• 68 block of block diagram 60
• 69 block of block diagram 60
• 70 block of block diagram 60
• 71 block of block diagram 60
• 72 block of block diagram 60
• 73 verification block of block diagram 60
• 80 block diagram
• 81 block of block diagram 80
• 82 block of block diagram 80
• 83 block of block diagram 80
• 84 block of block diagram 80
• 85 block of block diagram 80
• 86 block of block diagram 80
• 87 block of block diagram 80
• 88 block of block diagram 80
• 89 block of block diagram 80
• 90 block of block diagram 80
• 91 block of block diagram 80
• 92 block of block diagram 80
• 93 verification block of block diagram 80

Claims

1. A method of determining a plaintext M based upon a cipher C in a chip card microcontroller and using a secret key d, wherein the secret key d is used in binary form containing bits di, the method comprising: initializing a plaintext variable M, a second variable Mn, and a third variable x, whereby x is initialized to the cipher C iteratively, for each bit di of the secret key d depending on the value of the corresponding bit di:determining whether di is equal to 1:if di is equal to 1, then calculating a new value of M from x and a previous value of M;if di is not equal to 1, then calculating a new value of Mn from x and a previous value of Mn;calculating a new value for x such that a final value of M is the plaintext M;comparing x to a value depending on final values of both M and Mn; andverifying, after a last iteration, whether calculation of the plaintext M was disrupted.

2. The method as claimed in claim 1, further comprising: determining the plaintext M by means of exponentiation (mod N).

3. The method as claimed in claim 1, further comprising: using a multiplication/squaring algorithm to determine the plaintext M.

4. The method as in claim 1, further comprising: using an addition/doubling algorithm to determine the plaintext M.

5. The method as claimed in claim 1, further comprising: calculating M=x*M mod N for di=1 to determine the plaintext M.

6. The method as claimed in claim 5, further comprising: calculating Mn=Mn*x mod N for di=0 to detect ineffective multiplication.

7. The method as in claim 1, further comprising: calculating M=x+M for di=1 to determine the plaintext M.

8. The method as claimed in claim 5, calculating Mn=x+Mn for di=0 to detect ineffective addition.

9. The method as in claim 1, further comprising: calculating M*Mn mod N=x for verification purposes.

10. The method as in claim 1, further comprising: calculating M+Mn=x for verification purposes.

11. The method of claim 1, further comprising: using a checksum to verify calculation of M=Cd.

12. The method of claim 1, further comprising: calculating the new value of Mn during ineffective multiplication.

13. The method of claim 1, further comprising: determining x=x2 mod N.

14. The method of claim 1, further comprising: calculating whether M*Mn mod N is equal to x.

15. The method of claim 14, further comprising: when M*Mn mod N is equal to x, recognizing a non-disrupted calculation and returning a corresponding signal.

16. The method of claim 14, further comprising: when M*Mn mod N is not equal to x, recognizing a disrupted calculation and returning an error signal.

17. The method of claim 1, further comprising: initializing Mn with C.

18. The method of claim 1, further comprising: calculating whether M+Mn is equal to x.

19. The method of claim 18, further comprising: when M+Mn is equal to x, recognizing a non-disrupted calculation and returning a corresponding signal.

20. The method of claim 18, further comprising: when M+Mn is not equal to x, recognizing a disrupted calculation and returning an error signal.