CA2288837C - Improved method and apparatus for protecting public key schemes from timing and fault attacks - Google Patents
Improved method and apparatus for protecting public key schemes from timing and fault attacks Download PDFInfo
- Publication number
- CA2288837C CA2288837C CA002288837A CA2288837A CA2288837C CA 2288837 C CA2288837 C CA 2288837C CA 002288837 A CA002288837 A CA 002288837A CA 2288837 A CA2288837 A CA 2288837A CA 2288837 C CA2288837 C CA 2288837C
- Authority
- CA
- Canada
- Prior art keywords
- mod
- improvement
- phi
- computation
- timing
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L9/00—Cryptographic mechanisms or cryptographic arrangements for secret or secure communications; Network security protocols
- H04L9/30—Public key, i.e. encryption algorithm being computationally infeasible to invert or user's encryption keys not requiring secrecy
- H04L9/3006—Public key, i.e. encryption algorithm being computationally infeasible to invert or user's encryption keys not requiring secrecy underlying computational problems or public-key parameters
- H04L9/302—Public key, i.e. encryption algorithm being computationally infeasible to invert or user's encryption keys not requiring secrecy underlying computational problems or public-key parameters involving the integer factorization problem, e.g. RSA or quadratic sieve [QS] schemes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F7/00—Methods or arrangements for processing data by operating upon the order or content of the data handled
- G06F7/60—Methods or arrangements for performing computations using a digital non-denominational number representation, i.e. number representation without radix; Computing devices using combinations of denominational and non-denominational quantity representations, e.g. using difunction pulse trains, STEELE computers, phase computers
- G06F7/72—Methods or arrangements for performing computations using a digital non-denominational number representation, i.e. number representation without radix; Computing devices using combinations of denominational and non-denominational quantity representations, e.g. using difunction pulse trains, STEELE computers, phase computers using residue arithmetic
- G06F7/723—Modular exponentiation
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L9/00—Cryptographic mechanisms or cryptographic arrangements for secret or secure communications; Network security protocols
- H04L9/002—Countermeasures against attacks on cryptographic mechanisms
- H04L9/004—Countermeasures against attacks on cryptographic mechanisms for fault attacks
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L9/00—Cryptographic mechanisms or cryptographic arrangements for secret or secure communications; Network security protocols
- H04L9/002—Countermeasures against attacks on cryptographic mechanisms
- H04L9/005—Countermeasures against attacks on cryptographic mechanisms for timing attacks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2207/00—Indexing scheme relating to methods or arrangements for processing data by operating upon the order or content of the data handled
- G06F2207/72—Indexing scheme relating to groups G06F7/72 - G06F7/729
- G06F2207/7219—Countermeasures against side channel or fault attacks
- G06F2207/7223—Randomisation as countermeasure against side channel attacks
- G06F2207/7233—Masking, e.g. (A**e)+r mod n
- G06F2207/7247—Modulo masking, e.g. A**e mod (n*r)
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2207/00—Indexing scheme relating to methods or arrangements for processing data by operating upon the order or content of the data handled
- G06F2207/72—Indexing scheme relating to groups G06F7/72 - G06F7/729
- G06F2207/7219—Countermeasures against side channel or fault attacks
- G06F2207/7223—Randomisation as countermeasure against side channel attacks
- G06F2207/7257—Random modification not requiring correction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2207/00—Indexing scheme relating to methods or arrangements for processing data by operating upon the order or content of the data handled
- G06F2207/72—Indexing scheme relating to groups G06F7/72 - G06F7/729
- G06F2207/7219—Countermeasures against side channel or fault attacks
- G06F2207/7271—Fault verification, e.g. comparing two values which should be the same, unless a computational fault occurred
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L2209/00—Additional information or applications relating to cryptographic mechanisms or cryptographic arrangements for secret or secure communication H04L9/00
- H04L2209/08—Randomization, e.g. dummy operations or using noise
Abstract
Improved methods and apparatus for protecting pub-lic key schemes based on modular exponentiation (including RSA and Diffie-Hellman) from indirect cryptanalytic tech-niques such as timing and fault attacks. Known methods for making the implementation of number-theoretic schemes resistant to such attacks typically double their running time, whereas the novel methods and apparatus described in this patent add only negligible overhead. This improvement is particularly significant in smart card and software-based implementations, in which the modular exponentiation operation is quite slow, and doubling its time may be an unacceptable solution.
Description
IMPROVED METHOD AND APPARATUS FOR PROTECTING
PUBLIC KEY SCHEME.S FROM TIMING AND FAULT ATTACKS
FIELD OF INVENTION
The present invention relates to novel techniques, methods and apparatus, for maldng number-theoretic public key schemes (including encryption schemes, signature schemes, identification schemes, key management schemes, etc.) resistant to timing and fault attacks.
BACKGROUND OF IIWENTION .
1. Introduction.
The simplest attack on a given cryptosystem is to exhaustively search for the key. There are many variants of this attack (known ciphertext, known cleartext, chosen cleartext, etc.), but they are all based on a procedure which tries the keys one by one until the correct key is encountered. If the key consists of n random bits, the expected running time of this procedure is 2~(n-1). This attack can be easily foiled by using a large enough n (e.g., n> 100).
To attack cryptosystems with large keys, cryptanalysts try to- find mathematical or statistical weaknesses which reduce the size of the search space (preferably to 1). Although many techniques and results are classified for national security reasons, it is safe to assume that it is increasingly difficult to find such weaknesses in modern schemes designed by experienced cryptographers and implemented on high speed microprocessors.
To successfully attack strong cryptosystems, the cryptanalyst must use indirect techniques. This is best done when the cryptanalyst is either in close physical proximity to the cryptographic-device, or has it under his complete control.
The cryptographic device is assumed to be a black box which contains a known algorithm and an unknown key. The cryptanalyst cannot open this box and read its key, but he can observe its behavior under various circumstances.
One of the best known examples of such an indirect attack is TEMPEST, which tries to deduce the key by analyzing electromagnetic radiation emanating from the black box during the computation of the ciphertext. Techniques for applying and preventing such attacks have been extensively studied for more than 50 years, and by now this is a well understood problem.
Two powerful indirect attacks were discovered and published recently: In December 1995, P. Kocher, "Cryptanalysis of Dif6e-Hellman, RSA, DSS, and Other Systems Using Timing Attacks," technical report, 12/'7/95, described a timing attack, and in September 1996, D. Boneh, R.A. Demillo and R.J. Lipton, "On the Importance of Checldng Computations," technical report, 9/25/96 (an extended version appears in the Proceedings of Eurocrypt 97, May 1997) described a fault attack. Both attacks were originally designed for and are most successful against public key schemes based on number theoretic principles, such as RSA, but they were later extended to classical cryptosystems as well (e.g., by E. Biham and A. Shamir, "A New Cryptanalytic Attack on DF.S," technical report, 10/18/96. An extended version appears in the Proceedings of Crypto 97, August 1997).
Such attacks are particularly useful when the scheme is implemented on a smart card, which is distributed by a bank, computer network, cellular phone operator, or pay-TV broadcaster to its customers. Hackers do not usually have the financial and technical resources required to read the contents of the key registers inside the smart card, but they have complete control on the input/output, clock, reset, and power connections of the smart card. They can carefully measure the duration of the various operations, how much power they consume, what happens when the computation is interrupted or carried out under abnormal operating conditions, etc. Since the tests are carried out in the privacy of the customer's home, the card manufacturer cannot prevent them or even learn about their existence.
2. TimingAttacks.
Timing attacks are based on the assumption that some of the basic operations carried out during the cryptographic calculation require a non-constant amount of time which depends on the actual values being operated upon. This implies that some information about these unknown intermediate values leaks out by measuring the length of the cryptographic computation. If these intermediate values are computed from known cleartext bits and unknown key bits by a known cryptographic algorithm, the.attacker can try to use the leaked intermediate values to deduce the key.
The main difficulty in carrying out this attack is that the attacker knows only the total amount of time required to carry out the cryptographic computation, but not the timing of individual computational steps. Kocher's main contribution is in developing an efficient technique for handling this difficulty in many cases of practical interest.
For the sake of concreteness, we describe Kocher's attack on the RSA
PUBLIC KEY SCHEME.S FROM TIMING AND FAULT ATTACKS
FIELD OF INVENTION
The present invention relates to novel techniques, methods and apparatus, for maldng number-theoretic public key schemes (including encryption schemes, signature schemes, identification schemes, key management schemes, etc.) resistant to timing and fault attacks.
BACKGROUND OF IIWENTION .
1. Introduction.
The simplest attack on a given cryptosystem is to exhaustively search for the key. There are many variants of this attack (known ciphertext, known cleartext, chosen cleartext, etc.), but they are all based on a procedure which tries the keys one by one until the correct key is encountered. If the key consists of n random bits, the expected running time of this procedure is 2~(n-1). This attack can be easily foiled by using a large enough n (e.g., n> 100).
To attack cryptosystems with large keys, cryptanalysts try to- find mathematical or statistical weaknesses which reduce the size of the search space (preferably to 1). Although many techniques and results are classified for national security reasons, it is safe to assume that it is increasingly difficult to find such weaknesses in modern schemes designed by experienced cryptographers and implemented on high speed microprocessors.
To successfully attack strong cryptosystems, the cryptanalyst must use indirect techniques. This is best done when the cryptanalyst is either in close physical proximity to the cryptographic-device, or has it under his complete control.
The cryptographic device is assumed to be a black box which contains a known algorithm and an unknown key. The cryptanalyst cannot open this box and read its key, but he can observe its behavior under various circumstances.
One of the best known examples of such an indirect attack is TEMPEST, which tries to deduce the key by analyzing electromagnetic radiation emanating from the black box during the computation of the ciphertext. Techniques for applying and preventing such attacks have been extensively studied for more than 50 years, and by now this is a well understood problem.
Two powerful indirect attacks were discovered and published recently: In December 1995, P. Kocher, "Cryptanalysis of Dif6e-Hellman, RSA, DSS, and Other Systems Using Timing Attacks," technical report, 12/'7/95, described a timing attack, and in September 1996, D. Boneh, R.A. Demillo and R.J. Lipton, "On the Importance of Checldng Computations," technical report, 9/25/96 (an extended version appears in the Proceedings of Eurocrypt 97, May 1997) described a fault attack. Both attacks were originally designed for and are most successful against public key schemes based on number theoretic principles, such as RSA, but they were later extended to classical cryptosystems as well (e.g., by E. Biham and A. Shamir, "A New Cryptanalytic Attack on DF.S," technical report, 10/18/96. An extended version appears in the Proceedings of Crypto 97, August 1997).
Such attacks are particularly useful when the scheme is implemented on a smart card, which is distributed by a bank, computer network, cellular phone operator, or pay-TV broadcaster to its customers. Hackers do not usually have the financial and technical resources required to read the contents of the key registers inside the smart card, but they have complete control on the input/output, clock, reset, and power connections of the smart card. They can carefully measure the duration of the various operations, how much power they consume, what happens when the computation is interrupted or carried out under abnormal operating conditions, etc. Since the tests are carried out in the privacy of the customer's home, the card manufacturer cannot prevent them or even learn about their existence.
2. TimingAttacks.
Timing attacks are based on the assumption that some of the basic operations carried out during the cryptographic calculation require a non-constant amount of time which depends on the actual values being operated upon. This implies that some information about these unknown intermediate values leaks out by measuring the length of the cryptographic computation. If these intermediate values are computed from known cleartext bits and unknown key bits by a known cryptographic algorithm, the.attacker can try to use the leaked intermediate values to deduce the key.
The main difficulty in carrying out this attack is that the attacker knows only the total amount of time required to carry out the cryptographic computation, but not the timing of individual computational steps. Kocher's main contribution is in developing an efficient technique for handling this difficulty in many cases of practical interest.
For the sake of concreteness, we describe Kocher's attack on the RSA
cryptosystem. The black box is assumed to contain a publicly known modulus n and a secret exponent d. Given an input number x, the box performs the modular exponentiation x~ d (mod n) by using the standard square-and-multiply technique. In this description, the symbol " ~ " is exponentiation and the symbol "_ " is a subscript. The result (which can be the decryption of the ciphertext x, the signature of the message x, or the response to a random identification challenge x, depending on the application) is sent out as soon as it is produced, and thus the attacker can measure the total number of clock cycles taken by all the modular multiplications.
Standard implementations of modular multiplication require a non-constant amount of time, since they skip multiplication steps involving leading zeroes, and reduction steps when the result is smaller than the modulus. The attacker chooses a large number of random inputs x, and measures the actual timing distribution T
0 of the modular exponentiation operation carried out by the black box. He then measures for each x (by computer simulation, using his knowledge of how the scheme is implemented) the precise timing of an initial square-only operation, and separately, the precise timing of an initial square-and-multiply operation. The result is a pair of timing distributions T 1 and T 2, which are not identical. A11 the cryptographic computations carried out in the black box use the same exponent d, and its first bit determines which one of the two computed distributions T 1 and T 2 is the initial part of the experimentally computed T 0. The timing of the remaining steps of the computations can be assumed to be a random variable R, which is normally distributed and uncorrelated with either T 1 or T 2. Since T 0 is either T 1+R or T 2+R, the attacker can decide which case is more --~ T - -__ ____----likely by finding which one of xhe two distributions T 0-T 1 and T 0-T 2 has a lower variance.
After finding the first bit of the secret exponent d, the attacker knows the actual inputs to the second computational step, and thus he can apply the same technique (with properly modified experimental and simulated timing distributions T'_0, T'_l, and T' 2) to find the second bit of d. By repeating this procedure about times, he can compute all the bits of d, and thus break the RSA. scheme.
A similar timing attack can be applied to any cryptographic scheme in which the black box raises all its inputs x 1,x 2,... to the same secret power d modulo the same known n (which can be either a prime or a composite number). For example, in one of the variants of the Diffie-Hellman key distribution scheme, all the users agree on a prime modulus n and on a generator g of the multiplicative group Z~* n.
Each user chooses a random secret exponent d, and computes y = g ~ d (mod n) as his public key. To establish a common secret key with another user, the first user sends out his public key y=g ~ d (mod n), and receives a similarly computed public key x=g ~ e (mod n) from the other user. Their common cryptographic key is z=g ~(d*e) (mod n) which the first user computes by evaluating x~ d (mod n).
When the first user communicates with several parties, he raises several known values x 1,x 2,...to the same secret power d modulo the same known modulus n. By measuring the timing of sufficiently many such computations, the attacker can determine d and thus find all the cryptographic keys z 1,z_2,... employed by that user.
The timing attack has to be modified if the computation of x~ d (mod n) = CA 02288837 1999-11-08 for a composite modulus n=p*q is carried out by computing x~ d (mod p), x~ d (mod q), and combining the results by the Chinese Remainder Theorem (CRT). This is a common way of making the computation about 4 times faster when the factorization of n is known. The problem for the attacker is that he does not know the secret factors p and q of the public modulus n, and thus cannot simulate the timing distributions T 1 and T 2. Kocher's solution is to concentrate on the first step of the CRT
computation, in which the input x is reduced modulo p. If x is smaller than p, no modular reduction is required, and thus the computation is considerably faster than when x is larger than or equal to p. The attacker thus presents to the black box a large number of inputs x which are very close to each other, and uses the average time of such computations to decide whether these x's are above or below p. A decision procedure for this question can be repeatedly used to find the precise value of p by binary search.
Shortly after the discovery of this attack, researchers tried to develop implementations which are immune to it The simplest idea is to make sure that-all the cryptographic operations take exactly the same amount of time, regardless of the values of the cleartexts and keys. However, achieving this is surprisingly difficult for the following reasons:
(a) In many cases, the implementor wants to run the same algorithm in software on different (and perhaps unknown) machines. An implementation which is constant time on one microprocessor may be variable time on another microprocessor or even on an enhanced version of the same microprocessor.
(b) On a multitasldng machine the running time may depend on the amount of free memory, the cache hit rate, the number of external interrupts, etc. This can change a constant time implementation under one set of circumstances into a variable time implementation under another set of circumstances.
(c) If the implementor tries to use a real time clock to force all the computations to take the same amount of time, he must slow all of them down to their worst cases. Since he cannot use any input-dependent optimization technique, the implementation is likely to be unacceptably slow.
The best protective technique proposed so far against Kocher's timing attack on modular exponentiation is to replace each input x by a modified version y=x*r (mod n) where r is a secret random number between 1 and n-1. To compute x~ d (mod n), the program computes y~ d (mod n) and r~ d (mod n), and then uses the multiplicative property of modular exponentiation to compute x~ d (mod n) as y~ d/r ~ d (mod n). Since both y and r are unknown, the attacker cannot simulate these computations in order to find the successive bits of d in the non-CRT
computation, and cannot perform binary search in the CRT version of the computation.
Unfortunately, this randomization technique doubles the expected running time of the computation.
3. Fault Attacks.
Fault attacks try to introduce errors into the cryptographic computation, and to identify the key by analyzing the mathematical and statistical properties of the erroneously computed results. - Among the many techniques suggested so far for 'introducing such errors are the use of ionizing radiation, unusual operating temperatures, power and clock glitches, and laser-based chip microsurgery. Some of the attacks are differential (i.e., they carry out both a correct and an erroneous computation with the same input and analyze their differences), while other attacks just use the erroneous results.
The original fault attack on public key cryptosystems was described in Boneh, Demillo and Lipton, and required several cryptographic computations. We now describe an improved version of this attack, due to Arjen Lenstra, which requires a single faulty computation. We assume that the black box uses the RSA scheme to sign a given message x. The computation of x~ d (mod n) is carried out with the CRT method by first reducing x modulo p and q to get x 1 and x 2, then computing y_1=x 1~ d (mod p) and y_2=x 2~ d (mod q), and finally combining y_1 and y_2 to get the signature y (mod n) with the CRT method. We assume that a single error is introduced at a random time during this computation by applying mild physical stress to the black box.
Without loss of generality, we can assume that the error was introduced during the computation of x 1~ d (mod p), and thus instead of getting the correct y_I, the box computed an erroneous y'_l. When y'_1 and y_2 are combined by the CRT method, the box computes an incorrect signature y' which is provided to the attacker.
The main observation is that the attacker knows the signature verification exponent e, for which y~ e=x (mod n). Due to the error, y' ~ e-x is non-zero mod p, but zero mod q, and thus it is a multiple of q which is not a multiple of n. The attacker can T_ ~
thus factor n by computing the greatest common divisor of y' ~ e-x (mod n) and n, which is an easy computation.
To protect cryptographic schemes against fault attacks, Boneh, Demillo and Lipton recommend that each computation should be carried out twice (preferably by different algorithms). If any discrepancy is found between the two results, the box should not output anything. This provides strong protection from random faults (which are unlikely to affect the two computations in an identical way), but it slows down the computation by a factor of 2. Such a slowdown is particularly noticeable in smart card implementations of public key schemes, which are quite slow to begin with.
SUMMARY OF THE IIWENTION
The present invention relates to method and apparatus for providing protective techniques for public key schemes, which provide strong protection against the described timing and fault attacks without incurring the twofold slowdown made necessary by the previously known protective techniques.
The first technique is designed to protect non-CRT implementations of public key schemes against timing attacks. It is applicable to the RSA
cryptosystem, RSA digital signature scheme, Diffie-Hellman key distribution scheme, and any other number theoretic scheme in which the black box raises a known input x to a fixed secret exponent d modulo a public n whose factorization is known to the black box.
The second technique is designed to protect CRT-based implementations of public key schemes from both timing and fault attacks. The main problem is how to verify the correctness of the computations of x 1~ d (mod p) and x 2~ d (mod q) without repeating them a second time (or verifying each step separately, which again doubles the running time). The invention provides a novel error-detection technique for such number theoretic computations which is much more efficient.
Accordingly, in one aspect of the present invention there is provided in a method of implementing public key schemes containing a non-CRT form of the modular exponentiation operation x"d (mod n), for the purpose of making them more resistant to timing attacks, the improvement comprising the steps of:
a. computing or storing the computed value of t=phi(n), where phi is Euler's totient function of the modulus n;
b. selecting some secret integer i; and c. replacing the computation of x~d (mod n) by the computation of x~(d +
i*t) (mod n).
According to another aspect of the present invention there is provided in a method of implementing public key schemes containing a CRT form of the modular exponentiation operation x"d (mod n) where n=p*q, for the purpose of making them more resistant to timing and fault attacks, the improvement comprising the steps of:
a. selecting some secret integer j;
b. computing v_l=x (mod j*p), v_2=x (mod j*q), d_1=d (mod phi(j*p)), d 2=d (mod phi(j*q)), w_1=v_1~d_1 (mod j*p), and w_2=v_2~d_2 (mod j*q);
c. aborting the computation if w_1 and w 2 are not equal modulo j; and d. otherwise, computing y l=w_1 (mod p), y_2=w_2 (mod q), and combining them by the Chinese Remainder Theorem to obtain the result of x~d (mod n).
According to yet another aspect of the present invention there is provided in a method of implementing public key schemes containing a CRT form of the modular exponentiation operation x"d (mod n) where n=p*q, for the purpose of making them more resistant to timing and fault attacks, the improvement comprising the steps of:
a. selecting two secret integers j_1 and j_2;
b. computing v_1=x (mod j_1*p), v_2=x (mod j_2*q), d_1=d (mod phi(j_1*p)), d_2=d (mod phi(j_2*q)), w_1=v_1~d_1 (mod j_1*p), and w_2=v_2~d_2 (mod j_2*q);
c. computing v_3=x (mod j_l), v_4=x (mod j_2), d_3=d (mod j_1), d 4=d (mod j_2), w_3=v_3~d 3(mod j_1), and w 4=v 4~d 4(mod j_2);
d. aborting the computation if w 3 is not equal to w_1 modulo j_l, or if w_4 is not equal to w_2 modulo j_2; and e. otherwise, computing y 1=w_1 (mod p), y_2=w_2 (mod q), and combining them by the Chinese Remainder Theorem to obtain the result of x~d (mod n).
According to still yet another aspect of the present invention there is provided in an apparatus for implementing public key schemes containing a non-CRT form of the modular exponentiation operation x~d (mod n), for the purpose of making them more resistant to timing attacks, the improvement comprising:
a. means for computing or storing the computed value of t=phi(n), where phi is Euler's totient function of the modulus n;
b. means for selecting some secret integer i; and c. means for replacing the computation of x~d (mod n) by the computation of x~(d+i*t) (mod n).
10a According to still yet another aspect of the present invention there is provided in an apparatus for implementing public key schemes containing a CRT form of the modular exponentiation operation x~d (mod n) where n=p*q, for the purpose of making them more resistant to timing and fault attacks, the improvement comprising:
a. means for selecting some secret integer j;
b. means for computing v_1=x (mod j*p), v_2=x (mod j*q), d_1=d (mod phi(j *p)), d_2=d (mod phi(j *q)), w 1=v 1~d 1(mod j*p), and w_2=v 2~d 2(mod j*q);
c. means for aborting the computation if w 1 and w_2 are not equal modulo j; and d. otherwise, means for computing y-1=w_1 (mod p), y_2=w_2 (mod q), and combining them by the Chinese Remainder Theorem to obtain the result of x~d (mod n).
10b According to still yet another aspect of the present invention there is provided in an apparatus for implementing public key schemes containing a CRT form of the modular exponentiation operation x~d (mod n) where n=p*q, for the purpose of making them more resistant to timing and fault attacks, the improvement comprising:
a. selecting two secret integers j_1 and j_2;
b. computing v_1=x (mod j_1*p), v_2=x (mod j_2*q), d_1=d (mod phi(j_1 *p)), d_2=d (mod phi(j 2*q)), w 1=v l~d l(mod j_1 *p), and w_2=v_2~d 2 (mod j_2*q);
c. computing v_3=x (mod j_1), v_4=x (mod j_2), d_3=d (mod j_1), d_4=d (mod j_2), w_3=v_3~d_3 (mod j_1), and w_4=v_4~d 4(mod j_2);
d. aborting the computation if w 3 is not equal to w_1 modulo j_1, or if w 4 is not equal to w_2 modulo j_2; and e. otherwise, computing y_l=w_1 (mod p), y_2=w_2 (mod q), and combining them by the Chinese Remainder Theorem to obtain the result of x~d (mod n).
lOc BRIEF DESCRIPTION OF THE FIGURES
Embodiments of the present invention will now be described more fully with reference to the accompanying drawings in which:
Fig. 1 shows schematically the method and apparatus of the invention as it relates to a first technique designed to protect non-CRT implementations of a public key scheme against timing attacks.
Fig. 2 shows schematically the method and apparatus of the invention as it relates to a second technique designed to protect CRT-based implementations of a public key scheme against both timing and fault attacks.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
Referring now to the drawings, preferred embodiments will now be described in detail. Since public key schemes and computer hardware and software implementation are well known to those of skill in the art, no description of same is deemed necessary to a full, concise and exact understanding of the present invention.
We now describe two novel protective techniques for public key schemes, which provide strong protection against the described timing and fault attacks without incurring the twofold slowdown made necessary by the previously known protective techniques.
The first technique, shown in Fig. 1, is designed to protect non-CRT
implementations of public key schemes against timing attacks. It is applicable to the RSA cryptosystem, RSA digital signature scheme, Diffie-Hellman key distribution scheme, and any other number theoretic scheme in which the black box 10 raises a 10d known input x to a fixed secret exponent d modulo a public n whose factorization is known to the black box.
The main observation is that for each n there exists a number t=phi(n), box 12, such that for any x between 1 and n which is relatively prime to n, x~
t = 1 (mod n). This phi is called Euler's totient function: when n is prime, phi(n)=n-1, and when n=p*q,.phi(n)=(p-1)*(q-1). The implementation can thus replace the computation of x~ d(mod n) by the computation of x~(d+i*t) (mod n) for any integer i without changing the computed result, since x ~ (d+it) = (x ~ d) *(x ~ t) ~ i= (x ~ d) *1 ~ i =x ~ d (mod n), box 14.
This equality can be shown to be true even when x is not relatively prime to n, but this case is unlikely to arise in practice. Note also that any integral multiple of GCD((p-1), (q-1)) can replace phi(n) in our choice of t.
The proposed protection technique for public key schemes based on the modular exponentiation operation is to choose a new random secret i in each computation, and to replace the computation of x d (mod n) by the computation of x~ (d+i*t) (mod n) where t is the precomputed value of phi(n). Since the bits of these (d+i*t) for the various i are different, each exponentiation uses a different sequence of square-and-multiply steps, and thus the attacker cannot use Kocher's timing attack to analyze the timing distribution of several exponentiations, even though all of them compute the same d-th power of their inputs.
The efficiency of this technique is based on the fact that t is always smaller than n, and the randomizing element i can be chosen as a relatively small number. If n and d are 1024 bit numbers, and i is a random 32 bit number (which are the currently recommended sizes), d+i*t is a 1056 bit number. The process of raising the input x to the 1056 bit power d+i*t requires only 3% more square-and-multiply operations than the process of raising x to the original 1024 bit power d. This is much better than the alternative randoniization technique described earlier, which doubles the running time.
The second technique, shown in Fig. 2, is designed to protect CRT-based implementations of public key schemes from both timing and fault attacks. The main problem is how to verify t h e correctness of the computations of x 1~ d (mod p) and x 2~ d (mod q) without repeating them a second time (or verifying each step separately, which again doubles the running time). We now descnbe a novel error-detection technique for such number theoretic computations which is much more efficient.
In each computation, the black box 20 chooses a new random integer j (the recommended size of j is also 32 bit), box 22. Instead of computing x 1=x (mod p) and x 2=x (mod q) followed by y_1=x 1~ d (mod p) and y_2=x 2~ d (mod q), the box computes v_1=x (mod j*p), v_2=x (mod j*q), d_1=d (mod phi(j*p)), and d 2=d (mod phi(j*q)), box 24, followed by w 1=v_1 ~ d 1(mod j*p) and w_2=v 2~ d_2 (mod j*q), box26.
The main observation is that from w 1 and w 2 it is easy to derive y_1 and y_2 by further reductions (namely, y_1=w =w-1 (p) and y_2=w_2 (mod q)), box 28, and thus it is easy to compute the final result y by the Chinese remainder Theorem, box 30. However, we can also obtain the value of x~ d (mod j) in two different ways: as w 1(mod j) and as w 2(mod j), box 32. We can now use the equality of these two values (which were obtained from the two halves of the computation, and mixed with the 'derivation of y_l, y_2 in a very strong way), box 34, as a test of correctness: In a faultless computation the two values will always be the same, whereas in a faulty computation (with random faults) the probability that the two values will be the same is about 1/(2 ~ 32), see decision box 36 in which an abort is commanded for a faulty computation. This error detection technique is thus sufficient for any application in which the total number of modular exponentiations is significantly smaller than 2 32 (about 4 billion).
The overall time complexity of this implementation is higher by a few percent than the time complexity of standard implementations, since the exponentiations are carried out modulo 512+32=544 bit moduli j*p and j*q instead of 512 bit moduli p and q. However, this is much faster than repeating each exponentiation a second time to verify its correctness.
An additional benefit of this randomization technique is that it also provides protection from timing attacks at no extra cost. Kocher's original attack on CRT-based implementations concentrates on the initial modular reduction (mod p), and uses binary search to find increasingly accurate approximations of p from multiple computations.
By using our proposed technique, each computation uses a different modulus j*p in its initial reduction step, and thus the attacker cannot refine his knowledge of the modulus by analyzing a large number of exponentiations.
There are many optimizations and variations of this technique, which ~ CA 02288837 1999-11-08 should be obvious to anyone skilled in the art. For example, it is possible to impose additional restrictions (such as primality) on the choice of the small multiplier j which make it somewhat less likely that faulty computations will remain undetected.
Another modification of the technique is to test each half of the computation by a separate recomputation modulo a different small modulus, rather than by comparing the two results modulo a common small modulus. More precisely, the implementation can choose two small numbers j_1 and j 2, and then compute the foDowing quantities:
v 1=x (mod j_1*p), v -2=x (mod j_2*q), v -3=x (mod j_1), v-4=x (mod j_2) ;
d1=d (mod phi(j_1*p)), d_2=d (mod phi(j_2*q)), d_3=d (mod phi(j_1)), d_4=d (mod phi(j_2);
w_1=v_1 ~ d_1 (mod j_1 *p), w 2=v_2 ~ d_2 (mod j_2*P), w_3 =v_3 ~ d_3 (mod j_1), w 4=v_4 ~ d_4 (mod j_2).
To check the correctness of the computation, the black box verifies that w 1=w 3(mod j_i) and w 2=w 4(mod j 2). The only expensive operations are the computation of w 1 and w 2, since the small exponentiations in the computation of w 3 and w 4 are very efficient. This recomputation technique is slower than the original comparison technique, but may be slightly more resistant to certain types of non-random faults.
While the invention has been described with respect to certain embodiments thereof, it will be appreciated by one skilled in the art that variations and modifications may be made without departing from the spirit and scope of the invention.
_--- __ T
BIBLIOGRAPHY
1. E. Biham and A. Shamir, "A New Cryptanalytic Attack on DES", technical report, 10/18/96. An Extended version appears in the Proceedings of Crypto 97;
August 1997.
2. D. Boneh, R.A. Demillo and R.J. Lipton, "On the Importance of Checidng Computations", technical report, 9/25/96. An Extended version appears in the Proceedings of Eurocrypt 97, May 1997.
3. P. Kocher, "Cryptanalysis of Diffie-Hellman, RSA, DSS, and Other Systems Using Timing Attacks", technical report, 1217/95.
Standard implementations of modular multiplication require a non-constant amount of time, since they skip multiplication steps involving leading zeroes, and reduction steps when the result is smaller than the modulus. The attacker chooses a large number of random inputs x, and measures the actual timing distribution T
0 of the modular exponentiation operation carried out by the black box. He then measures for each x (by computer simulation, using his knowledge of how the scheme is implemented) the precise timing of an initial square-only operation, and separately, the precise timing of an initial square-and-multiply operation. The result is a pair of timing distributions T 1 and T 2, which are not identical. A11 the cryptographic computations carried out in the black box use the same exponent d, and its first bit determines which one of the two computed distributions T 1 and T 2 is the initial part of the experimentally computed T 0. The timing of the remaining steps of the computations can be assumed to be a random variable R, which is normally distributed and uncorrelated with either T 1 or T 2. Since T 0 is either T 1+R or T 2+R, the attacker can decide which case is more --~ T - -__ ____----likely by finding which one of xhe two distributions T 0-T 1 and T 0-T 2 has a lower variance.
After finding the first bit of the secret exponent d, the attacker knows the actual inputs to the second computational step, and thus he can apply the same technique (with properly modified experimental and simulated timing distributions T'_0, T'_l, and T' 2) to find the second bit of d. By repeating this procedure about times, he can compute all the bits of d, and thus break the RSA. scheme.
A similar timing attack can be applied to any cryptographic scheme in which the black box raises all its inputs x 1,x 2,... to the same secret power d modulo the same known n (which can be either a prime or a composite number). For example, in one of the variants of the Diffie-Hellman key distribution scheme, all the users agree on a prime modulus n and on a generator g of the multiplicative group Z~* n.
Each user chooses a random secret exponent d, and computes y = g ~ d (mod n) as his public key. To establish a common secret key with another user, the first user sends out his public key y=g ~ d (mod n), and receives a similarly computed public key x=g ~ e (mod n) from the other user. Their common cryptographic key is z=g ~(d*e) (mod n) which the first user computes by evaluating x~ d (mod n).
When the first user communicates with several parties, he raises several known values x 1,x 2,...to the same secret power d modulo the same known modulus n. By measuring the timing of sufficiently many such computations, the attacker can determine d and thus find all the cryptographic keys z 1,z_2,... employed by that user.
The timing attack has to be modified if the computation of x~ d (mod n) = CA 02288837 1999-11-08 for a composite modulus n=p*q is carried out by computing x~ d (mod p), x~ d (mod q), and combining the results by the Chinese Remainder Theorem (CRT). This is a common way of making the computation about 4 times faster when the factorization of n is known. The problem for the attacker is that he does not know the secret factors p and q of the public modulus n, and thus cannot simulate the timing distributions T 1 and T 2. Kocher's solution is to concentrate on the first step of the CRT
computation, in which the input x is reduced modulo p. If x is smaller than p, no modular reduction is required, and thus the computation is considerably faster than when x is larger than or equal to p. The attacker thus presents to the black box a large number of inputs x which are very close to each other, and uses the average time of such computations to decide whether these x's are above or below p. A decision procedure for this question can be repeatedly used to find the precise value of p by binary search.
Shortly after the discovery of this attack, researchers tried to develop implementations which are immune to it The simplest idea is to make sure that-all the cryptographic operations take exactly the same amount of time, regardless of the values of the cleartexts and keys. However, achieving this is surprisingly difficult for the following reasons:
(a) In many cases, the implementor wants to run the same algorithm in software on different (and perhaps unknown) machines. An implementation which is constant time on one microprocessor may be variable time on another microprocessor or even on an enhanced version of the same microprocessor.
(b) On a multitasldng machine the running time may depend on the amount of free memory, the cache hit rate, the number of external interrupts, etc. This can change a constant time implementation under one set of circumstances into a variable time implementation under another set of circumstances.
(c) If the implementor tries to use a real time clock to force all the computations to take the same amount of time, he must slow all of them down to their worst cases. Since he cannot use any input-dependent optimization technique, the implementation is likely to be unacceptably slow.
The best protective technique proposed so far against Kocher's timing attack on modular exponentiation is to replace each input x by a modified version y=x*r (mod n) where r is a secret random number between 1 and n-1. To compute x~ d (mod n), the program computes y~ d (mod n) and r~ d (mod n), and then uses the multiplicative property of modular exponentiation to compute x~ d (mod n) as y~ d/r ~ d (mod n). Since both y and r are unknown, the attacker cannot simulate these computations in order to find the successive bits of d in the non-CRT
computation, and cannot perform binary search in the CRT version of the computation.
Unfortunately, this randomization technique doubles the expected running time of the computation.
3. Fault Attacks.
Fault attacks try to introduce errors into the cryptographic computation, and to identify the key by analyzing the mathematical and statistical properties of the erroneously computed results. - Among the many techniques suggested so far for 'introducing such errors are the use of ionizing radiation, unusual operating temperatures, power and clock glitches, and laser-based chip microsurgery. Some of the attacks are differential (i.e., they carry out both a correct and an erroneous computation with the same input and analyze their differences), while other attacks just use the erroneous results.
The original fault attack on public key cryptosystems was described in Boneh, Demillo and Lipton, and required several cryptographic computations. We now describe an improved version of this attack, due to Arjen Lenstra, which requires a single faulty computation. We assume that the black box uses the RSA scheme to sign a given message x. The computation of x~ d (mod n) is carried out with the CRT method by first reducing x modulo p and q to get x 1 and x 2, then computing y_1=x 1~ d (mod p) and y_2=x 2~ d (mod q), and finally combining y_1 and y_2 to get the signature y (mod n) with the CRT method. We assume that a single error is introduced at a random time during this computation by applying mild physical stress to the black box.
Without loss of generality, we can assume that the error was introduced during the computation of x 1~ d (mod p), and thus instead of getting the correct y_I, the box computed an erroneous y'_l. When y'_1 and y_2 are combined by the CRT method, the box computes an incorrect signature y' which is provided to the attacker.
The main observation is that the attacker knows the signature verification exponent e, for which y~ e=x (mod n). Due to the error, y' ~ e-x is non-zero mod p, but zero mod q, and thus it is a multiple of q which is not a multiple of n. The attacker can T_ ~
thus factor n by computing the greatest common divisor of y' ~ e-x (mod n) and n, which is an easy computation.
To protect cryptographic schemes against fault attacks, Boneh, Demillo and Lipton recommend that each computation should be carried out twice (preferably by different algorithms). If any discrepancy is found between the two results, the box should not output anything. This provides strong protection from random faults (which are unlikely to affect the two computations in an identical way), but it slows down the computation by a factor of 2. Such a slowdown is particularly noticeable in smart card implementations of public key schemes, which are quite slow to begin with.
SUMMARY OF THE IIWENTION
The present invention relates to method and apparatus for providing protective techniques for public key schemes, which provide strong protection against the described timing and fault attacks without incurring the twofold slowdown made necessary by the previously known protective techniques.
The first technique is designed to protect non-CRT implementations of public key schemes against timing attacks. It is applicable to the RSA
cryptosystem, RSA digital signature scheme, Diffie-Hellman key distribution scheme, and any other number theoretic scheme in which the black box raises a known input x to a fixed secret exponent d modulo a public n whose factorization is known to the black box.
The second technique is designed to protect CRT-based implementations of public key schemes from both timing and fault attacks. The main problem is how to verify the correctness of the computations of x 1~ d (mod p) and x 2~ d (mod q) without repeating them a second time (or verifying each step separately, which again doubles the running time). The invention provides a novel error-detection technique for such number theoretic computations which is much more efficient.
Accordingly, in one aspect of the present invention there is provided in a method of implementing public key schemes containing a non-CRT form of the modular exponentiation operation x"d (mod n), for the purpose of making them more resistant to timing attacks, the improvement comprising the steps of:
a. computing or storing the computed value of t=phi(n), where phi is Euler's totient function of the modulus n;
b. selecting some secret integer i; and c. replacing the computation of x~d (mod n) by the computation of x~(d +
i*t) (mod n).
According to another aspect of the present invention there is provided in a method of implementing public key schemes containing a CRT form of the modular exponentiation operation x"d (mod n) where n=p*q, for the purpose of making them more resistant to timing and fault attacks, the improvement comprising the steps of:
a. selecting some secret integer j;
b. computing v_l=x (mod j*p), v_2=x (mod j*q), d_1=d (mod phi(j*p)), d 2=d (mod phi(j*q)), w_1=v_1~d_1 (mod j*p), and w_2=v_2~d_2 (mod j*q);
c. aborting the computation if w_1 and w 2 are not equal modulo j; and d. otherwise, computing y l=w_1 (mod p), y_2=w_2 (mod q), and combining them by the Chinese Remainder Theorem to obtain the result of x~d (mod n).
According to yet another aspect of the present invention there is provided in a method of implementing public key schemes containing a CRT form of the modular exponentiation operation x"d (mod n) where n=p*q, for the purpose of making them more resistant to timing and fault attacks, the improvement comprising the steps of:
a. selecting two secret integers j_1 and j_2;
b. computing v_1=x (mod j_1*p), v_2=x (mod j_2*q), d_1=d (mod phi(j_1*p)), d_2=d (mod phi(j_2*q)), w_1=v_1~d_1 (mod j_1*p), and w_2=v_2~d_2 (mod j_2*q);
c. computing v_3=x (mod j_l), v_4=x (mod j_2), d_3=d (mod j_1), d 4=d (mod j_2), w_3=v_3~d 3(mod j_1), and w 4=v 4~d 4(mod j_2);
d. aborting the computation if w 3 is not equal to w_1 modulo j_l, or if w_4 is not equal to w_2 modulo j_2; and e. otherwise, computing y 1=w_1 (mod p), y_2=w_2 (mod q), and combining them by the Chinese Remainder Theorem to obtain the result of x~d (mod n).
According to still yet another aspect of the present invention there is provided in an apparatus for implementing public key schemes containing a non-CRT form of the modular exponentiation operation x~d (mod n), for the purpose of making them more resistant to timing attacks, the improvement comprising:
a. means for computing or storing the computed value of t=phi(n), where phi is Euler's totient function of the modulus n;
b. means for selecting some secret integer i; and c. means for replacing the computation of x~d (mod n) by the computation of x~(d+i*t) (mod n).
10a According to still yet another aspect of the present invention there is provided in an apparatus for implementing public key schemes containing a CRT form of the modular exponentiation operation x~d (mod n) where n=p*q, for the purpose of making them more resistant to timing and fault attacks, the improvement comprising:
a. means for selecting some secret integer j;
b. means for computing v_1=x (mod j*p), v_2=x (mod j*q), d_1=d (mod phi(j *p)), d_2=d (mod phi(j *q)), w 1=v 1~d 1(mod j*p), and w_2=v 2~d 2(mod j*q);
c. means for aborting the computation if w 1 and w_2 are not equal modulo j; and d. otherwise, means for computing y-1=w_1 (mod p), y_2=w_2 (mod q), and combining them by the Chinese Remainder Theorem to obtain the result of x~d (mod n).
10b According to still yet another aspect of the present invention there is provided in an apparatus for implementing public key schemes containing a CRT form of the modular exponentiation operation x~d (mod n) where n=p*q, for the purpose of making them more resistant to timing and fault attacks, the improvement comprising:
a. selecting two secret integers j_1 and j_2;
b. computing v_1=x (mod j_1*p), v_2=x (mod j_2*q), d_1=d (mod phi(j_1 *p)), d_2=d (mod phi(j 2*q)), w 1=v l~d l(mod j_1 *p), and w_2=v_2~d 2 (mod j_2*q);
c. computing v_3=x (mod j_1), v_4=x (mod j_2), d_3=d (mod j_1), d_4=d (mod j_2), w_3=v_3~d_3 (mod j_1), and w_4=v_4~d 4(mod j_2);
d. aborting the computation if w 3 is not equal to w_1 modulo j_1, or if w 4 is not equal to w_2 modulo j_2; and e. otherwise, computing y_l=w_1 (mod p), y_2=w_2 (mod q), and combining them by the Chinese Remainder Theorem to obtain the result of x~d (mod n).
lOc BRIEF DESCRIPTION OF THE FIGURES
Embodiments of the present invention will now be described more fully with reference to the accompanying drawings in which:
Fig. 1 shows schematically the method and apparatus of the invention as it relates to a first technique designed to protect non-CRT implementations of a public key scheme against timing attacks.
Fig. 2 shows schematically the method and apparatus of the invention as it relates to a second technique designed to protect CRT-based implementations of a public key scheme against both timing and fault attacks.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
Referring now to the drawings, preferred embodiments will now be described in detail. Since public key schemes and computer hardware and software implementation are well known to those of skill in the art, no description of same is deemed necessary to a full, concise and exact understanding of the present invention.
We now describe two novel protective techniques for public key schemes, which provide strong protection against the described timing and fault attacks without incurring the twofold slowdown made necessary by the previously known protective techniques.
The first technique, shown in Fig. 1, is designed to protect non-CRT
implementations of public key schemes against timing attacks. It is applicable to the RSA cryptosystem, RSA digital signature scheme, Diffie-Hellman key distribution scheme, and any other number theoretic scheme in which the black box 10 raises a 10d known input x to a fixed secret exponent d modulo a public n whose factorization is known to the black box.
The main observation is that for each n there exists a number t=phi(n), box 12, such that for any x between 1 and n which is relatively prime to n, x~
t = 1 (mod n). This phi is called Euler's totient function: when n is prime, phi(n)=n-1, and when n=p*q,.phi(n)=(p-1)*(q-1). The implementation can thus replace the computation of x~ d(mod n) by the computation of x~(d+i*t) (mod n) for any integer i without changing the computed result, since x ~ (d+it) = (x ~ d) *(x ~ t) ~ i= (x ~ d) *1 ~ i =x ~ d (mod n), box 14.
This equality can be shown to be true even when x is not relatively prime to n, but this case is unlikely to arise in practice. Note also that any integral multiple of GCD((p-1), (q-1)) can replace phi(n) in our choice of t.
The proposed protection technique for public key schemes based on the modular exponentiation operation is to choose a new random secret i in each computation, and to replace the computation of x d (mod n) by the computation of x~ (d+i*t) (mod n) where t is the precomputed value of phi(n). Since the bits of these (d+i*t) for the various i are different, each exponentiation uses a different sequence of square-and-multiply steps, and thus the attacker cannot use Kocher's timing attack to analyze the timing distribution of several exponentiations, even though all of them compute the same d-th power of their inputs.
The efficiency of this technique is based on the fact that t is always smaller than n, and the randomizing element i can be chosen as a relatively small number. If n and d are 1024 bit numbers, and i is a random 32 bit number (which are the currently recommended sizes), d+i*t is a 1056 bit number. The process of raising the input x to the 1056 bit power d+i*t requires only 3% more square-and-multiply operations than the process of raising x to the original 1024 bit power d. This is much better than the alternative randoniization technique described earlier, which doubles the running time.
The second technique, shown in Fig. 2, is designed to protect CRT-based implementations of public key schemes from both timing and fault attacks. The main problem is how to verify t h e correctness of the computations of x 1~ d (mod p) and x 2~ d (mod q) without repeating them a second time (or verifying each step separately, which again doubles the running time). We now descnbe a novel error-detection technique for such number theoretic computations which is much more efficient.
In each computation, the black box 20 chooses a new random integer j (the recommended size of j is also 32 bit), box 22. Instead of computing x 1=x (mod p) and x 2=x (mod q) followed by y_1=x 1~ d (mod p) and y_2=x 2~ d (mod q), the box computes v_1=x (mod j*p), v_2=x (mod j*q), d_1=d (mod phi(j*p)), and d 2=d (mod phi(j*q)), box 24, followed by w 1=v_1 ~ d 1(mod j*p) and w_2=v 2~ d_2 (mod j*q), box26.
The main observation is that from w 1 and w 2 it is easy to derive y_1 and y_2 by further reductions (namely, y_1=w =w-1 (p) and y_2=w_2 (mod q)), box 28, and thus it is easy to compute the final result y by the Chinese remainder Theorem, box 30. However, we can also obtain the value of x~ d (mod j) in two different ways: as w 1(mod j) and as w 2(mod j), box 32. We can now use the equality of these two values (which were obtained from the two halves of the computation, and mixed with the 'derivation of y_l, y_2 in a very strong way), box 34, as a test of correctness: In a faultless computation the two values will always be the same, whereas in a faulty computation (with random faults) the probability that the two values will be the same is about 1/(2 ~ 32), see decision box 36 in which an abort is commanded for a faulty computation. This error detection technique is thus sufficient for any application in which the total number of modular exponentiations is significantly smaller than 2 32 (about 4 billion).
The overall time complexity of this implementation is higher by a few percent than the time complexity of standard implementations, since the exponentiations are carried out modulo 512+32=544 bit moduli j*p and j*q instead of 512 bit moduli p and q. However, this is much faster than repeating each exponentiation a second time to verify its correctness.
An additional benefit of this randomization technique is that it also provides protection from timing attacks at no extra cost. Kocher's original attack on CRT-based implementations concentrates on the initial modular reduction (mod p), and uses binary search to find increasingly accurate approximations of p from multiple computations.
By using our proposed technique, each computation uses a different modulus j*p in its initial reduction step, and thus the attacker cannot refine his knowledge of the modulus by analyzing a large number of exponentiations.
There are many optimizations and variations of this technique, which ~ CA 02288837 1999-11-08 should be obvious to anyone skilled in the art. For example, it is possible to impose additional restrictions (such as primality) on the choice of the small multiplier j which make it somewhat less likely that faulty computations will remain undetected.
Another modification of the technique is to test each half of the computation by a separate recomputation modulo a different small modulus, rather than by comparing the two results modulo a common small modulus. More precisely, the implementation can choose two small numbers j_1 and j 2, and then compute the foDowing quantities:
v 1=x (mod j_1*p), v -2=x (mod j_2*q), v -3=x (mod j_1), v-4=x (mod j_2) ;
d1=d (mod phi(j_1*p)), d_2=d (mod phi(j_2*q)), d_3=d (mod phi(j_1)), d_4=d (mod phi(j_2);
w_1=v_1 ~ d_1 (mod j_1 *p), w 2=v_2 ~ d_2 (mod j_2*P), w_3 =v_3 ~ d_3 (mod j_1), w 4=v_4 ~ d_4 (mod j_2).
To check the correctness of the computation, the black box verifies that w 1=w 3(mod j_i) and w 2=w 4(mod j 2). The only expensive operations are the computation of w 1 and w 2, since the small exponentiations in the computation of w 3 and w 4 are very efficient. This recomputation technique is slower than the original comparison technique, but may be slightly more resistant to certain types of non-random faults.
While the invention has been described with respect to certain embodiments thereof, it will be appreciated by one skilled in the art that variations and modifications may be made without departing from the spirit and scope of the invention.
_--- __ T
BIBLIOGRAPHY
1. E. Biham and A. Shamir, "A New Cryptanalytic Attack on DES", technical report, 10/18/96. An Extended version appears in the Proceedings of Crypto 97;
August 1997.
2. D. Boneh, R.A. Demillo and R.J. Lipton, "On the Importance of Checidng Computations", technical report, 9/25/96. An Extended version appears in the Proceedings of Eurocrypt 97, May 1997.
3. P. Kocher, "Cryptanalysis of Diffie-Hellman, RSA, DSS, and Other Systems Using Timing Attacks", technical report, 1217/95.
Claims (22)
1. In a method of implementing public key schemes containing a non-CRT form of the modular exponentiation operation x~d (mod n), for the purpose of making them more resistant to timing attacks, the improvement comprising the steps of:
a. computing or storing the computed value of t=phi(n), where phi is Euler's totient function of the modulus n;
b. selecting some secret integer i; and c. replacing the computation of x~d (mod n) by the computation of x~(d +
i*t) (mod n).
a. computing or storing the computed value of t=phi(n), where phi is Euler's totient function of the modulus n;
b. selecting some secret integer i; and c. replacing the computation of x~d (mod n) by the computation of x~(d +
i*t) (mod n).
2. In the method of claim 1, the improvement where i is chosen as a random number in the range [0,(2~k)-1] for some k.
3. In the method of claim 2, the improvement where k=32.
4. In a method of implementing public key schemes containing a CRT form of the modular exponentiation operation x~d (mod n) where n=p*q, for the purpose of making them more resistant to timing and fault attacks, the improvement comprising the steps of a. selecting some secret integer j;
b. computing v_1=x (mod j*p), v_2=x (mod j*q), d_1=d (mod phi(j*p)), d_2=d (mod phi(j*q)), w_1=v_1~d_1 (mod j*p), and w_2=v_2~d_2 (mod j*q);
c. aborting the computation if w_1 and w_2 are not equal modulo j; and d. otherwise, computing y_1=w_1 (mod p), y 2=w_2 (mod q), and combining them by the Chinese Remainder Theorem to obtain the result of x~d (mod n).
b. computing v_1=x (mod j*p), v_2=x (mod j*q), d_1=d (mod phi(j*p)), d_2=d (mod phi(j*q)), w_1=v_1~d_1 (mod j*p), and w_2=v_2~d_2 (mod j*q);
c. aborting the computation if w_1 and w_2 are not equal modulo j; and d. otherwise, computing y_1=w_1 (mod p), y 2=w_2 (mod q), and combining them by the Chinese Remainder Theorem to obtain the result of x~d (mod n).
5. In the method of claim 4, the improvement where j is chosen as a prime number.
6. In the method of claim 4, the improvement where j is chosen as a random number in the range [0,(2~k)-11 for some k.
7. In the method of claim 6, the improvement where k=32.
8. In a method of implementing public key schemes containing a CRT form of the modular exponentiation operation x~d (mod n) where n=p*q, for the purpose of making them more resistant to timing and fault attacks, the improvement comprising the steps of:
a. selecting two secret integers j_1 and j_2;
b. computing v_1=x (mod j_1 *p), v_2=x (mod j_2*q), d_1=d (mod phi(j_1*p)), d_2=d (mod phi(j_2*q)), w_1=v 1~d 1(mod j_1*p), and w_2=v 2~d 2 (mod j_2*q);
c. computing v_3=x (mod j_1), v_4=x (mod j_2), d_3=d (mod j_1), d 4=d (mod j_2), w_3=v_3~d 3(mod j_1), and w_4=v_4~d_4 (mod j_2);
d. aborting the computation if w_3 is not equal to w_1 modulo j_1, or if w 4 is not equal to w_2 modulo j_2; and e. otherwise, computing y_1=w_1 (mod p), y_2=w_2 (mod q), and combining them by the Chinese Remainder Theorem to obtain the result of x~d (mod n).
a. selecting two secret integers j_1 and j_2;
b. computing v_1=x (mod j_1 *p), v_2=x (mod j_2*q), d_1=d (mod phi(j_1*p)), d_2=d (mod phi(j_2*q)), w_1=v 1~d 1(mod j_1*p), and w_2=v 2~d 2 (mod j_2*q);
c. computing v_3=x (mod j_1), v_4=x (mod j_2), d_3=d (mod j_1), d 4=d (mod j_2), w_3=v_3~d 3(mod j_1), and w_4=v_4~d_4 (mod j_2);
d. aborting the computation if w_3 is not equal to w_1 modulo j_1, or if w 4 is not equal to w_2 modulo j_2; and e. otherwise, computing y_1=w_1 (mod p), y_2=w_2 (mod q), and combining them by the Chinese Remainder Theorem to obtain the result of x~d (mod n).
9. In the method of claim 8, the improvement where j_1 and j_2 are prime numbers.
10. In the method of claim 8, the improvement where j_1 and j_2 are chosen as random numbers in the range [0,(2~k)-1] for some k.
11. In the method of claim 10, the improvement where k=32.
12. In an apparatus for implementing public key schemes containing a non-CRT
form of the modular exponentiation operation x"d (mod n), for the purpose of making them more resistant to timing attacks, the improvement comprising:
a. means for computing or storing the computed value of t=phi(n), where phi is Euler's totient function of the modulus n;
b. means for selecting some secret integer i; and c. means for replacing the computation of x~d (mod n) by the computation of x~(d+i*t) (mod n).
form of the modular exponentiation operation x"d (mod n), for the purpose of making them more resistant to timing attacks, the improvement comprising:
a. means for computing or storing the computed value of t=phi(n), where phi is Euler's totient function of the modulus n;
b. means for selecting some secret integer i; and c. means for replacing the computation of x~d (mod n) by the computation of x~(d+i*t) (mod n).
13. In an apparatus according to claim 12, the improvement where i is chosen as a random number in the range [0,(2~k)-1] for some k.
14. In an apparatus according to claim 13, the improvement where k=32.
15. In an apparatus for implementing public key schemes containing a CRT form of the modular exponentiation operation x~d (mod n) where n=p*q, for the purpose of making them more resistant to timing and fault attacks, the improvement comprising:
a. means for selecting some secret integer j;
b. means for computing v_1=x (mod j*p), v_2=x (mod j*q), d_1=d (mod phi(j*p)), d_2=d (mod phi(j*q)), w 1=v 1~d 1(mod j*p), and w_2=v 2~d 2(mod j*q);
c. means for aborting the computation if w_1 and w_2 are not equal modulo j; and d. otherwise, means for computing y-1=w_1 (mod p), y_2=w_2 (mod q), and combining them by the Chinese Remainder Theorem to obtain the result of x~d (mod n).
a. means for selecting some secret integer j;
b. means for computing v_1=x (mod j*p), v_2=x (mod j*q), d_1=d (mod phi(j*p)), d_2=d (mod phi(j*q)), w 1=v 1~d 1(mod j*p), and w_2=v 2~d 2(mod j*q);
c. means for aborting the computation if w_1 and w_2 are not equal modulo j; and d. otherwise, means for computing y-1=w_1 (mod p), y_2=w_2 (mod q), and combining them by the Chinese Remainder Theorem to obtain the result of x~d (mod n).
16. In an apparatus according to claim 15, the improvement where j is chosen as a prime number.
17. In an apparatus according to claim 15, the improvement where j is chosen as a random number in the range [0,(2 ~ k)-1] for some k.
18. In an apparatus according to claim 17, the improvement where k=32.
19. In an apparatus for implementing public key schemes containing a CRT form of the modular exponentiation operation x~d (mod n) where n=p*q, for the purpose of making them more resistant to timing and fault attacks, the improvement comprising:
a. selecting two secret integers j_1 and j_2;
b. computing v_1=x (mod j_1*p), v_2=x (mod j_2*q), d_1=d (mod phi(j_1 *p)), d_2=d (mod phi(j 2*q)), w_1=v 1~d 1(mod j_1 *p), and w_2=v_2~d 2 (mod j_2*q);
c. computing v_3=x (mod j_1), v_4=x (mod j_2), d_3=d (mod j_1), d_4=d (mod j_2), w_3=v 3~d d-3 (mod j_1), and w 4=v 4~d 4 (mod j_2);
d. aborting the computation if w 3 is not equal to w_1 modulo j_l, or if w 4 is not equal to w_2 modulo j_2; and e. otherwise, computing y_1=w_1 (mod p), y_2=w_2 (mod q), and combining them by the Chinese Remainder Theorem to obtain the result of x~d (mod n).
a. selecting two secret integers j_1 and j_2;
b. computing v_1=x (mod j_1*p), v_2=x (mod j_2*q), d_1=d (mod phi(j_1 *p)), d_2=d (mod phi(j 2*q)), w_1=v 1~d 1(mod j_1 *p), and w_2=v_2~d 2 (mod j_2*q);
c. computing v_3=x (mod j_1), v_4=x (mod j_2), d_3=d (mod j_1), d_4=d (mod j_2), w_3=v 3~d d-3 (mod j_1), and w 4=v 4~d 4 (mod j_2);
d. aborting the computation if w 3 is not equal to w_1 modulo j_l, or if w 4 is not equal to w_2 modulo j_2; and e. otherwise, computing y_1=w_1 (mod p), y_2=w_2 (mod q), and combining them by the Chinese Remainder Theorem to obtain the result of x~d (mod n).
20. In an apparatus according to claim 19, the improvement where j_1 and j_2 are prime numbers.
21. In an apparatus according to claim 19, the improvement where j_1 and j_2 are chosen as random numbers in the range [0,(2~k)-11 for some k.
22. In an apparatus according to claim 21, the improvement where k=32.
Applications Claiming Priority (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US08/854,464 | 1997-05-12 | ||
| US08/854,464 US5991415A (en) | 1997-05-12 | 1997-05-12 | Method and apparatus for protecting public key schemes from timing and fault attacks |
| PCT/US1998/009593 WO1998052319A1 (en) | 1997-05-12 | 1998-05-12 | Improved method and apparatus for protecting public key schemes from timing and fault attacks |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CA2288837A1 CA2288837A1 (en) | 1998-11-19 |
| CA2288837C true CA2288837C (en) | 2008-11-18 |
Family
ID=25318766
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CA002288837A Expired - Fee Related CA2288837C (en) | 1997-05-12 | 1998-05-12 | Improved method and apparatus for protecting public key schemes from timing and fault attacks |
Country Status (8)
| Country | Link |
|---|---|
| US (1) | US5991415A (en) |
| EP (1) | EP0986873B1 (en) |
| AU (1) | AU7568598A (en) |
| CA (1) | CA2288837C (en) |
| DE (1) | DE69828787T2 (en) |
| ES (1) | ES2236903T3 (en) |
| IL (1) | IL132816A (en) |
| WO (1) | WO1998052319A1 (en) |
Families Citing this family (106)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6965673B1 (en) * | 1997-09-19 | 2005-11-15 | Telcordia Technologies, Inc. | Method of using transient faults to verify the security of a cryptosystem |
| DE69840959D1 (en) * | 1997-12-17 | 2009-08-20 | Nippon Telegraph & Telephone | Encryption and decryption devices for public key cryptosystems and recording media having associated processing programs thereon. |
| US7587044B2 (en) | 1998-01-02 | 2009-09-08 | Cryptography Research, Inc. | Differential power analysis method and apparatus |
| ATE429748T1 (en) | 1998-01-02 | 2009-05-15 | Cryptography Res Inc | LEAK RESISTANT CRYPTOGRAPHIC METHOD AND APPARATUS |
| DE19811175A1 (en) * | 1998-03-14 | 1999-09-16 | Philips Patentverwaltung | Process and arrangement for the defense against cryptanalytical investigations |
| FR2776445A1 (en) * | 1998-03-17 | 1999-09-24 | Schlumberger Ind Sa | Cryptographic algorithm security technique |
| US6214372B1 (en) | 1998-05-04 | 2001-04-10 | Con Lin Co., Inc. | Method of using isomer enriched conjugated linoleic acid compositions |
| CA2332350C (en) | 1998-05-18 | 2015-07-07 | Giesecke & Devrient Gmbh | Access-protected data carrier |
| US6144740A (en) * | 1998-05-20 | 2000-11-07 | Network Security Technology Co. | Method for designing public key cryptosystems against fault-based attacks with an implementation |
| US20030118190A1 (en) * | 1998-05-29 | 2003-06-26 | Siemens Aktiengesellschaft | Method and apparatus for processing data where a part of the current supplied is supplied to an auxiliary circuit |
| CA2333095C (en) | 1998-06-03 | 2005-05-10 | Cryptography Research, Inc. | Improved des and other cryptographic processes with leak minimization for smartcards and other cryptosystems |
| ATE360866T1 (en) | 1998-07-02 | 2007-05-15 | Cryptography Res Inc | LEAK-RESISTANT UPDATING OF AN INDEXED CRYPTOGRAPHIC KEY |
| CA2243761C (en) * | 1998-07-21 | 2009-10-06 | Certicom Corp. | Timing attack resistant cryptographic system |
| CA2259089C (en) * | 1999-01-15 | 2013-03-12 | Robert J. Lambert | Method and apparatus for masking cryptographic operations |
| CA2258338C (en) * | 1999-01-11 | 2009-02-24 | Certicom Corp. | Method and apparatus for minimizing differential power attacks on processors |
| US7092523B2 (en) | 1999-01-11 | 2006-08-15 | Certicom Corp. | Method and apparatus for minimizing differential power attacks on processors |
| US7599491B2 (en) | 1999-01-11 | 2009-10-06 | Certicom Corp. | Method for strengthening the implementation of ECDSA against power analysis |
| FR2789535B1 (en) | 1999-02-04 | 2001-09-28 | Bull Cp8 | METHOD FOR SECURING AN ELECTRONIC ASSEMBLY OF SECRET KEY CRYPTOGRAPHY AGAINST ATTACKS BY PHYSICAL ANALYSIS |
| FR2792141B1 (en) | 1999-04-09 | 2001-06-15 | Bull Cp8 | METHOD FOR SECURING ONE OR MORE ELECTRONIC ASSEMBLIES IMPLEMENTING THE SAME CRYPTOGRAPHIC ALGORITHM WITH SECRET KEY, A USE OF THE METHOD AND THE ELECTRONIC ASSEMBLY |
| US6298135B1 (en) * | 1999-04-29 | 2001-10-02 | Motorola, Inc. | Method of preventing power analysis attacks on microelectronic assemblies |
| DE19921633A1 (en) * | 1999-05-10 | 2000-11-16 | Deutsche Telekom Ag | Procedure for implementing cryptographic algorithms |
| US6804782B1 (en) * | 1999-06-11 | 2004-10-12 | General Instrument Corporation | Countermeasure to power attack and timing attack on cryptographic operations |
| NL1012581C2 (en) * | 1999-07-13 | 2001-01-16 | Koninkl Kpn Nv | Method of protecting a portable card. |
| FR2799851B1 (en) * | 1999-10-14 | 2002-01-25 | Gemplus Card Int | COUNTER-MEASUREMENT METHOD IN AN ELECTRONIC COMPONENT USING A RSA-TYPE PUBLIC KEY CRYPTOGRAPHY ALGORITHM |
| FR2800478B1 (en) * | 1999-10-28 | 2001-11-30 | Bull Cp8 | METHOD FOR SECURING AN ELECTRONIC CRYPTOGRAPHY ASSEMBLY BASED ON MODULAR EXPONENTIATION AGAINST ATTACKS BY PHYSICAL ANALYSIS |
| US6724894B1 (en) * | 1999-11-05 | 2004-04-20 | Pitney Bowes Inc. | Cryptographic device having reduced vulnerability to side-channel attack and method of operating same |
| DE19963407A1 (en) * | 1999-12-28 | 2001-07-12 | Giesecke & Devrient Gmbh | Portable data carrier with access protection through message alienation |
| DE19963408A1 (en) * | 1999-12-28 | 2001-08-30 | Giesecke & Devrient Gmbh | Portable data carrier with access protection by key division |
| DE10024325B4 (en) * | 2000-05-17 | 2005-12-15 | Giesecke & Devrient Gmbh | Cryptographic method and cryptographic device |
| FR2810178B1 (en) * | 2000-06-13 | 2004-10-29 | Gemplus Card Int | CRYPTOGRAPHIC CALCULATION PROCESS INCLUDING A MODULAR EXPONENTIATION ROUTINE |
| DE10042234C2 (en) * | 2000-08-28 | 2002-06-20 | Infineon Technologies Ag | Method and device for performing a modular exponentiation in a cryptographic processor |
| KR100396740B1 (en) * | 2000-10-17 | 2003-09-02 | 학교법인 한국정보통신학원 | Provably secure public key encryption scheme based on computational diffie-hellman assumption |
| FR2818772A1 (en) * | 2000-12-21 | 2002-06-28 | Bull Cp8 | METHOD OF SECURING A LOGIC OR MATHEMATICAL OPERATOR IMPLANTED IN A MICROPROCESSOR ELECTRONIC MODULE, AND THE ASSOCIATED ELECTRONIC MODULE AND THE ON-LINE SYSTEM |
| FR2819663B1 (en) * | 2001-01-18 | 2003-04-11 | Gemplus Card Int | DEVICE AND METHOD FOR EXECUTING A CRYPTOGRAPHIC ALGORITHM |
| FR2824209B1 (en) * | 2001-04-30 | 2003-08-29 | St Microelectronics Sa | INTERFERENCE OF A CALCULATION IMPLEMENTING A MODULAR FUNCTION |
| FR2828608B1 (en) * | 2001-08-10 | 2004-03-05 | Gemplus Card Int | SECURE PROCESS FOR PERFORMING A MODULAR EXPONENTIATION OPERATION |
| DE10143728B4 (en) * | 2001-09-06 | 2004-09-02 | Infineon Technologies Ag | Device and method for calculating a result of a modular exponentiation |
| FR2829335A1 (en) * | 2001-09-06 | 2003-03-07 | St Microelectronics Sa | METHOD FOR INTERFERING A QUANTITY SECRET CALCULATION |
| US7436953B2 (en) * | 2001-09-13 | 2008-10-14 | Nds Limited | Hacking prevention system |
| FR2830146B1 (en) * | 2001-09-24 | 2003-10-31 | Gemplus Card Int | METHOD FOR IMPLEMENTING, IN AN ELECTRONIC COMPONENT, A CRYPTOGRAPHIC ALGORITHM AND CORRESPONDING COMPONENT |
| WO2003034649A2 (en) * | 2001-10-17 | 2003-04-24 | Infineon Technologies Ag | Method and device for guaranteeing a calculation in a cryptographic algorithm |
| WO2003034268A2 (en) * | 2001-10-17 | 2003-04-24 | Infineon Technologies Ag | Method and device for securing an exponentiation calculation by means of the chinese remainder theorem (crt) |
| DE10162584A1 (en) * | 2001-10-17 | 2003-05-08 | Infineon Technologies Ag | Method for validating an exponentiation result with the Chinese remainder theorem forms extra modules with two primary numbers for calculating extra values to work out a modular exponentiation to match the product of the values. |
| GB0126317D0 (en) * | 2001-11-02 | 2002-01-02 | Comodo Res Lab Ltd | Improvements in and relating to cryptographic methods and apparatus in which an exponentiation is used |
| WO2003069841A1 (en) * | 2001-12-28 | 2003-08-21 | Gemplus | Method for detection of attacks on cryptographic algorithms by trial and error |
| KR100431286B1 (en) * | 2002-01-14 | 2004-05-12 | 한국정보보호진흥원 | Method for preventing the CRT-based fault attack and apparatus thereof |
| JP2003241659A (en) * | 2002-02-22 | 2003-08-29 | Hitachi Ltd | Information processing method |
| KR100431047B1 (en) * | 2002-02-26 | 2004-05-12 | 주홍정보통신주식회사 | Digital signature method using RSA public-key cryptographic based on CRT and apparatus therefor |
| FR2838262B1 (en) | 2002-04-08 | 2004-07-30 | Oberthur Card Syst Sa | METHOD FOR SECURING ELECTRONICS WITH ENCRYPTED ACCESS |
| FR2840083A1 (en) * | 2002-05-24 | 2003-11-28 | St Microelectronics Sa | TEST OF AN ALGORITHM EXECUTED BY AN INTEGRATED CIRCUIT |
| DE10250810A1 (en) * | 2002-09-11 | 2004-03-25 | Giesecke & Devrient Gmbh | Cryptographic computation method for running protected computation in smart cards produces a code with two or more code parameters |
| CN1682484B (en) * | 2002-09-11 | 2012-03-21 | 德国捷德有限公司 | Protected cryptographic calculation |
| FR2844891A1 (en) * | 2002-09-20 | 2004-03-26 | St Microelectronics Sa | Microprocessor handled digital data masking method involves factorizing digital data by residue number system based on finite base of numbers or polynomials prime to one another |
| DE10304451B3 (en) * | 2003-02-04 | 2004-09-02 | Infineon Technologies Ag | Modular exponentiation with randomized exponent |
| GB2399904B (en) * | 2003-03-28 | 2005-08-17 | Sharp Kk | Side channel attack prevention in data processing apparatus |
| US6880752B2 (en) * | 2003-04-16 | 2005-04-19 | George V. Tarnovsky | System for testing, verifying legitimacy of smart card in-situ and for storing data therein |
| DE10328860B4 (en) * | 2003-06-26 | 2008-08-07 | Infineon Technologies Ag | Device and method for encrypting data |
| FR2858496B1 (en) | 2003-07-31 | 2005-09-30 | Gemplus Card Int | METHOD FOR SECURELY IMPLEMENTING AN RSA-TYPE CRYPTOGRAPHY ALGORITHM AND CORRESPONDING COMPONENT |
| DE602004027943D1 (en) | 2003-11-16 | 2010-08-12 | Sandisk Il Ltd | IMPROVED NATURAL MONTGOMERY EXPONENT MASKING |
| US7590880B1 (en) * | 2004-09-13 | 2009-09-15 | National Semiconductor Corporation | Circuitry and method for detecting and protecting against over-clocking attacks |
| US7477741B1 (en) | 2004-10-01 | 2009-01-13 | The United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration | Analysis resistant cipher method and apparatus |
| CA2592875C (en) | 2005-01-18 | 2016-09-06 | Certicom Corp. | Accelerated verification of digital signatures and public keys |
| US8467535B2 (en) | 2005-01-18 | 2013-06-18 | Certicom Corp. | Accelerated verification of digital signatures and public keys |
| US7620187B1 (en) | 2005-03-30 | 2009-11-17 | Rockwell Collins, Inc. | Method and apparatus for ad hoc cryptographic key transfer |
| FR2884004B1 (en) * | 2005-03-30 | 2007-06-29 | Oberthur Card Syst Sa | DATA PROCESSING METHOD INVOLVING MODULAR EXPONENTIATION AND ASSOCIATED DEVICE |
| FR2884088B1 (en) * | 2005-03-31 | 2007-06-22 | Gemplus Sa | METHOD AND CRYPTOGRAPHIC DEVICE FOR PROTECTING THE LOGIC OF PUBLIC KEYS AGAINST FAULT ATTACKS |
| US7729459B1 (en) | 2005-03-31 | 2010-06-01 | National Semiconductor Corporation | System and method for providing a robust ultra low power serial interface with digital clock and data recovery circuit for power management systems |
| FR2887351A1 (en) * | 2005-06-16 | 2006-12-22 | St Microelectronics Sa | PROTECTION OF A MODULAR EXPONENTIATION CALCULATION CARRIED OUT BY AN INTEGRATED CIRCUIT |
| EP1920324A1 (en) * | 2005-08-19 | 2008-05-14 | Nxp B.V. | Circuit arrangement for and method of performing an inversion operation in a cryptographic calculation |
| JP2009505147A (en) * | 2005-08-19 | 2009-02-05 | エヌエックスピー ビー ヴィ | Circuit apparatus and method for performing cryptographic calculation |
| US7822207B2 (en) * | 2006-12-22 | 2010-10-26 | Atmel Rousset S.A.S. | Key protection mechanism |
| US8280041B2 (en) * | 2007-03-12 | 2012-10-02 | Inside Secure | Chinese remainder theorem-based computation method for cryptosystems |
| WO2008114310A1 (en) * | 2007-03-16 | 2008-09-25 | Fujitsu Limited | Incorporating device having fault attack countermeasure function |
| US7974409B2 (en) * | 2007-06-28 | 2011-07-05 | Samsung Electronics Co., Ltd. | Changing the order of public key cryptographic computations |
| US7936871B2 (en) * | 2007-06-28 | 2011-05-03 | Samsung Electronics Co., Ltd. | Altering the size of windows in public key cryptographic computations |
| FR2919739B1 (en) | 2007-08-03 | 2009-12-04 | Oberthur Card Syst Sa | FAT GENERATION ATTACKED DATA PROCESSING METHOD AND ASSOCIATED DEVICE |
| US8670557B2 (en) | 2007-09-10 | 2014-03-11 | Spansion Llc | Cryptographic system with modular randomization of exponentiation |
| US8139763B2 (en) | 2007-10-10 | 2012-03-20 | Spansion Llc | Randomized RSA-based cryptographic exponentiation resistant to side channel and fault attacks |
| US7907724B2 (en) | 2007-10-25 | 2011-03-15 | Infineon Technologies Ag | Method and apparatus for protecting an RSA calculation on an output by means of the chinese remainder theorem |
| US8774400B2 (en) | 2008-01-03 | 2014-07-08 | Spansion Llc | Method for protecting data against differntial fault analysis involved in rivest, shamir, and adleman cryptography using the chinese remainder theorem |
| KR100953715B1 (en) | 2008-01-22 | 2010-04-19 | 고려대학교 산학협력단 | Digital signature method, Digital signature apparatus using CRT-RSA modula exponentiation algorithm and Recording medium using by the same |
| FR2926652B1 (en) * | 2008-01-23 | 2010-06-18 | Inside Contactless | COUNTER-MEASUREMENT METHOD AND DEVICES FOR ASYMMETRIC CRYPTOGRAPHY WITH SIGNATURE SCHEMA |
| US7958354B1 (en) | 2008-02-14 | 2011-06-07 | Rockwell Collins, Inc. | High-order knowledge sharing system to distribute secret data |
| US8600047B2 (en) * | 2008-05-07 | 2013-12-03 | Irdeto Corporate B.V. | Exponent obfuscation |
| US7522723B1 (en) | 2008-05-29 | 2009-04-21 | Cheman Shaik | Password self encryption method and system and encryption by keys generated from personal secret information |
| EP2154604A1 (en) | 2008-08-06 | 2010-02-17 | Gemalto SA | Countermeasure securing exponentiation based cryptography |
| EP2222013A1 (en) | 2009-02-19 | 2010-08-25 | Thomson Licensing | Method and device for countering fault attacks |
| EP2228715A1 (en) * | 2009-03-13 | 2010-09-15 | Thomson Licensing | Fault-resistant calculcations on elliptic curves |
| WO2011068996A1 (en) * | 2009-12-04 | 2011-06-09 | Cryptography Research, Inc. | Verifiable, leak-resistant encryption and decryption |
| US9300475B2 (en) * | 2010-12-24 | 2016-03-29 | Mitsubishi Electric Corporation | Signature generation by calculating a remainder modulo public information |
| US8817974B2 (en) * | 2011-05-11 | 2014-08-26 | Nxp B.V. | Finite field cryptographic arithmetic resistant to fault attacks |
| EP2535804A1 (en) | 2011-06-17 | 2012-12-19 | Thomson Licensing | Fault-resistant exponentiation algorithm |
| FR2977953A1 (en) * | 2011-07-13 | 2013-01-18 | St Microelectronics Rousset | PROTECTION OF A MODULAR EXPONENTIATION CALCULATION BY ADDING A RANDOM QUANTITY |
| DE102011115082A1 (en) | 2011-09-19 | 2013-03-21 | Giesecke & Devrient Gmbh | Against spying on protectable secret RSA encryption exponent |
| US8745376B2 (en) | 2011-10-14 | 2014-06-03 | Certicom Corp. | Verifying implicit certificates and digital signatures |
| US8334705B1 (en) | 2011-10-27 | 2012-12-18 | Certicom Corp. | Analog circuitry to conceal activity of logic circuitry |
| US8635467B2 (en) | 2011-10-27 | 2014-01-21 | Certicom Corp. | Integrated circuit with logic circuitry and multiple concealing circuits |
| US8861718B2 (en) * | 2012-02-10 | 2014-10-14 | Electronics And Telecommunications Research Institute | Method of preventing fault-injection attacks on Chinese Remainder Theorem-Rivest Shamir Adleman cryptographic operations and recording medium for storing program implementing the same |
| EP2667539A1 (en) * | 2012-05-21 | 2013-11-27 | Koninklijke Philips N.V. | Key sharing methods, device and system for configuration thereof. |
| CN103051454A (en) * | 2012-12-19 | 2013-04-17 | 天地融科技股份有限公司 | Electronic equipment and password authentication method thereof |
| US9087192B2 (en) | 2013-09-10 | 2015-07-21 | Infineon Technologies Ag | Electronic circuit and method for monitoring a data processing |
| KR102180029B1 (en) | 2014-02-03 | 2020-11-17 | 삼성전자 주식회사 | CRT-RSA encryption method and apparatus and computer readable storage medium therefore |
| EP3188401B1 (en) | 2015-12-29 | 2019-12-18 | Secure-IC SAS | Method and system for protecting a cryptographic operation |
| EP3242202A1 (en) | 2016-05-04 | 2017-11-08 | Gemalto Sa | Countermeasure to safe-error fault injection attacks on cryptographic exponentiation algorithms |
| TW201810989A (en) * | 2016-05-18 | 2018-03-16 | 納格維遜股份有限公司 | Method and device to protect a cryptographic exponent |
| EP3503459B1 (en) | 2017-12-22 | 2021-04-21 | Secure-IC SAS | Device and method for protecting execution of a cryptographic operation |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5504817A (en) * | 1994-05-09 | 1996-04-02 | Yeda Research And Development Co. Ltd. At The Weizmann Institute Of Science | Method and apparatus for memory efficient variants of public key encryption and identification schemes for smart card applications |
| US5764766A (en) * | 1996-06-11 | 1998-06-09 | Digital Equipment Corporation | System and method for generation of one-time encryption keys for data communications and a computer program product for implementing the same |
-
1997
- 1997-05-12 US US08/854,464 patent/US5991415A/en not_active Expired - Lifetime
-
1998
- 1998-05-12 WO PCT/US1998/009593 patent/WO1998052319A1/en active IP Right Grant
- 1998-05-12 AU AU75685/98A patent/AU7568598A/en not_active Abandoned
- 1998-05-12 CA CA002288837A patent/CA2288837C/en not_active Expired - Fee Related
- 1998-05-12 ES ES98923379T patent/ES2236903T3/en not_active Expired - Lifetime
- 1998-05-12 DE DE69828787T patent/DE69828787T2/en not_active Expired - Lifetime
- 1998-05-12 EP EP98923379A patent/EP0986873B1/en not_active Expired - Lifetime
- 1998-05-12 IL IL13281698A patent/IL132816A/en not_active IP Right Cessation
Also Published As
| Publication number | Publication date |
|---|---|
| ES2236903T3 (en) | 2005-07-16 |
| CA2288837A1 (en) | 1998-11-19 |
| EP0986873A1 (en) | 2000-03-22 |
| DE69828787D1 (en) | 2005-03-03 |
| IL132816A (en) | 2003-05-29 |
| WO1998052319A1 (en) | 1998-11-19 |
| EP0986873B1 (en) | 2005-01-26 |
| AU7568598A (en) | 1998-12-08 |
| US5991415A (en) | 1999-11-23 |
| IL132816A0 (en) | 2001-03-19 |
| DE69828787T2 (en) | 2006-04-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CA2288837C (en) | Improved method and apparatus for protecting public key schemes from timing and fault attacks | |
| den Boer et al. | A DPA attack against the modular reduction within a CRT implementation of RSA | |
| Giraud | An RSA implementation resistant to fault attacks and to simple power analysis | |
| Fouque et al. | The doubling attack–why upwards is better than downwards | |
| Vigilant | RSA with CRT: A new cost-effective solution to thwart fault attacks | |
| Boscher et al. | CRT RSA algorithm protected against fault attacks | |
| US10361854B2 (en) | Modular multiplication device and method | |
| Hess et al. | Information leakage attacks against smart card implementations of cryptographic algorithms and countermeasures–a survey | |
| US7860242B2 (en) | Method of securely implementing a cryptography algorithm of the RSA type, and a corresponding component | |
| EP2332040B1 (en) | Countermeasure securing exponentiation based cryptography | |
| Walter | Precise bounds for Montgomery modular multiplication and some potentially insecure RSA moduli | |
| Kim et al. | How can we overcome both side channel analysis and fault attacks on RSA-CRT? | |
| Walter | Simple power analysis of unified code for ECC double and add | |
| Fumaroli et al. | Blinded fault resistant exponentiation | |
| US7907724B2 (en) | Method and apparatus for protecting an RSA calculation on an output by means of the chinese remainder theorem | |
| KR100652377B1 (en) | A modular exponentiation algorithm, a record device including the algorithm and a system using the algorithm | |
| Fournaris et al. | Protecting CRT RSA against fault and power side channel attacks | |
| EP3503459B1 (en) | Device and method for protecting execution of a cryptographic operation | |
| US7123717B1 (en) | Countermeasure method in an electronic component which uses an RSA-type public key cryptographic algorithm | |
| EP1347596B1 (en) | Digital signature methods and apparatus | |
| Walter | Longer keys may facilitate side channel attacks | |
| KR100431286B1 (en) | Method for preventing the CRT-based fault attack and apparatus thereof | |
| Joye | Protecting RSA against fault attacks: The embedding method | |
| US11606195B2 (en) | Method of verifying integrity of a pair of cryptographic keys and cryptographic device | |
| Mahanta et al. | Modular exponentiation with inner product to resist higher-order DPA attacks |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| EEER | Examination request | ||
| MKLA | Lapsed |