Method and apparatus for encrypting and decrypting data

[0001] The present invention relates to cryptography and in particular to a method and apparatus for encrypting and decrypting digital data for the purpose of protecting or securing its contents.

[0002] There exists a need to transfer data confidentially over an open channel or to store such data securely in an unsecure location. Whilst such transfer or storage may be achieved by physical means, it is more effective and/or flexible to use cryptographic means.

[0003] In the prior art, to send private communications between two parties, the parties need to share a cryptographic key and use a symmetric-key cipher to encrypt and decrypt data. Various ciphers including block ciphers and stream ciphers have been proposed in the past. A stream cipher handles messages of arbitrary size by ciphering individual elements, such as bits or bytes of data. This avoids the need to accumulate data into a block before ciphering as is necessary in a block cipher. A conventional block cipher requires an accumulation of a certain amount of data or multiple data elements for ciphering to complete. Examples of block ciphers include DES (see ANSI X3.92, “American National Standard for Data Encryption Algorithm (DEA),” American National Standards Institute, 1981), IDEA (see X. Lai, J. Massey, and S. Murphy, “Markov ciphers and differential cryptanalysis,” Advances in Cryptology—EUROCRYPT '91 Proceedings, Springer-Verlag, 1991, pp. 17–38), SAFER (see J. Massey. SAFER K-64: One year later. In B. Preneel, editor, Fast Software Encryption—Proceedings of Second International Workshop, LNCS 1008, pages 212–241, Springer Verlag, 1995), and RC5 (see R. Rivest, “The RC5 encryption algorithm,” Dr. Dobb's Journal, Vol. 20, No. 1, January 1995, pp. 146–148). A typical data encryption speed for these ciphers is several million bits per second (Mb/s) on a Pentium 266 MHz processor.

[0004] Due to the pervasiveness of high-speed networking and multimedia communications, the demand for high-speed ciphers is ever increasing. For example, data rates over Asynchronous Data Transfer networks range from several tens of Mb/s to 1 Gb/s. Software implementations of existing block ciphers cannot reach these kinds of data rates.

[0005] In general, stream ciphers are much faster than block ciphers. However, stream ciphers are usually not sufficiently analyzed and are perceived to be weaker in security than block ciphers. Many stream ciphers that we believed to be very secure were subsequently broken. The design of secure and efficient high-speed ciphers remains a highly challenging problem.

[0006] Many powerful cryptanalytical methods have been developed during the past decade or so. It may be observed that the success of many of these methods in attacking a cipher depends on the availability of a large quantity of ciphertexts/plaintenxts under a particular encryption key. Normally, the likelihood of successfully attacking a cipher, i.e., discovering the key, diminishes as the amount of available ciphertexts/plaintexts decreases. The present invention, is motivated by the above observation, and provides an improved method and apparatus for data encryption and decryption.

[0007] The method of the present invention may employ a combination of at least two cryptographic algorithms to achieve relatively high throughput without compromizing security. A first algorithm may be a cryptographic pseudo random sequence (or number) generator with strong security, and a second algorithm may be a cipher capable of high-speed operation, but may be weak in security when used alone. The first algorithm may be used to systematically and periodically generate “segment keys” and the second algorithm may be used to encrypt a data segment or plaintext segment using a segment key. Each data segment may be encrypted using a different segment key. By limiting the sizes of the data segments, an attacker may not have sufficient plaintexts or ciphertexts under a given segment key to carry out meaningful cryptanalysis against the second algorithm. In doing so, the present invention may achieve high throughput in data encryption and decryption without compromising overall security of the system.

[0008] According to one aspect of the present invention there is provided a method of encrypting data suitable for sending to a decrypting party, said method including the steps of:

• • [0009] (a) dividing said data into data segments;
  • [0010] (b) accepting at least a cryptographic key k shared with the decrypting party;
  • [0011] (c) for the ith data segment (i=1, 2, . . . ,) to be encrypted, generating the ith segment key s_i using a first function with said cryptographic key k and some accessory data strings as inputs;
  • [0012] (d) encrypting the ith data segment using a second function with s_i as the encryption key to form the ith ciphertext segment; and
  • [0013] (e) outputting the ith ciphertext segment, and at least a part of said accessory data strings for sending data to the decrypting party, and if more data segments are to be encrypted, repeating steps (c), (d) and (e).

[0014] The accessory data strings may include a single string v_i derived from the previous value v_i-1 in a predetermined fashion. The string v_i may be derived according to the relation v_i=F(v_i-1), i=1, 2, . . . , wherein F( ) maps v_i-1 to v_i and v_o is an initialization value made known to the decrypting party.

[0015] According to a further aspect of the present invention there is provided a method of decrypting data encrypted by an encrypting party, said method including the steps of:

• • [0016] (a) accepting at least a cryptographic key k being shared with the encrypting party;
  • [0017] (b) for the ith ciphertext segment (i=1, 2, . . . ,) to be decrypted, generating the ith segment key s_i using a first function with said cryptographic key k and some accessory data strings as inputs;
  • [0018] (c) decrypting the ith ciphertext segment using a second function with s_i as the decryption key;
  • [0019] (d) outputting the decrypted ith ciphertext segment, and if more ciphertext segments are to be decrypted, repeating steps (b), (c) and (d).

[0020] According to a still further aspect of the present invention there is provided apparatus for encrypting data suitable for sending to a decrypting party, said apparatus including:

• • [0021] (a) means for dividing said data into data segments;
  • [0022] (b) means for accepting at least a cryptographic key k shared with the decrypting party;
  • [0023] (c) means for generating for the ith data segment (i=1, 2, . . . ,) to be encrypted, the ith segment key s_i using a first function with said cryptographic key k and some accessory data strings as inputs;
  • [0024] (d) means for encrypting the ith data segment using a second function with s_i as the encryption key to form the ith ciphertext segment; and
  • [0025] (e) means for outputting the ith ciphertext segment, and at least a part of said accessory data strings for sending data to the decrypting party.

[0026] According to a still further aspect of the present invention there is provided apparatus for decrypting data encrypted by an encrypting party, said apparatus including:

• • [0027] (a) means for accepting at least a cryptographic key k being shared with the encrypting party;
  • [0028] (b) means for generating as inputs for the ith ciphertext segment (i=1, 2, . . . ,) to be decrypted, the ith segment key s_i using a first function with said cryptographic key k and some accessory data strings;
  • [0029] (c) means for decrypting the ith ciphertext segment using a second function with s_i as the decryption key; and
  • [0030] (d) means for outputting the decrypted ith ciphertext segment.

[0031] The apparatus of the present invention may be conveniently embodied by means of a suitably programmed general purpose digital computer. It is well within the capability of persons skilled in the art of programming digital computers to develop software programs for implementing the encrypting/decrypting methods described herein. Alternatively the apparatus may be implemented via dedicated hardware.

[0035] FIG. 1 shows the operation of the present invention at the encrypting end of a communication channel. Data encryption is performed using two cryptographic algorithms, the first being a cryptographic pseudo random sequence generator R( ) which is a sequence generating function and the second being a high-speed cipher E( ) which is functionally a ciphering function. The high-speed cipher may be relatively weak in security when used alone. The pseudo random sequence generator accepts two inputs k and v and outputs a pseudo random sequence s=R(k, v). The high-speed cipher accepts a secret key s and a data segment d and produces the ciphertext c=E(s, d). In addition, the illustrative embodiment uses a pre-determined function F( ) to update an initial value, i.e., v_i=F(V_i-1). It is assumed that the encrypting end and decrypting ends share a secret key k, an initial value v₀, and the functions F( ) and R( ). Moreover, it is assumed that the decrypting end knows the decrypting algorithm D( ) corresponding to the encrypting algorithm E( ).

[0036] As shown in FIG. 1, at step 100, a program at the encrypting end divides the data to be encrypted into segments of equal or unequal sizes: d₁, d₂, . . . , d_i, . . . . In the former case the last segment may be padded with random data if necessary; while in the latter case, the sizes of the data segments normally need to be known by the decrypting end to facilitate decryption. Furthermore, the program accepts the shared secret key k and the shared initial value v₀ as inputs, and sets the index i=0.

[0037] At 110, the program inspects if there is any data segment available for encryption, and if not, the program terminates. Assuming that there is a data segment available, the program, at 120, increments the index i by 1, gets an updated initial value v_i using a hash function F( ) where v_i=F(v_i-1), generates a segment key s_i=R(k, v_i), and uses the segment key to encrypt the data segment to get the ciphertext segment c_i=E(s_i, d_i) in a manner that is well known to those skilled in the art.

[0038] At 130, the program transmits the ciphertext segment, and optionally the size of the corresponding data segment, to the decrypting end. The program then goes back to 110 to see if more data segments need to be encrypted. If so, the preceding process is repeated.

[0039] The function F( ) is used to update the initial value. One example is v_i=v_i-1+1 and another example is a cryptographic hash function.

[0040] Those skilled in the art will see that the shared secret key is protected by the cryptographic pseudo random generator R(k, v_i). To obtain good security, it is required that R( ) be secure against all known attacks to the key k. R( ) is preferably a secure one-way function or one-way hash function in k. That is, given R(k, v_i) and v_i, it should be computationally hard to find k. One example of a pseudo random generator is a keyed one-way hash function h(k, v_i,) or h(k, p, v_i, k) where h( ) is a one-way hash function and where p pads k to a full input block as specified by some hash functions. Examples of one-way hash functions are MD5 and SHA, (refer respectively, R. Rivest, “The MD5 message digest algorithm,” IETF RFC 1321, April 1992 and National Institute of Standards and Technology, NIST FIPS PUB 186, “Digital Signature Standard,” U.S. Department of Commerce, May 1994). Another example of a cryptographic pseudo random generator is a strong encryption algorithm such as IDEA with k as the encryption key, v_i as plaintext, and the ciphertext output as the pseudo random sequence.

[0041] In the illustrative embodiment for encryption, the segment key s_i is used by the cipher E( ) to encrypt only one data segment d_i. This implies that only the corresponding ciphertext segment c_i and in some cases part of the corresponding data segment are available to an attacker to cryptanalyze the cipher. One selection criteria for E( ) is that it should be capable of operating at a high-speed. Another selection criteria for E( ) is that given the limited amount of ciphertexts and even part of the corresponding data segment under a segment key, the cipher E( ) should be capable of resisting all known attacks. As a consequence, there is a tradeoff between the size of the data segment and system throughput; the larger the size of a data segment, the higher the throughput. On the other hand, a larger data segment implies that more ciphertexts or plaintexts under a segment key are available to an attacker to cryptanalyze the cipher E( ). Examples of E( ) are high-speed stream ciphers or block ciphers with fewer rounds of iterations than that when they are used alone. In the latter case, the notation E(s_i, d_i) represents the encryption of data segment d_i using a block cipher even when the size of the data segment d_i is larger than the block size of the underlying block cipher and the encryption may be performed in various modes, such as Electronic Code Book or Cipher Block Chaining Mode.

[0042] One specific example of E( ) is the following high-speed stream cipher. Let N( ) be a function defined as N(s,x)=((((x+s₁)⊕s₂)×s₃⊕s₄)>>>, where s=s₁s₂s₃s₄ (consisting of four 32-bit strings) is a 128 bit secret key, x is a 32-bit string, ⊕ is the bit-wise exclusive-or, + and × are mod 2³² addition and multiplication, and >>> is to reverse a 32 bit string into opposite ranking. Let b₁b₂Λ b_mΛ be the data to be encrypted which is a concatenation of 32 bit strings, the corresponding ciphertexts are given by d_i=b_i⊕N(s,N(s,N(s,d_i-1)⊕b_i-1)⊕d_i-2), where the initial values d₋₁, d₋₂, d₋₃ can be set to s₂,s₃,s₄.

[0043] Another specific example of E( ) is Serpent with a reduced number of rounds. Serpent is a block cipher with 128 bit block length, variable key lengths, and 32 rounds of operations (see R. Anderson, E. Biham, and L. Knudsen, “Serpent: A Proposal for the Advanced Encryption Standard”, http://www.ci.cam.ac.uk/˜ria14/serpent.html). Its inventors showed that to attack 6 round Serpent successfully, it would require 2⁵⁶ and 2¹¹⁶ plaintext blocks using linear and differential cryptanalysis, respectively. Hence, if a 6 round Serpent is used as E( ) to encrypt data, it should resist both linear and differential cryptanalysis as long as the data segment size is less than 2⁵⁶ 128 bit blocks. At the same time, this E( ) is about 5 times faster than the 32 round Serpent.

[0044] FIG. 2 depicts a flowchart of the operation of the present invention at the data decrypting end of a communication channel. As shown in FIG. 2, at step 200, a program at the decrypting end accepts the shared secret key k and the shared initial value v₀ as inputs, and sets the index i=0.

[0045] The program then checks at 210 to see if there is any ciphertext segment available for decryption and if not, the program halts its operation. Assuming that a ciphertext segment is received, the program, at 220, increments the index i by 1, updates the initial value v_i=F(v_i-1), computes a segment key s_i=R(k, v_i), and uses the segment key to decrypt the ciphertext segment to get the data segment d_i=D(s_i, c_i) in a fashion that is well known in the art.

[0046] As shown at 230, the program preferably outputs the data segment and then goes back to 210 to see if there is more ciphertext segment available for decryption. If so, the preceding steps are repeated.

[0047] The embodiment described above is merely one illustrative example of realizing the present invention; there can be many variants of this. For example, it is well within the capability of persons skilled in the art to suggest alternative ways of generating segment keys using a pseudo random generator, where the current segment key may depend not only on the cryptographic key k, but also on other variables such as part of the plaintext, part of the ciphertext, a time stamp, and previous segment keys.

[0048] Finally, it is to be understood that various alterations, modifications and/or additions may be introduced into the constructions and arrangements of parts previously described without departing from the spirit or ambit of the present invention.