CIS 5371 Cryptography 6. Practical Constructions of Symmetric-Key Primitives Based on: Jonathan Katz and Yehuda Lindell Introduction to Modern Cryptography 1 Stream ciphers A stream cipher is a pair of deterministic algorithms (Init, GetBits), where Init takes input a seed 𝑠 and an optional 𝐼𝑉 and outputs an initial state 𝑠𝑡0 . That is, 𝑠𝑡0 := Init (𝑠, 𝐼𝑉) GetBits takes as input 𝑠𝑡𝑖 and outputs a bit 𝑦𝑖 and an updated state 𝑠𝑡𝑖+1 . That is, (𝑦𝑖 , 𝑠𝑡𝑖 ) := GetBits(𝑠𝑡𝑖−1 ), 𝑖 = 1,2, . . . 2 Linear Feedback Shift Registers (LFSR) Linear feedback (𝑛 = 5) 𝑠𝑠34 x 𝑠𝑠33 x 𝑠𝑠32 x 𝑠𝑠31 x 𝑠𝑠30 x 𝑡 𝑠𝑖𝑡+1 := 𝑠𝑖+1 , 𝑖 = 0, … , 𝑛 − 2 𝑡 𝑡+1 𝑠𝑛−1 := 𝑛−1 𝑐 𝑠 𝑖=0 𝑖 𝑖 Output: 𝑡 𝑦𝑖 = 𝑠𝑖−1 , 𝑡 = 1, … , 𝑛, 𝑖 = 𝑡 𝑛−1 𝑦𝑖 = 𝑗=0 𝑐𝑗 𝑦𝑖−𝑛+𝑗−1 , 𝑖 > 𝑛 3 Reconstruction attacks Solve for unknowns: 𝑐0 , . . . , 𝑐𝑛−1 𝑦𝑛+1 = 𝑐𝑛−1 𝑦𝑛 ⋯ 𝑐0 𝑦1 ⋮ 𝑦2𝑛 = 𝑐𝑛−1 𝑦2𝑛−1 ⋯ 𝑐0 𝑦𝑛 So we must use nonlinear feedback 𝑡 𝑠𝑖𝑡+1 := 𝑠𝑖+1 , 𝑖 = 0, … , 𝑛 − 2 𝑡+1 𝑡 𝑠𝑛−1 := 𝑔(𝑠0𝑡 , . . . , 𝑠𝑛−1 ), some nonlinear function 𝑔 4 Self-shrinking generator The self-shrinking generator uses alternating output bits of a single register to control its final output. 1. 2. 3. 4. 5. Clock two bits from the LFSR. If the pair is 10 output a zero. If the pair is 11 output a one. Otherwise, output nothing. Return to step one. 5 Self-shrinking generator, Example Use polynomial: x8 + x4 + x3 + x2 + 1 Initial state: 1 0 1 1 0 1 1 0. t 8 7 6 5 4 3 2 1 Out1 Out2 0 1 0 1 1 0 1 1 0 n/a n/a 1 1 1 0 1 1 0 1 1 0 2 1 1 1 0 1 1 0 1 1 3 1 1 1 1 0 1 1 0 1 4 1 1 1 1 1 0 1 1 0 n/a 0 6 Other nonlinear stream ciphers Trivium, eSTREAM project --see textbook These are hardware implementations of PRNG Next we shall consider a software implementation. 7 RC4 Init for RC4 (key scheduling) Algorithm 6.1 Input 16 byte key 𝑘 Output Initial state (𝑆, 𝑖, 𝑗), 𝑆 is a permutation of 0, . . . , 255, 𝑖, 𝑗 𝜖 {0, . . . , 255} for 𝑖 = 0 to 255 𝑆 𝑖 ≔ 𝑖, 𝑘 𝑖 ≔ 𝑘[𝑖 𝑚𝑜𝑑 16] 𝑗 ≔0 for 𝑖 = 0 to 255 𝑗 ≔ 𝑗 + 𝑆 𝑖 + 𝑘[𝑖] Swap 𝑆[𝑖] and 𝑆 𝑗 Return 𝑆, 𝑖, 𝑗 8 RC4 GetBits for RC4 (Algorithm 6.2) Input: (𝑆, 𝑖, 𝑗) Output: byte y, updated state (𝑆, 𝑖, 𝑗) 𝑖 ≔𝑖+1 𝑗 ≔𝑗+𝑆 𝑖 Swap 𝑆[𝑖] and 𝑆 𝑗 𝑡 ≔ 𝑆 𝑖 + 𝑆[𝑗] 𝑦 ≔ 𝑆[𝑡] Return 𝑆, 𝑖, 𝑗 , 𝑦 𝑦 9 Attacks on RC4 There are several attacks on RC4 known for some time and therefore this stream cipher should not be used anymore. A serious attack occurs when an IV is prepended to the to the key. This attack can be used to recover the key (regardless of it length). This attack was used to break the WEP encryption standard, and was influential in getting the standard replaced---see textbook for details of the attack. 10 Block ciphers A block cipher is an efficient keyed permutation 𝐹 ∶ {0,1}𝑛 × {0,1}𝑙 → {0,1}𝑙 𝐹𝑘 𝑥 ≝ 𝐹(𝑘, 𝑥) is a bijection, and 𝐹𝑘 and its inverse 𝐹𝑘−1 are efficiently computable given 𝑘. Block ciphers should be viewed as pseudorandom permutations rather than as encryption schemes. They are a basic building blocks for symmetric key applications. 11 Block ciphers We refer to 𝑛 as the key length and 𝑙 as the block length of 𝐹. These are now constants (fixed) whereas earlier they where functions of the security parameter. This takes us away from the asymptotic security to concrete security. 12 Substitution-Permutation Networks A block cipher must behave like a random permutation. However there are 2𝑙 ! permutations on 𝑙-bit strings, so representing an arbitrary permutation with block length 𝑙 requires roughly log(2𝑙 !) ≈ 𝑙 ∙ 2𝑙 bits. Thus, we need to somehow construct a concise function that behaves like a random function 13 The confusion−diffusion paradigm Idea (Shannon): construct a random looking permutation 𝐹 with large block length using smaller random looking substitutions {𝑓𝑖 } with small length. A substitution-permutation network is an implementation of this paradigm. 14 The confusion−diffusion paradigm The substitution component refers to small random functions 𝑓𝑖 called S-boxes and the permutation component refers to the mixing of the outputs of the random functions. The permutation component involves the reordering of the output bits and are called mixing permutations. 15 The confusion−diffusion paradigm An example, 1 Suppose we want 𝐹 to have block length 128 bits, and use 16 substitutions 𝑓1 , . . . , 𝑓16 that have block length 8 bits. The key 𝑘 will specify the 16 substitutions. For input 𝑥 ∈ {0,1}128 we parse 𝑥 as 𝑥1 , … , 𝑥16 and set 𝐹𝑘 𝑥 = 𝑓1 𝑥1 || ⋯ ||𝑓16 𝑥16 The “round” functions {𝑓𝑖 } are said to introduce confusion. 16 The confusion−diffusion paradigm An example, 2 A diffusion step then mixes the bits of the output. For example the bits of 𝐹𝑘 𝑥 are shuffled to get 𝑥′. The confusion-diffusion process is repeated several times A substitution-permutation network is an implementation of this paradigm. 17 The confusion−diffusion paradigm An example, 3 Consider an SPN network with 64 bit block length and 8 bit 𝑆-boxes, 𝑆1 , … , 𝑆8 . Evaluating the cipher proceeds in a number of rounds in which: Key mixing: set 𝑥 ≔ 𝑥 𝑘, where 𝑘 is the current “round sub-key”. Substitution: set 𝑥 ≔ 𝑆1 (𝑥1 )|| ⋯ ||𝑆8 𝑥8 . Permutation: Permute the bits of 𝑥 to get the output for the next round. 18 Substitution-permutation network Example 3, single round 19 The confusion−diffusion paradigm The basic idea is to break the input up into small parts and then feed these parts through different S-boxes (random permutations). The outputs are then mixed together. The process is repeated a given number of times, called a rounds. The S-boxes introduce confusion into the construction. In order to spread the confusion throughout, the results are mixed together, achieving diffusion. 20 Any SPN is invertible (given the key) It suffices to invert each round. Given the SPN output for a round and the key we: First invert the mixing permutation Then invert the 𝑆-box permutations Finally XOR the result with the appropriate sub-key to get the round input. 21 The avalanche effect An important property in any block cipher is that small changes to the input must result in large changes to the output. To ensure this, block ciphers are designed so that small changes in the input propagate quickly to very large changes in the intermediate values. 22 The avalanche effect It is easy to demonstrate that the avalanche effect holds in a substitution-permutation network, when the following hold: 1. The 𝑆-boxes are designed so that any change of at least a single bit to the input to an 𝑆-box results in a change of at least two bits in the output. 2. The mixing permutations are designed so that the output bits of any given 𝑆-box are spread into different 𝑆-boxes in the next round. 23 Feistel Networks A Feistel* network is an alternative way of constructing a block cipher. The low-level building blocks (S-boxes, mixing permutations and key schedule) are the same. The difference is in the high-level design. The advantage of Feistel networks over substitution permutation networks is that they enable the use of S-boxes that are not necessarily invertible. * Horst Feistel who did pioneering research while working for IBM 24 Feistel Networks This is important because a good block cipher has chaotic behavior (it should look random). Requiring that all of the components of the construction be invertible inherently introduces structure, which contradicts the need for chaos. 25 Feistel Networks A Feistel network is thus a way of constructing an invertible function from non-invertible components. This seems like a contradiction in terms---if you cannot invert the components, how can you invert the overall structure. Nevertheless, the Feistel design ingeniously overcomes this obstacle. 26 A Feistel network 1. For input 𝑥, denote by 𝑥1 and 𝑥2 the first and second halves of 𝑥 respectively. 2. Let 𝑣1 = 𝑥1 and 𝑣2 = 𝑥2 . 3. For 𝑖 = 1 to 𝑟 (where 𝑟 is the number of rounds in the network): a) Let 𝑤1 = 𝑣2 and 𝑤2 = 𝑣1 𝑓𝑖 (𝑣2 ), where 𝑓𝑖 denotes the 𝑓-function in the 𝑖-th round of the network. b) Let 𝑣1 = 𝑤1 and 𝑣2 = 𝑤2 . c) The output 𝑦 is (𝑣1 , 𝑣2 ). 27 Feistel Network . mm mmm mm mmm m mmm mmm 28