180 likes | 345 Views
CS G513 / SS G513 Network Security. Agenda. Stream Ciphers One-time Pads Synchronous Stream Ciphers Self-Synchronizing Stream Ciphers Feedback Shift Registers Linear Feedback Shift Registers Non-Linear Feedback Shift Registers Non-Linear Combination Generators
E N D
CS G513 / SS G513 Network Security Agenda Stream Ciphers One-time Pads Synchronous Stream Ciphers Self-Synchronizing Stream Ciphers Feedback Shift Registers Linear Feedback Shift Registers Non-Linear Feedback Shift Registers Non-Linear Combination Generators Non-Linear Filter Generators Software Stream Ciphers
Stream Ciphers • Block Ciphers vs. Stream Ciphers: • Block Ciphers • process plaintext in large blocks and • repeat the same process (of encryption) across blocks • Thus, block ciphers are typically stateless/memoryless. • With Stream ciphers encryption function may vary as plaintext is processed and • Stream Ciphers process plaintext in small blocks (say one character or often just one bit)
Stream Ciphers: One-time Pad • A Vernam cipher (over the binary alphabet) is defined as: • Ci = Mi Ki where for I=1,2,3,… Mi refers to the keystream, Ki the keystream, Ci the cipherstream. • (Vernam’s) One-time Pad: • The keystream is generated independently and randomly. • This is unconditionally secure against a cipher-text only attack. • Observation: • Key is (at least) as long as the message – uncertainty of the key is at least as great as that of the message Sundar B.
Synchronous Stream Ciphers • Keystream is generated independently of the plaintext message and of the ciphertext. • Encryption process: • Si+1 = f(Si, k) Zi = g(Si, k) Ci = h(Zi, Mi) • where S0 is the initital state (may depend on k) • and Zi is the keystream • and Ci is the cipher-stream • Decryption process: • Si+1 = f(Si, k) Zi = g(Si, k) Mi = h-1(Zi, Ci) • E.g.: Binary Additive Stream Cipher: • streams are binary and h is Sundar B.
Synchronous Stream Ciphers • Properties: • Synchronization Requirements: • Sender and Receiver must be synchronized – using the same key and operating at the same state within that key • Insertion/Deletion may cause loss of synchronization • Re-synchronization may need re-initialization and/or special marks in the stream at regular intervals. • No Error Propagation: • Modified digit does not affect decryption of other digits • Active Attacks: • Insertion/Deletion/Replay cause loss of sync. And hence detected by decryptor • Chosen plaintext attacks possible due to lack of error propagation.
Self-synchronizing Stream Ciphers • Keystream is generated as a function of the key and a fixed number of previous ciphertext digits. • Encryption process: • Si = (Ci-t, Ci-t+1, …, Ci-1) • Zi = g(Si, k) Ci = h(Zi, Mi) • where S0 = (C-t, C-t+1, …, C-1) is the initial state • and Zi is the keystream • and Ci is the cipher-stream • Decryption process: • Si = (Ci-t, Ci-t+1, …, Ci-1) • Zi = g(Si, k) Mi = h-1(Zi, Ci)
Self-synchronizing Stream Ciphers • Properties: • Self-synchronization: • possible with insertions/deletions (at most t digits may be lost) • Limited Error Propagation: • 1 digit modification/insertion/deletion may cause incorrect decryption of up to t digits. • Active Attacks • Modification can be detected due to incorrect decryption – better than synchronous stream ciphers. • It is more difficult than for synch. stream ciphers to detect insertion / deletion / replay of ciphertext digits. • Diffusion of plaintext statistics: Better Sundar B.
Linear Feedback Shift Registers • An LFSR of length L consists of • L stages (or delay elements) capable of storing 1 bit each and • a clock controlling the movement of data. • During each unit of time: • Content of stage 0 is output • Content of stage j is moved to stage j-1 for each j (1 to L-1) • New content of stage L-1 is the feedback bit computed as sum without carry of previous contents of a fixed subset of stages. Sundar B.
Linear Feedback Shift Registers Figure from Menezes et. al.
Linear Feedback Shift Registers • If the initial state of the LFSR in the Figure (prev. slide) is • [sL-1, …, s1, s0] • then the output sequence is determined by the recursion • sj = (c1sj-1 + c2sj-2 + … + cLsj-L) Sundar B.
Feedback Shift Registers • Linear Feedback Shift Registers (LFSRs) are used in key-stream generation because: • they are well-suited for hardware implementation • they can produce sequences of large period • they can produce sequences w/ good statistical properties • they can be readily analyzed by algebraic techniques Reading exercise – Refer to Menezes et. al. Chapter 6 Sundar B.
LFSRs - Limitations • Primary Weakness: Linear Complexity – no of stages needed for generating a sequence • At most n for a n-bit sequence • At most N for a sequence with period N • Claim: If an inifinite bit stream s is generated by an LFSR of L stages, then there is an O(n2) algorithm that determines the linear combination used for generating s given a substream t of s if the length of t > 2L. • Proof: Berlekamp-Massey algo. • Refer to Menezes et. al (Chapter 6) Sundar B.
(Generalized) Feedback Shift Registers • An LFSR of length L consists of • L stages (or delay elements) capable of storing 1 bit each and • a clock controlling the movement of data. • During each unit of time: • Content of stage 0 is output • Content of stage j is moved to stage j-1 for each j (1 to L-1) • New content of stage L-1 is the feedback bit Sj = f(Sj-1, Sj-2, … Sj-L) where f is a boolean function and Sj-i is the previous content of stage L-i, 1<= i <= L Sundar B.
(Generalized) Feedback Shift Registers Figure from Menezes et. al. If f is a non-linear function, then this is a non-linear FSR Sundar B.
Non-linear Combination Generators f is a non-linear combining function Figure from Menezes et. al. Note that f must be suitable for destroying linearity of LFSR outputs Sundar B.
Non-linear Filter Generators f is a non-linear filtering function Figure from Menezes et. al. Sundar B.
Non-linear Filter Generators • Example: Knapsack Generator • Secret key consists of L knapsack integer weights a1,a2,…,aL each of bit length L • At time j, the knapsack sum Sj is computed as: • Sj = sum from k=1 to L of (xkak mod 2L) where [xL, …, x2,x1] is the state of the LFSR at time j. • Selected bits of Sj are extracted to form part of the keystream. • Determining a subset of weights from a given sum is a hard problem. (Subset Sum is NP-hard). Sundar B.
Other Stream Ciphers • LFSR based stream ciphers are well-suited for hardware implementation • But they are not necessarily amenable for software implementations. • SEAL (Software Optimized Encryption Algorithm) • A binary additive stream cipher designed for efficient software implementation in 32-bit processors (circa 1993) • Maps a 32-bit sequence number n into a L-bit keystream under control of a 160-bit secret key a. • Refer to Menezes et. al (Chapter 6) Sundar B.