# Rijndael-GF¶

Rijndael-GF is an algebraic implementation of the AES cipher which seeks to provide a fully generalized algebraic representation of both the whole AES cipher as well as its individual components.

This class is an algebraic implementation of the Rijndael-GF extension of the AES cipher, as described in [DR2002]. The AES cipher itself is defined to operate on a state in $$(\GF{2})^{8 n_t}$$ where $$n_t \in \{16, 20, 24, 28, 32\}$$. Rijndael-GF is a generalization of AES which allows for operations in $$(\GF{2^8})^{n_t}$$, enabling more algebraically sophisticated study of AES and its variants. This implementation of Rijndael-GF is suitable for learning purposes, for comparison to other algebraic ciphers, and for studying various techniques of algebraic cryptanalysis of AES. This cipher is different from Mini-AES, which is a teaching tool for beginners to understand the basic structure of AES.

An algebraic implementation of Rijndael-GF is achieved by recognizing that for each round component function $$\phi$$ of AES (SubBytes, ShiftRows, etc.) operating on state matrices, every entry of the output matrix $$B = \phi(A)$$ is representable as a polynomial with variables being the entries of the input state matrix $$A$$. Correspondingly, this implementation of Rijndael-GF provides a RijndaelGF.Round_Component_Poly_Constr class which allows for creation of these such polynomials. For each round component function $$\phi$$ of Rijndael-GF there exists a Round_Component_Poly_Constr object with a __call__ method of the form __call__(i, j) which returns a polynomial representing $$\phi(A)_{i,j}$$ in terms of the entries of $$A$$. There additionally are various methods provided which allow for easy polynomial evaluation and for simple creation of Round_Component_Poly_Constr objects representing more complex aspects of the cipher.

This approach to implementing Rijndael-GF bears some similarity to the multivariate quadratic (MQ) systems utilized in SR, in that the MQ systems also seek to describe the AES cipher as a system of algebraic equations. Despite this initial similarity though, Rijndael-GF and SR are quite different as this implementation seeks to provide a fully generalized algebraic representation of both the whole AES cipher as well as its individual components, while SR is instead a family of parameterizable variants of the AES suitable as a framework for comparing different cryptanalytic techniques that can be brought to bear on the AES.

AUTHORS:

• Thomas Gagne (2015-06): initial version

EXAMPLES

We build Rijndael-GF with a block length of 4 and a key length of 6:

sage: from sage.crypto.mq.rijndael_gf import RijndaelGF
sage: rgf = RijndaelGF(4, 6)


We can encrypt plaintexts and decrypt and ciphertexts by calling the encrypt and decrypt methods or by calling the Rijndael-GF object explicitly. Note that the default input format is a hex string.

sage: plaintext = '00112233445566778899aabbccddeeff'
sage: key = '000102030405060708090a0b0c0d0e0f1011121314151617'
sage: rgf.encrypt(plaintext, key)
'dda97ca4864cdfe06eaf70a0ec0d7191'
sage: rgf.decrypt('dda97ca4864cdfe06eaf70a0ec0d7191', key)
'00112233445566778899aabbccddeeff'


We can also use binary strings as input and output.

sage: plain = '11101011100111110000000111001100' * 4
sage: key = '01100010111101101000110010111010' * 6
sage: ciphertext = rgf(plain, key, format='binary')
sage: ciphertext
'11010011000010011010110001000011101110110100110100110010011011111100011011100111110011100111010011001110110100011100000011111011'
sage: rgf(ciphertext, key, algorithm='decrypt', format='binary') == plain
True


[DR2002] demonstrates an example of encryption which takes the plaintext ‘3243f6a8885a308d313198a2e0370734’ and the key ‘2b7e151628aed2a6abf7158809cf4f3c’ and returns the ciphertext ‘3902dc1925dc116a8409850b1dfb9732’. We can use this example to demonstrate the correctness of this implementation:

sage: rgf = RijndaelGF(4, 4) # change dimensions for this example
sage: plain = '3243f6a8885a308d313198a2e0370734'
sage: key = '2b7e151628aed2a6abf7158809cf4f3c'
sage: expected_ciphertext = '3925841d02dc09fbdc118597196a0b32'
sage: rgf.encrypt(plain, key) == expected_ciphertext
True

sage: rgf = RijndaelGF(4, 6) # revert to previous dimensions


To build polynomials representing entries of the output matrix $$B = \phi(A)$$ for any round component function $$\phi$$, each of the round component functions (SubBytes, ShiftRows, and MixColumns) have a Round_Component_Poly_Constr object associated with it for building polynomials. These objects can be accessed by calling their getter functions: rgf.sub_bytes_poly(), rgf.shift_rows_poly(), and rgf.mix_columns_poly(). Each returned object has a __call__ method which takes an index i,j and an algorithm flag (‘encrypt’ or ‘decrypt’) and returns a polynomial representing $$\phi(A)_{i,j}$$ in terms of the entries of $$A$$, where $$A$$ is an arbitrary state matrix and $$\phi$$ is the round component function associated with that particular Round_Component_Poly_Constr object. Some of these objects’ __call__ methods also have additional keywords to modify their behavior, and so we describe the usage of each object below.

rgf.shift_rows_poly() and rgf.mix_columns_poly() do not have any additional keywords for their __call__ methods and we can call them as such:

sage: sr_pc = rgf.shift_rows_poly_constr()
sage: sr_pc(1, 2)
a13
sage: sr_pc(2, 3, algorithm='decrypt')
a21

sage: mc_pc = rgf.mix_columns_poly_constr()
sage: mc_pc(1, 2)
a02 + (x)*a12 + (x + 1)*a22 + a32
sage: mc_pc(2, 3, algorithm='decrypt')
(x^3 + x^2 + 1)*a03 + (x^3 + 1)*a13 + (x^3 + x^2 + x)*a23 + (x^3 + x + 1)*a33


rgf.sub_bytes_poly() has a single keyword no_inversion=False, which when set to True returns only the affine transformation step of SubBytes. Below describes the usage of rgf.sub_bytes_poly()

sage: sb_pc = rgf.sub_bytes_poly_constr()
sage: sb_pc(1, 2)
(x^2 + 1)*a12^254 +
(x^3 + 1)*a12^253 +
(x^7 + x^6 + x^5 + x^4 + x^3 + 1)*a12^251 +
(x^5 + x^2 + 1)*a12^247 +
(x^7 + x^6 + x^5 + x^4 + x^2)*a12^239 +
a12^223 +
(x^7 + x^5 + x^4 + x^2 + 1)*a12^191 +
(x^7 + x^3 + x^2 + x + 1)*a12^127 +
(x^6 + x^5 + x + 1)
sage: sb_pc(2, 3, no_inversion=True)
(x^7 + x^3 + x^2 + x + 1)*a23^128 +
(x^7 + x^5 + x^4 + x^2 + 1)*a23^64 +
a23^32 +
(x^7 + x^6 + x^5 + x^4 + x^2)*a23^16 +
(x^5 + x^2 + 1)*a23^8 +
(x^7 + x^6 + x^5 + x^4 + x^3 + 1)*a23^4 +
(x^3 + 1)*a23^2 +
(x^2 + 1)*a23 +
(x^6 + x^5 + x + 1)


Because of the order of the affine transformation and the inversion step in SubBytes, calling rgf.sub_bytes_poly()(i, j, algorithm='decrypt') results in a polynomial with thousands of terms which takes a very long time to compute. Hence, when using the decryption version of rgf.sub_bytes_poly() with the intention of evaluating the polynomials it constructs, it is recommended to first call rgf.sub_bytes_poly()(i, j, algorithm='decrypt', no_inversion=True) to get a polynomial representing only the inverse affine transformation, evaluate this polynomial for a particular input block, then finally perform the inversion step after the affine transformation polynomial has been evaluated.

sage: inv_affine = sb_pc(1, 2, algorithm='decrypt',
....: no_inversion=True)
sage: state = rgf._hex_to_GF('ff87968431d86a51645151fa773ad009')
sage: evaluated = inv_affine(state.list())
sage: result = evaluated * -1
sage: rgf._GF_to_hex(result)
'79'


We can see how the variables of these polynomials are organized in $$A$$:

sage: rgf.state_vrs
[a00 a01 a02 a03]
[a10 a11 a12 a13]
[a20 a21 a22 a23]
[a30 a31 a32 a33]


The final Round_Component_Poly_Constr object we have not discussed yet is add_round_key_poly, which corresponds to the AddRoundKey round component function. This object differs from the other Round_Component_Poly_Constr objects in that it returns polynomials with variables being entries of an input state $$A$$ as well as entries of various subkeys. Since there are $$N_r$$ subkeys to choose from, add_round_key_poly has a keyword of round=0 to select which subkey to use variables from.

sage: ark_pc = rgf.add_round_key_poly_constr()
sage: ark_pc(1, 2)
a12 + k012
sage: ark_pc(1, 2, algorithm='decrypt')
a12 + k012
sage: ark_pc(2, 3, round=7)
a23 + k723


We can see how key variables are organized in the original key (the key used to build the rest of the subkeys) below. Note that because key variables are subkey entries, if the key length is longer than the block length we will have entries from multiple subkeys in the original key matrix.

sage: rgf.key_vrs
[k000 k001 k002 k003 k100 k101]
[k010 k011 k012 k013 k110 k111]
[k020 k021 k022 k023 k120 k121]
[k030 k031 k032 k033 k130 k131]


We can evaluate any of these constructed polynomials for a particular input state (in essence, calculate $$\phi(A)_{i,j}$$) as such:

sage: rgf = RijndaelGF(4, 6)
sage: state = rgf._hex_to_GF('fe7b5170fe7c8e93477f7e4bf6b98071')
sage: poly = mc_pc(3, 2, algorithm='decrypt')
sage: poly(state.list())
x^7 + x^6 + x^5 + x^2 + x


We can use the apply_poly method to build a matrix whose $$i,j$$ th entry equals the polynomial phi_poly(i, j) evaluated for a particular input state, where phi_poly is the Round_Component_Poly_Constr object associated with the round component function $$\phi$$. Essentially, apply_poly calculates $$\phi(A)$$, where $$A$$ is our input state. Calling apply_poly is equivalent to applying the round component function associated this Round_Component_Poly_Constr object to $$A$$.

sage: state = rgf._hex_to_GF('c4cedcabe694694e4b23bfdd6fb522fa')
sage: result = rgf.apply_poly(state, rgf.sub_bytes_poly_constr())
sage: rgf._GF_to_hex(result)
'1c8b86628e22f92fb32608c1a8d5932d'
sage: result == rgf.sub_bytes(state)
True


Alternatively, we can pass a matrix of polynomials as input to apply_poly, which will then return another matrix of polynomials. For example, rgf.state_vrs can be used as input to make each i,j th entry of the output matrix equal phi_poly_constr(i, j), where phi_poly_constr is our inputted Round_Component_Poly_Constr object. This matrix can then be passed through again and so on, demonstrating how one could potentially build a matrix of polynomials representing the entire cipher.

sage: state = rgf.apply_poly(rgf.state_vrs, rgf.shift_rows_poly_constr())
sage: state
[a00 a01 a02 a03]
[a11 a12 a13 a10]
[a22 a23 a20 a21]
[a33 a30 a31 a32]
[a00 + k000 a01 + k001 a02 + k002 a03 + k003]
[a11 + k010 a12 + k011 a13 + k012 a10 + k013]
[a22 + k020 a23 + k021 a20 + k022 a21 + k023]
[a33 + k030 a30 + k031 a31 + k032 a32 + k033]


For any of these Round_Component_Poly_Constr objects, we can change the keywords of its __call__ method when apply_poly invokes it by passing apply_poly a dictionary mapping keywords to their values.

sage: rgf.apply_poly(rgf.state_vrs, rgf.add_round_key_poly_constr(),
....: poly_constr_attr={'round' : 5})
[a00 + k500 a01 + k501 a02 + k502 a03 + k503]
[a10 + k510 a11 + k511 a12 + k512 a13 + k513]
[a20 + k520 a21 + k521 a22 + k522 a23 + k523]
[a30 + k530 a31 + k531 a32 + k532 a33 + k533]


We can build our own Round_Component_Poly_Constr objects which correspond to the composition of multiple round component functions with the compose method. To do this, if we pass two Round_Component_Poly_Constr objects to compose where the first object corresponds to the round component function $$f$$ and the second to the round component function $$g$$, compose will return a new Round_Component_Poly_Constr object corresponding to the function $$g \circ f$$. This returned Round_Component_Poly_Constr object will have the arguments of __call__(row, col, algorithm='encrypt') and when passed an index i,j will return $$g(f(A))_{i,j}$$ in terms of the entries of $$A$$.

sage: rcpc = rgf.compose(rgf.shift_rows_poly_constr(),
....: rgf.mix_columns_poly_constr())
sage: rcpc
A polynomial constructor of a round component of Rijndael-GF block cipher with block length 4, key length 6, and 12 rounds.
sage: rcpc(2, 1)
a01 + a12 + (x)*a23 + (x + 1)*a30

sage: state = rgf._hex_to_GF('afb73eeb1cd1b85162280f27fb20d585')
sage: result = rgf.apply_poly(state, rcpc)
sage: new_state = rgf.shift_rows(state)
sage: new_state = rgf.mix_columns(new_state)
sage: result == new_state
True

sage: rcpc = rgf.compose(rgf.mix_columns_poly_constr(),
....: rgf.shift_rows_poly_constr())
sage: result = rgf.apply_poly(state, rcpc, algorithm='decrypt')
sage: new_state = rgf.mix_columns(state, algorithm='decrypt')
sage: new_state = rgf.shift_rows(new_state, algorithm='decrypt')
sage: new_state == result
True


Alternatively, we can use compose to build the polynomial output of a Round_Component_Poly_Constr object corresponding to the composition of multiple round functions like above without having to explicitly build our own Round_Component_Poly_Constr object. To do this, we simply make the first input a Round_Component_Poly_Constr object corresponding to a round component function $$f$$ and make the second input a polynomial representing $$g(A)_{i,j}$$ for a round component function $$g$$. Given this, compose will return a polynomial representing $$g(f(A))_{i,j}$$ in terms of the entries of $$A$$.

sage: poly = rgf.mix_columns_poly_constr()(0, 3)
sage: poly
(x)*a03 + (x + 1)*a13 + a23 + a33
sage: rgf.compose(rgf.sub_bytes_poly_constr(), poly)
(x^3 + x)*a03^254 +
(x^3 + x^2 + x + 1)*a13^254 +
(x^2 + 1)*a23^254 +
(x^2 + 1)*a33^254 +
(x^4 + x)*a03^253 +
(x^4 + x^3 + x + 1)*a13^253 +
(x^3 + 1)*a23^253 +
(x^3 + 1)*a33^253 +
(x^7 + x^6 + x^5 + x^3 + 1)*a03^251 +
(x^4)*a13^251 +
(x^7 + x^6 + x^5 + x^4 + x^3 + 1)*a23^251 +
(x^7 + x^6 + x^5 + x^4 + x^3 + 1)*a33^251 +
(x^6 + x^3 + x)*a03^247 +
(x^6 + x^5 + x^3 + x^2 + x + 1)*a13^247 +
(x^5 + x^2 + 1)*a23^247 +
(x^5 + x^2 + 1)*a33^247 +
(x^7 + x^6 + x^5 + x^4 + x + 1)*a03^239 +
(x^2 + x + 1)*a13^239 +
(x^7 + x^6 + x^5 + x^4 + x^2)*a23^239 +
(x^7 + x^6 + x^5 + x^4 + x^2)*a33^239 +
(x)*a03^223 +
(x + 1)*a13^223 +
a23^223 +
a33^223 +
(x^6 + x^5 + x^4 + 1)*a03^191 +
(x^7 + x^6 + x^2)*a13^191 +
(x^7 + x^5 + x^4 + x^2 + 1)*a23^191 +
(x^7 + x^5 + x^4 + x^2 + 1)*a33^191 +
(x^2 + 1)*a03^127 +
(x^7 + x^3 + x)*a13^127 +
(x^7 + x^3 + x^2 + x + 1)*a23^127 +
(x^7 + x^3 + x^2 + x + 1)*a33^127 +
(x^6 + x^5 + x + 1)


If we use algorithm='decrypt' as an argument to compose, then the value of algorithm will be passed directly to the first argument of compose (a Round_Component_Poly_Constr object) when it is called, provided the second argument is a polynomial. Setting this flag does nothing if both arguments are Round_Component_Poly_Constr objects, since the returned Round_Component_Poly_Constr object’s __call__ method must have its own algorithm keyword defaulted to ‘encrypt’.

sage: poly = rgf.shift_rows_poly_constr()(2, 1)
sage: rgf.compose(rgf.mix_columns_poly_constr(), poly, algorithm='decrypt')
(x^3 + x^2 + 1)*a03 + (x^3 + 1)*a13 + (x^3 + x^2 + x)*a23 + (x^3 + x + 1)*a33

sage: state = rgf._hex_to_GF('80121e0776fd1d8a8d8c31bc965d1fee')
sage: with_decrypt = rgf.compose(rgf.sub_bytes_poly_constr(),
....: rgf.shift_rows_poly_constr(), algorithm='decrypt')
sage: result_wd = rgf.apply_poly(state, with_decrypt)
sage: no_decrypt = rgf.compose(rgf.sub_bytes_poly_constr(),
....: rgf.shift_rows_poly_constr())
sage: result_nd = rgf.apply_poly(state, no_decrypt)
sage: result_wd == result_nd
True


We can also pass keyword dictionaries of f_attr and g_attr to compose to make f and g use those keywords during polynomial creation.

sage: rcpc = rgf.compose(rgf.add_round_key_poly_constr(),
....: f_attr={'round' : 4}, g_attr={'round' : 7})
sage: rcpc(1, 2)
a12 + k412 + k712


In addition to building polynomial representations of state matrices, we can also build polynomial representations of elements of the expanded key with the expand_key_poly method. However, since the key schedule is defined recursively, it is impossible to build polynomials for the key schedule in the same manner as we do for the round component functions. Consequently, expand_round_key_poly() is not a Round_Component_Poly_Constr object. Instead, expand_key_poly is a method which takes an index i,j and a round number round, and returns a polynomial representing the $$i,j$$ th entry of the round th round key. This polynomial’s variables are entries of the original key we built above.

sage: rgf.expand_key_poly(1, 2, 0)
k012
sage: rgf.expand_key_poly(1, 1, 1)
k111
sage: rgf.expand_key_poly(1, 2, 1)
(x^2 + 1)*k121^254 +
(x^3 + 1)*k121^253 +
(x^7 + x^6 + x^5 + x^4 + x^3 + 1)*k121^251 +
(x^5 + x^2 + 1)*k121^247 +
(x^7 + x^6 + x^5 + x^4 + x^2)*k121^239 +
k121^223 +
(x^7 + x^5 + x^4 + x^2 + 1)*k121^191 +
(x^7 + x^3 + x^2 + x + 1)*k121^127 +
k010 +
(x^6 + x^5 + x)


Since expand_key_poly is not actually a Round_Component_Poly_Constr object, we cannot use it as input to apply_poly or compose.

sage: rgf.apply_poly(state, rgf.expand_key_poly)
Traceback (most recent call last):
...
TypeError: keyword 'poly_constr' must be a Round_Component_Poly_Constr
sage: rgf.compose(rgf.expand_key_poly, rgf.sub_bytes_poly_constr())
Traceback (most recent call last):
...
TypeError: keyword 'f' must be a Round_Component_Poly_Constr

class sage.crypto.mq.rijndael_gf.RijndaelGF(Nb, Nk, state_chr='a', key_chr='k')

An algebraically generalized version of the AES cipher.

INPUT:

• Nb – The block length of this instantiation. Must be between 4 and 8.
• Nk – The key length of this instantiation. Must be between 4 and 8.
• state_chr – The variable name for polynomials representing elements from state matrices.
• key_chr – The variable name for polynomials representing elements of the key schedule.

EXAMPLES:

sage: from sage.crypto.mq.rijndael_gf import RijndaelGF
sage: rgf = RijndaelGF(6, 8)
sage: rgf
Rijndael-GF block cipher with block length 6, key length 8, and 14 rounds.


By changing state_chr we can alter the names of variables in polynomials representing elements from state matrices.

sage: rgf = RijndaelGF(4, 6, state_chr='myChr')
sage: rgf.mix_columns_poly_constr()(3, 2)
(x + 1)*myChr02 + myChr12 + myChr22 + (x)*myChr32


We can also alter the name of variables in polynomials representing elements from round keys by changing key_chr.

sage: rgf = RijndaelGF(4, 6, key_chr='myKeyChr')
sage: rgf.expand_key_poly(1, 2, 1)
(x^2 + 1)*myKeyChr121^254 +
(x^3 + 1)*myKeyChr121^253 +
(x^7 + x^6 + x^5 + x^4 + x^3 + 1)*myKeyChr121^251 +
(x^5 + x^2 + 1)*myKeyChr121^247 +
(x^7 + x^6 + x^5 + x^4 + x^2)*myKeyChr121^239 +
myKeyChr121^223 +
(x^7 + x^5 + x^4 + x^2 + 1)*myKeyChr121^191 +
(x^7 + x^3 + x^2 + x + 1)*myKeyChr121^127 +
myKeyChr010 +
(x^6 + x^5 + x)

class Round_Component_Poly_Constr(polynomial_constr, rgf, round_component_name=None)

An object which constructs polynomials representing round component functions of a RijndaelGF object.

INPUT:

• polynomial_constr – A function which takes an index row,col and returns a polynomial representing the row,col th entry of a matrix after a specific round component function has been applied to it. This polynomial must be in terms of entries of the input matrix to that round component function and of entries of various subkeys. polynomial_constr must have arguments of the form polynomial_constr(row, col, algorithm='encrypt', **kwargs) and must be able to be called as polynomial_constr(row, col).
• rgf – The RijndaelGF object whose state entries are represented by polynomials returned from polynomial_constr.
• round_component_name – The name of the round component function this object corresponds to as a string. Used solely for display purposes.

EXAMPLES:

sage: from sage.crypto.mq.rijndael_gf import \
....: RijndaelGF
sage: rgf = RijndaelGF(4, 4)
sage: rcpc = RijndaelGF.Round_Component_Poly_Constr(
....: rgf._shift_rows_pc, rgf, "Shift Rows")
sage: rcpc
A polynomial constructor for the function 'Shift Rows' of Rijndael-GF block cipher with block length 4, key length 4, and 10 rounds.


If $$\phi$$ is the round component function to which this object corresponds to, then __call__(i,j) $$= \phi(A)_{i,j}$$, where $$A$$ is an arbitrary input matrix. Note that the polynomial returned by __call__(i,j) will be in terms of the entries of $$A$$.

sage: rcpc = RijndaelGF.Round_Component_Poly_Constr(
....: rgf._mix_columns_pc, rgf, "Mix Columns")
sage: poly = rcpc(1, 2); poly
a02 + (x)*a12 + (x + 1)*a22 + a32
sage: state = rgf._hex_to_GF('d1876c0f79c4300ab45594add66ff41f')
sage: result = rgf.mix_columns(state)
sage: result[1,2] == poly(state.list())
True


Invoking this objects __call__ method passes its arguments directly to polynomial_constr and returns the result. In a sense, Round_Component_Poly_Constr acts as a wrapper for the polynomial_constr method and helps ensure that each Round_Component_Poly_Constr object will act similarly.

sage: all(rgf._mix_columns_pc(i, j) == rcpc(i, j)
....: for i in range(4) for j in range(4))
True


Since all keyword arguments of polynomial_constr must have a default value except for row and col, we can always call a Round_Component_Poly_Constr object by __call__(row, col). Because of this, methods such as apply_poly and compose will only call __call__(row, col) when passed a Round_Component_Poly_Constr object. In order to change this object’s behavior and force methods such as apply_poly to use non-default values for keywords we can pass dictionaries mapping keywords to non-default values as input to apply_poly and compose.

sage: rgf.apply_poly(rgf.state_vrs,
....: poly_constr_attr={'round' : 9})
[a00 + k900 a01 + k901 a02 + k902 a03 + k903]
[a10 + k910 a11 + k911 a12 + k912 a13 + k913]
[a20 + k920 a21 + k921 a22 + k922 a23 + k923]
[a30 + k930 a31 + k931 a32 + k932 a33 + k933]

sage: fn = rgf.compose(rgf.add_round_key_poly_constr(),
....: f_attr={'round' : 3}, g_attr={'round' : 7})
sage: fn(2, 3)
a23 + k323 + k723


Because all Round_Component_Poly_Constr objects are callable as __call__(row, col, algorithm), __call__ will check the validity of these three arguments automatically. Any other keywords, however, must be checked in polynomial_constr.

sage: def my_poly_constr(row, col, algorithm='encrypt'):
....:     return x * rgf._F.one() # example body with no checks
....:
sage: rcpc = RijndaelGF.Round_Component_Poly_Constr(
....: my_poly_constr, rgf, "My Poly Constr")
sage: rcpc(-1, 2)
Traceback (most recent call last):
...
ValueError: keyword 'row' must be in range 0 - 3
sage: rcpc(1, 2, algorithm=5)
Traceback (most recent call last):
...
ValueError: keyword 'algorithm' must be either 'encrypt' or 'decrypt'

add_round_key(state, round_key)

Returns the round-key addition of matrices state and round_key.

INPUT:

• state – The state matrix to have round_key added to.
• round_key – The round key to add to state.

OUTPUT:

• A state matrix which is the round key addition of state and round_key. This transformation is simply the entrywise addition of these two matrices.

EXAMPLES:

sage: from sage.crypto.mq.rijndael_gf import RijndaelGF
sage: rgf = RijndaelGF(4, 4)
sage: state = rgf._hex_to_GF('36339d50f9b539269f2c092dc4406d23')
sage: key = rgf._hex_to_GF('7CC78D0E22754E667E24573F454A6531')
sage: key_schedule = rgf.expand_key(key)
sage: result = rgf.add_round_key(state, key_schedule[0])
sage: rgf._GF_to_hex(result)
'4af4105edbc07740e1085e12810a0812'

add_round_key_poly_constr()

Return the Round_Component_Poly_Constr object corresponding to AddRoundKey.

EXAMPLES:

sage: from sage.crypto.mq.rijndael_gf import RijndaelGF
sage: rgf = RijndaelGF(4, 4)
sage: ark_pc = rgf.add_round_key_poly_constr()
sage: ark_pc
A polynomial constructor for the function 'Add Round Key' of Rijndael-GF block cipher with block length 4, key length 4, and 10 rounds.
sage: ark_pc(0, 1)
a01 + k001


When invoking the returned object’s __call__ method, changing the value of algorithm='encrypt' does nothing, since the AddRoundKey round component function is its own inverse.

sage: with_encrypt = ark_pc(1, 1, algorithm='encrypt')
sage: with_decrypt = ark_pc(1, 1, algorithm='decrypt')
sage: with_encrypt == with_decrypt
True


When invoking the returned object’s __call__ method, one can change the round subkey used in the returned polynomial by changing the round=0 keyword.

sage: ark_pc(2, 1, round=7)
a21 + k721


When passing the returned object to methods such as apply_poly and compose, we can make these methods use a non-default value for round=0 by passing in a dictionary mapping round to a different value.

sage: rgf.apply_poly(rgf.state_vrs, ark_pc,
....: poly_constr_attr={'round' : 6})
[a00 + k600 a01 + k601 a02 + k602 a03 + k603]
[a10 + k610 a11 + k611 a12 + k612 a13 + k613]
[a20 + k620 a21 + k621 a22 + k622 a23 + k623]
[a30 + k630 a31 + k631 a32 + k632 a33 + k633]

sage: rcpc = rgf.compose(ark_pc, ark_pc,
....: f_attr={'round' : 3}, g_attr={'round' : 5})
sage: rcpc(3, 1)
a31 + k331 + k531

apply_poly(state, poly_constr, algorithm='encrypt', keys=None, poly_constr_attr=None)

Returns a state matrix where poly_method is applied to each entry.

INPUT:

• state – The state matrix over $$\GF{2^8}$$ to which poly_method is applied to.
• poly_constr – The Round_Component_Poly_Constr object to build polynomials during evaluation.
• algorithm – (default: “encrypt”) Passed directly to rcpc to select encryption or decryption. The encryption flag is “encrypt” and the decrypt flag is “decrypt”.
• keys – (default: None) An array of $$N_r$$ subkey matrices to replace any key variables in any polynomials returned by poly_method. Must be identical to the format returned by expand_key. If any polynomials have key variables and keys is not supplied, the key variables will remain as-is.
• poly_constr_attr – (default:None) A dictionary of keyword attributes to pass to rcpc when it is called.

OUTPUT:

• A state matrix in $$\GF{2^8}$$ whose $$i,j$$ th entry equals the polynomial poly_constr(i, j, algorithm, **poly_constr_attr) evaluated by setting its variables equal to the corresponding entries of state.

EXAMPLES:

sage: from sage.crypto.mq.rijndael_gf import RijndaelGF
sage: rgf = RijndaelGF(4, 4)
sage: state = rgf._hex_to_GF('3b59cb73fcd90ee05774222dc067fb68')
sage: result = rgf.apply_poly(state, rgf.shift_rows_poly_constr())
sage: rgf._GF_to_hex(result)
'3bd92268fc74fb735767cbe0c0590e2d'


Calling apply_poly with the Round_Component_Poly_Constr object of a round component (e.g. sub_bytes_poly) is identical to calling that round component function itself.

sage: state = rgf._hex_to_GF('4915598f55e5d7a0daca94fa1f0a63f7')
sage: apply_poly_result = rgf.apply_poly(state,
....: rgf.sub_bytes_poly_constr())
sage: direct_result = rgf.sub_bytes(state)
sage: direct_result == apply_poly_result
True


If the Round_Component_Poly_Constr object’s __call__ method returns a polynomial with state variables as well as key variables, we can supply a list of $$N_r$$ round keys keys whose elements are evaluated as the key variables. If this is not provided, the key variables will remain as is.:

sage: state = rgf._hex_to_GF('14f9701ae35fe28c440adf4d4ea9c026')
sage: key = rgf._hex_to_GF('54d990a16ba09ab596bbf40ea111702f')
sage: keys = rgf.expand_key(key)
sage: result = rgf.apply_poly(state,
sage: result == rgf.add_round_key(state, key)
True

k000 + (x^4 + x^2)


We can change the value of the keywords of poly_constr ‘s __call__ method when apply_poly calls it by passing in a dictionary poly_constr_attr mapping keywords to their values.

sage: rgf.apply_poly(rgf.state_vrs,
....: poly_constr_attr={'round' : 5})
[a00 + k500 a01 + k501 a02 + k502 a03 + k503]
[a10 + k510 a11 + k511 a12 + k512 a13 + k513]
[a20 + k520 a21 + k521 a22 + k522 a23 + k523]
[a30 + k530 a31 + k531 a32 + k532 a33 + k533]

block_length()

Returns the block length of this instantiation of Rijndael-GF.

EXAMPLES:

sage: from sage.crypto.mq.rijndael_gf import RijndaelGF
sage: rgf = RijndaelGF(4, 6)
sage: rgf.block_length()
4

compose(f, g, algorithm='encrypt', f_attr=None, g_attr=None)

Return a Round_Component_Poly_Constr object corresponding to $$g \circ f$$ or the polynomial output of this object’s __call__ method.

INPUT:

• f – A Round_Component_Poly_Constr object corresponding to a round component function $$f$$.
• g – A Round_Component_Poly_Constr object corresponding to a round component function $$g$$ or a polynomial output of this object’s __call__ method.
• algorithm – (default: “encrypt”) Whether f and g should use their encryption transformations or their decryption transformations. Does nothing if g is a Round_Component_Poly_Constr object. The encryption flag is “encrypt” and the decryption flag is “decrypt”.
• f_attr – (default: None) A dictionary of keyword attributes to pass to f when it is called.
• g_attr – (default: None) A dictionary of keyword attributes to pass to g when it is called. Does nothing if g is a polynomial.

OUTPUT:

• If g is a Round_Component_Poly_Constr object corresponding to a round component function $$g$$, then compose returns a Round_Component_Poly_Constr corresponding to the round component function $$g \circ f$$, where $$f$$ is the round component function corresponding to the first argument f. On the other hand, if g $$= g(A)_{i,j}$$ for a round component function $$g$$, then compose returns $$g(f(A))_{i,j}$$, where $$A$$ is an arbitrary input state matrix.

EXAMPLES

This function allows us to determine the polynomial representations of entries across multiple round functions. For example, if we wanted a polynomial representing the 1,3 entry of a matrix after we first apply ShiftRows and then MixColumns to that matrix, we do:

sage: from sage.crypto.mq.rijndael_gf import RijndaelGF
sage: rgf = RijndaelGF(4, 4)
sage: mcp = rgf.mix_columns_poly_constr()(1, 3); mcp
a03 + (x)*a13 + (x + 1)*a23 + a33
sage: result = rgf.compose(rgf.shift_rows_poly_constr(), mcp)
sage: result
a03 + (x)*a10 + (x + 1)*a21 + a32


We can test the correctness of this:

sage: state = rgf._hex_to_GF('fa636a2825b339c940668a3157244d17')
sage: new_state = rgf.shift_rows(state)
sage: new_state = rgf.mix_columns(new_state)
sage: result(state.list()) == new_state[1,3]
True


We can also use compose to build a new Round_Component_Poly_Constr object corresponding to the composition of multiple round functions as such:

sage: fn = rgf.compose(rgf.shift_rows_poly_constr(),
....: rgf.mix_columns_poly_constr())
sage: fn(1, 3)
a03 + (x)*a10 + (x + 1)*a21 + a32


If we use compose to make a new Round_Component_Poly_Constr object, we can use that object as input to apply_poly and compose:

sage: state = rgf._hex_to_GF('36400926f9336d2d9fb59d23c42c3950')
sage: result = rgf.apply_poly(state, fn)
sage: rgf._GF_to_hex(result)
'f4bcd45432e554d075f1d6c51dd03b3c'

sage: new_state = rgf.shift_rows(state)
sage: new_state = rgf.mix_columns(new_state)
sage: result == new_state
True

sage: fn2 = rgf.compose(rgf.sub_bytes_poly_constr(), fn)


If the second argument is a polynomial, then the value of algorithm is passed directly to the first argument $$f$$ during evaluation. However, if the second argument is a Round_Component_Poly_Constr object, changing algorithm does nothing since the returned object has its own algorithm='encrypt' keyword.

sage: f = rgf.compose(rgf.sub_bytes_poly_constr(),
....: rgf.mix_columns_poly_constr(), algorithm='decrypt')
sage: g = rgf.compose(rgf.sub_bytes_poly_constr(),
....: rgf.mix_columns_poly_constr())
sage: all(f(i,j) == g(i,j) for i in range(4) for j in range(4))
True


We can change the keyword attributes of the __call__ methods of f and g by passing dictionaries f_attr and g_attr to compose.

sage: fn = rgf.compose(rgf.add_round_key_poly_constr(),
....: f_attr={'round' : 4}, g_attr={'round' : 7})
sage: fn(1, 2)
a12 + k412 + k712

decrypt(ciphertext, key, format='hex')

Returns the ciphertext ciphertext decrypted with the key key.

INPUT:

• ciphertext – The ciphertext to be decrypted.
• key – The key to decrypt ciphertext with.
• format – (default: hex) The string format that both ciphertext and key must be in, either “hex” or “binary”.

OUTPUT:

• A string in the format format of ciphertext decrypted with key key.

EXAMPLES

sage: from sage.crypto.mq.rijndael_gf import RijndaelGF
sage: rgf = RijndaelGF(4, 4)
sage: key = '2dfb02343f6d12dd09337ec75b36e3f0'
sage: ciphertext = '54d990a16ba09ab596bbf40ea111702f'
sage: expected_plaintext = '1e1d913b7274ad9b5a4ab1a5f9133b93'
sage: rgf.decrypt(ciphertext, key) == expected_plaintext
True


We can also decrypt messages using binary strings.

sage: key = '00011010000011100011000000111101' * 4
sage: ciphertext = '00110010001110000111110110000001' * 4
sage: expected_plaintext = ('101111111010011100111100101010100111'
....: '1111010000101101100001101000000000000000010000000100111011'
....: '0100001111100011010001101101001011')
sage: result = rgf.decrypt(ciphertext, key, format='binary')
sage: result == expected_plaintext
True

encrypt(plain, key, format='hex')

Returns the plaintext plain encrypted with the key key.

INPUT:

• plain – The plaintext to be encrypted.
• key – The key to encrypt plain with.
• format – (default: hex) The string format of key and plain, either “hex” or “binary”.

OUTPUT:

• A string of the plaintext plain encrypted with the key key.

EXAMPLES:

sage: from sage.crypto.mq.rijndael_gf import RijndaelGF
sage: rgf = RijndaelGF(4, 4)
sage: key = 'c81677bc9b7ac93b25027992b0261996'
sage: plain = 'fde3bad205e5d0d73547964ef1fe37f1'
sage: expected_ciphertext = 'e767290ddfc6414e3c50a444bec081f0'
sage: rgf.encrypt(plain, key) == expected_ciphertext
True


We can encrypt binary strings as well.

sage: key = '10010111110000011111011011010001' * 4
sage: plain = '00000000101000000000000001111011' * 4
sage: expected_ciphertext = ('11010111100100001010001011110010111'
....: '1110011000000011111100100011011100101000000001000111000010'
....: '00100111011011001000111101111110100')
sage: result = rgf.encrypt(plain, key, format='binary')
sage: result == expected_ciphertext
True

expand_key(key)

Returns the expanded key schedule from key.

INPUT:

• key – The key to build a key schedule from. Must be a matrix over $$\GF{2^8}$$ of dimensions $$4 \times N_k$$.

OUTPUT:

• A length $$Nr$$ list of $$4 \times N_b$$ matrices corresponding to the expanded key. The $$n$$ th entry of the list corresponds to the matrix used in the add_round_key step of the $$n$$ th round.

EXAMPLES:

sage: from sage.crypto.mq.rijndael_gf import RijndaelGF
sage: rgf = RijndaelGF(4, 6)
sage: key = '331D0084B176C3FB59CAA0EDA271B565BB5D9A2D1E4B2892'
sage: key_state = rgf._hex_to_GF(key)
sage: key_schedule = rgf.expand_key(key_state)
sage: rgf._GF_to_hex(key_schedule[0])
'331d0084b176c3fb59caa0eda271b565'
sage: rgf._GF_to_hex(key_schedule[6])
'5c5d51c4121f018d0f4f3e408ae9f78c'

expand_key_poly(row, col, round)

Returns a polynomial representing the row,col th entry of the round th round key.

INPUT:

• row – The row position of the element represented by this polynomial.
• col – The column position of the element represented by this polynomial.

OUTPUT:

• A polynomial representing the row,col th entry of the round th round key in terms of entries of the input key.

EXAMPLES:

sage: from sage.crypto.mq.rijndael_gf import RijndaelGF
sage: rgf = RijndaelGF(4, 4)
sage: rgf.expand_key_poly(1, 2, 0)
k012
sage: rgf.expand_key_poly(1, 2, 1)
(x^2 + 1)*k023^254 +
(x^3 + 1)*k023^253 +
(x^7 + x^6 + x^5 + x^4 + x^3 + 1)*k023^251 +
(x^5 + x^2 + 1)*k023^247 +
(x^7 + x^6 + x^5 + x^4 + x^2)*k023^239 +
k023^223 +
(x^7 + x^5 + x^4 + x^2 + 1)*k023^191 +
(x^7 + x^3 + x^2 + x + 1)*k023^127 +
k010 +
k011 +
k012 +
(x^6 + x^5 + x)


It should be noted that expand_key_poly cannot be used with apply_poly or compose, since expand_key_poly is not a Round_Component_Poly_Constr object.

sage: rgf.compose(rgf.sub_bytes_poly_constr(), rgf.expand_key_poly)
Traceback (most recent call last):
...
TypeError: keyword 'g' must be a Round_Component_Poly_Constr or a polynomial over Finite Field in x of size 2^8

sage: state = rgf._hex_to_GF('00000000000000000000000000000000')
sage: rgf.apply_poly(state, rgf.expand_key_poly)
Traceback (most recent call last):
...
TypeError: keyword 'poly_constr' must be a Round_Component_Poly_Constr

key_length()

Returns the key length of this instantiation of Rijndael-GF.

EXAMPLES:

sage: from sage.crypto.mq.rijndael_gf import RijndaelGF
sage: rgf = RijndaelGF(4, 8)
sage: rgf.key_length()
8

mix_columns(state, algorithm='encrypt')

Returns the application of MixColumns to the state matrix state.

INPUT:

• state – The state matrix to apply MixColumns to.
• algorithm – (default: “encrypt”) Whether to perform the encryption version of MixColumns, or its decryption inverse. The encryption flag is “encrypt” and the decryption flag is “decrypt”.

OUTPUT:

• The state matrix over $$\GF{2^8}$$ which is the result of applying MixColumns to state.

EXAMPLES:

sage: from sage.crypto.mq.rijndael_gf import RijndaelGF
sage: rgf = RijndaelGF(4, 4)
sage: state = rgf._hex_to_GF('cd54c7283864c0c55d4c727e90c9a465')
sage: result = rgf.mix_columns(state)
sage: rgf._GF_to_hex(result)
'921f748fd96e937d622d7725ba8ba50c'
sage: decryption = rgf.mix_columns(result, algorithm='decrypt')
sage: decryption == state
True

mix_columns_poly_constr()

Return a Round_Component_Poly_Constr object corresponding to MixColumns.

EXAMPLES:

sage: from sage.crypto.mq.rijndael_gf import RijndaelGF
sage: rgf = RijndaelGF(4, 4)
sage: mc_pc = rgf.mix_columns_poly_constr()
sage: mc_pc
A polynomial constructor for the function 'Mix Columns' of Rijndael-GF block cipher with block length 4, key length 4, and 10 rounds.
sage: mc_pc(1, 2)
a02 + (x)*a12 + (x + 1)*a22 + a32
sage: mc_pc(1, 0, algorithm='decrypt')
(x^3 + 1)*a00 + (x^3 + x^2 + x)*a10 + (x^3 + x + 1)*a20 + (x^3 + x^2 + 1)*a30


The returned object’s __call__ method has no additional keywords, unlike sub_bytes_poly_constr() and add_round_key_poly_constr().

number_rounds()

Returns the number of rounds used in this instantiation of Rijndael-GF.

EXAMPLES:

sage: from sage.crypto.mq.rijndael_gf import RijndaelGF
sage: rgf = RijndaelGF(5, 4)
sage: rgf.number_rounds()
11

shift_rows(state, algorithm='encrypt')

Returns the application of ShiftRows to the state matrix state.

INPUT:

• state – A state matrix over $$\GF{2^8}$$ to which ShiftRows is applied to.
• algorithm – (default: “encrypt”) Whether to perform the encryption version of ShiftRows or its decryption inverse. The encryption flag is “encrypt” and the decryption flag is “decrypt”.

OUTPUT:

• A state matrix over $$\GF{2^8}$$ which is the application of ShiftRows to state.

EXAMPLES:

sage: from sage.crypto.mq.rijndael_gf import RijndaelGF
sage: rgf = RijndaelGF(4, 4)
sage: state = rgf._hex_to_GF('adcb0f257e9c63e0bc557e951c15ef01')
sage: result = rgf.shift_rows(state)
sage: rgf._GF_to_hex(result)
sage: decryption = rgf.shift_rows(result, algorithm='decrypt')
sage: decryption == state
True

shift_rows_poly_constr()

Return a Round_Component_Poly_Constr object corresponding to ShiftRows.

EXAMPLES:

sage: from sage.crypto.mq.rijndael_gf import RijndaelGF
sage: rgf = RijndaelGF(4, 4)
sage: sr_pc = rgf.shift_rows_poly_constr()
sage: sr_pc(3, 0)
a33
sage: sr_pc(2, 1, algorithm='decrypt')
a23


The returned object’s __call__ method has no additional keywords, unlike sub_bytes_poly_constr() and add_round_key_poly_constr.

sub_bytes(state, algorithm='encrypt')

Returns the application of SubBytes to the state matrix state.

INPUT:

• state – The state matrix to apply SubBytes to.
• algorithm – (default: “encrypt”) Whether to apply the encryption step of SubBytes or its decryption inverse. The encryption flag is “encrypt” and the decryption flag is “decrypt”.

OUTPUT:

• The state matrix over $$\GF{2^8}$$ where SubBytes has been applied to every entry of state.

EXAMPLES:

sage: from sage.crypto.mq.rijndael_gf import RijndaelGF
sage: rgf = RijndaelGF(4, 4)
sage: result = rgf.sub_bytes(state)
sage: rgf._GF_to_hex(result)
'3e1c22c0b6fcbf768da85067f6170495'
sage: decryption = rgf.sub_bytes(result, algorithm='decrypt')
sage: decryption == state
True

sub_bytes_poly_constr()

Return the Round_Component_Poly_Constr object corresponding to SubBytes.

EXAMPLES:

sage: from sage.crypto.mq.rijndael_gf import RijndaelGF
sage: rgf = RijndaelGF(4, 4)
sage: sb_pc = rgf.sub_bytes_poly_constr()
sage: sb_pc
A polynomial constructor for the function 'SubBytes' of Rijndael-GF block cipher with block length 4, key length 4, and 10 rounds.
sage: sb_pc(2, 3)
(x^2 + 1)*a23^254 +
(x^3 + 1)*a23^253 +
(x^7 + x^6 + x^5 + x^4 + x^3 + 1)*a23^251 +
(x^5 + x^2 + 1)*a23^247 +
(x^7 + x^6 + x^5 + x^4 + x^2)*a23^239 +
a23^223 +
(x^7 + x^5 + x^4 + x^2 + 1)*a23^191 +
(x^7 + x^3 + x^2 + x + 1)*a23^127 +
(x^6 + x^5 + x + 1)


The returned object’s __call__ method has an additional keyword of no_inversion=False, which causes the returned polynomial to represent only the affine transformation step of SubBytes.

sage: sb_pc(1, 0, no_inversion=True)
(x^7 + x^3 + x^2 + x + 1)*a10^128 +
(x^7 + x^5 + x^4 + x^2 + 1)*a10^64 +
a10^32 +
(x^7 + x^6 + x^5 + x^4 + x^2)*a10^16 +
(x^5 + x^2 + 1)*a10^8 +
(x^7 + x^6 + x^5 + x^4 + x^3 + 1)*a10^4 +
(x^3 + 1)*a10^2 +
(x^2 + 1)*a10 +
(x^6 + x^5 + x + 1)


We can build a polynomial representing the inverse transformation by setting the keyword algorithm='decrypt'. However, the order of the affine transformation and the inversion step in SubBytes means that this polynomial has thousands of terms and is very slow to compute. Hence, if one wishes to build the decryption polynomial with the intention of evaluating that polynomial for a particular input, it is strongly recommended to first call sb_pc(i, j, algorithm='decrypt', no_inversion=True) to build a polynomial representing only the inverse affine transformation, evaluate this polynomial for your intended input, then finally calculate the inverse of the result.

sage: poly = sb_pc(1, 2, algorithm='decrypt', no_inversion=True)
sage: state = rgf._hex_to_GF('39daee38f4f1a82aaf432410c36d45b9')
sage: result = poly(state.list())
sage: rgf._GF_to_hex(result * -1)
'49'


When passing the returned object to apply_poly and compose, we can make those methods change the keyword no_inversion of this object’s __call__ method by passing the dictionary {'no_inversion' : True} to them.

sage: result = rgf.apply_poly(state, sb_pc,
....: poly_constr_attr={'no_inversion' : True})
sage: rgf._GF_to_hex(result)

sage: rcpc = rgf.compose(sb_pc, rgf.shift_rows_poly_constr(),

Note that if we set algorithm='decrypt' for apply_poly, it will perform the necessary performance enhancement described above automatically. The structure of compose, however, unfortunately does not allow this enhancement to be employed.