crypto-attacks

Introduction

Python implementations of cryptographic attacks and utilities.

Requirements

• SageMath with Python 3.9
• PyCryptodome

You can check your SageMath Python version using the following command:

$ sage -python --version
Python 3.9.0

If your SageMath Python version is older than 3.9.0, some features in some scripts might not work.

Usage

Unit tests are located in the test directory and can be executed using the unittest module or using pytest. This should not take very long, perhaps a few minutes depending on your machine.

To run a specific attack, you must add the code to the proper file before executing it.

Example

For example, you want to attack RSA using the Boneh-Durfee attack, with the following parameters (taken from test_rsa.py):

N = 88320836926176610260238895174120738360949322009576866758081671082752401596826820274141832913391890604999466444724537056453777218596634375604879123818123658076245218807184443147162102569631427096787406420042132112746340310992380094474893565028303466135529032341382899333117011402408049370805729286122880037249
e = 36224751658507610673165956970793195381480143363550601971796688201449789736497322700382657163240771111376677180786660893671085854060092736865293791299460933460067267613023891500397200389824179925263846148644777638774319680682025117466596019474987378275216579013846855328009375540444176771945272078755317168511

You add the following code at the bottom of the boneh_durfee.py file:

import logging

# Some logging so we can see what's happening.
logging.basicConfig(level=logging.DEBUG)

N = 88320836926176610260238895174120738360949322009576866758081671082752401596826820274141832913391890604999466444724537056453777218596634375604879123818123658076245218807184443147162102569631427096787406420042132112746340310992380094474893565028303466135529032341382899333117011402408049370805729286122880037249
e = 36224751658507610673165956970793195381480143363550601971796688201449789736497322700382657163240771111376677180786660893671085854060092736865293791299460933460067267613023891500397200389824179925263846148644777638774319680682025117466596019474987378275216579013846855328009375540444176771945272078755317168511
p_bits = 512
delta = 0.26

p, q = attack(N, e, p_bits, delta=delta, m=3)
assert p * q == N
print(f"Found {p = } and {q = }")

Then you can simply execute the file using Sage. It does not matter where you execute it from, the Python path is automagically set (you can also call the attacks from other Python files, but then you'll have to fix the Python path yourself):

[crypto-attacks]$ sage -python attacks/rsa/boneh_durfee.py
INFO:root:Trying m = 3, t = 1...
DEBUG:root:Generating shifts...
DEBUG:root:Creating a lattice with 11 shifts (order = 'invlex', sort_shifts_reverse = False, sort_monomials_reverse = False)...
DEBUG:root:Reducing a 11 x 11 lattice...
DEBUG:root:Reconstructing polynomials (divide_original = True, modulus_bound = False, divide_gcd = True)...
DEBUG:root:Polynomial at row 8 is constant, ignoring...
DEBUG:root:Reconstructed polynomial has gcd 1312232632720549890113031660369306919929075823824696839212183146130434668203517349691252841557097914064120078389640402109017308806168467714230057403815071456395553717020189622129706447677967264344568789118172311850383406340547579993263937406518074980025897726255316031512238322022839331135299265704052474541497687419350763703993630899191179705015113329644753599872380152055902238937889027950089072598069861391599563222633064848996619752054685734260976071760984100109990150069201501748622288840900421607423175114026653242500476408861976142751384898489130281755466581359057847077651502734556259387442296763474369957121 with polynomial at 8, dividing...
DEBUG:root:Reconstructed 10 polynomials
DEBUG:root:Computing pairwise gcds to find trivial roots...
DEBUG:root:Using Groebner basis method to find roots...
DEBUG:root:Sequence length: 10, Groebner basis length: 1
DEBUG:root:Sequence length: 9, Groebner basis length: 1
DEBUG:root:Sequence length: 8, Groebner basis length: 1
DEBUG:root:Sequence length: 7, Groebner basis length: 2
DEBUG:root:Found Groebner basis with length 2, trying to find roots...
Found p = 7866790440964395011005623971351568677139336343167390105188826934257986271072664643571727955882500173182140478082778193338086048035817634545367411924942763 and q = 11227048386374621771175649743442169526805922745751610531569607663416378302561807690656370394330458335919244239976798600743588701676542461805061598571009923

The parameters m and t as shown in the output log deserve special attention. These parameters are used in many lattice-based (small roots) algorithms to tune the lattice size. Conceptually, m (sometimes called k) and t represent the number of "shifts" used in the lattice, which is roughly equal or proportional to the number of rows. Therefore, increasing m and t will increase the size of the lattice, which also increases the time required to perform lattice reduction (currently using LLL). On the other hand, if m and t are too low, it is possible that the lattice reduction will not result in appropriate vectors, therefore wasting the time spent reducing. Hence, this is a trade-off.

In the current version of the project, m must always be provided by the user (the default value is set to 1). t can, in some cases, be computed based on the specific small roots method used by the attack. However it can still be tweaked by the user. In general, there are two ways to use these kinds of parameters:

• Implement a loop which starts at m = 1 until an answer is found (example below). This is a simple approach, but risks wasting time on futile computations with too small lattices.

m = 1
while True:
    res = attack(..., m=m)
    if res is not None:
        # The attack succeeded!
        break
    m += 1

• Implement a debug version of the attack you're trying to use (with known results), and determine the m value which results in good lattice vectors. Then directly call the attack method with the correct m value.

Implemented attacks

Approximate Common Divisor

• Multivariate polynomial attack [^acd_mp]
• Orthogonal based attack [^acd_ol]
• Simultaneous Diophantine approximation attack [^acd_sda]

CBC

• Bit flipping attack
• IV recovery attack
• Padding oracle attack

CBC + CBC-MAC

• Key reuse attack (encrypt-and-MAC)
• Key reuse attack (encrypt-then-MAC)
• Key reuse attack (MAC-then-encrypt)

CBC-MAC

• Length extension attack