Extended RandCrack – Advanced Python random module cracker / predictor

This library is based on randcrack but provides extended functionality, including support for 64-bit number submission and 64-bit state offsetting, making it even more powerful for reverse-engineering or simulating Python's random module.

What It Does

This script clones Python’s built-in random number generator based on the Mersenne Twister. After observing 624 outputs of 32-bit integers, it reconstructs the internal state and can then:

• Predict all future outputs
• Predict past outputs
• Handle 64-bit values directly with submit_64() and offset_64()

⚠️ The Mersenne Twister is not cryptographically secure. Do not use it for secure cryptographic applications. More info

Installation

pip install extendedrandcrack

Usage

You must provide exactly 624 32-bit integers or 312 64-bit integers to reconstruct the RNG state. Use either:

• submit(value) – for 32-bit values
• submit_64(value) – for 64-bit values (automatically splits into two 32-bit chunks)

Basic Example (32-bit)

import random
import time
from extendedrandcrack import RandCrack

random.seed(time.time())
rc = RandCrack()

for \_ in range(624):
rc.submit(random.getrandbits(32))

print("Random:", random.getrandbits(32))
print("Predicted:", rc.predict_getrandbits(32))

Using 64-bit Submissions

import random
from extendedrandcrack import RandCrack

rc = RandCrack()
random.seed(12345)
values = [random.getrandbits(64) for _ in range(312)]

for v in values:
rc.submit_64(v)

print("Next random 64-bit value:")
high = rc.predict_getrandbits(32)
low = rc.predict_getrandbits(32)
predicted = (high << 32) | low
print(predicted)

Predicting Past Outputs

You can go back in time using:

• offset(n) – where n is number of 32-bit steps
• offset_64(n) – where n is number of 64-bit steps

Negative values go back, positive values move forward.

Example – Rewind and Predict

import random
from extendedrandcrack import RandCrack

random.seed(42)
original = [random.getrandbits(32) for _ in range(10)]

rc = RandCrack()
for \_ in range(624):
rc.submit(random.getrandbits(32))

rc.offset(-624) # rewind to where we started
rc.offset(-10) # rewind before the known sequence

print("Original:", original)
print("Recovered:", [rc.predict_getrandbits(32) for _ in range(10)])

Accuracy Note on High-level Functions

The following prediction methods are available:

• predict_getrandbits(k)
• predict_randbelow(n)
• predict_randrange(start, stop[, step])
• predict_randint(a, b)
• predict_choice(seq)
• predict_random()

⚠️ randbelow(), randint(), randrange(), and choice() may consume more than one random number per call, so rewinding past these calls is imprecise. Use getrandbits() or random() for exact state tracking.

Testing

The script includes a test block that checks both forward and backward prediction across 1000 values and asserts 100% accuracy.

License

MIT. Based on original randcrack by tna0y, extended for 64-bit support and improved state handling.