About

This project is a port of Luca Mingarelli's cryptpandas package and allows users to easily encrypt and decrypt numpy arrays in addition to pandas dataframes.

Installation

You can install cryptnumpy with pip:

pip install cryptnumpy

NumPy example

You can encrypt and decrypt a single numpy array as follows:

import numpy as np
import cryptnumpy as crp

my_array = np.arange(10)

crp.to_encrypted(my_array, password='APassWord', path='file.crypt')
decrypted_array = crp.read_encrypted('file.crypt', password='APassWord')

print(np.all(my_array == decrypted_array))

A dictionary of numpy arrays works just the same:

import numpy as np
import cryptnumpy as crp

my_array_dict = dict(
    arry1=np.array(['foo', 'bar', 'baz']),
    arry2=np.array(['qux', 'quux'])
)

crp.to_encrypted(my_array_dict, password='APassWord', path='file.crypt')
decrypted_dict = crp.read_encrypted('file.crypt', password='APassWord')

for name, original_array in my_array_dict.items():
    decrypted_array = decrypted_dict[name]
    print(np.all(original_array == decrypted_array))

Pandas example

For convenience, cryptnumpy maintains the original functionality of Luca Mingarelli's cryptpandas package. Specifically, you can encrypt and decrypt a pandas dataframe as follows:

import pandas as pd
import cryptnumpy as crp

my_df = pd.DataFrame(
    {'A': [1, 2, 3],
     'B': ['foo', 'bar', 'baz']
     }
)

crp.to_encrypted(my_df, password='somePassword', path='file.crypt')
decrpyted_df = crp.read_encrypted(
    'file.crypt', 
    password='somePassword', 
    use_pandas=True 
)
print(decrpyted_df.equals(my_df))

Note that, by default, the read_encrypted function will assume your encrypted data is a numpy array. To load an encrypted dataframe instead, you need to set use_pandas=True (see example above).

If you are only looking to encrypt and decrypt dataframes, and do not need support for numpy arrays, you should install the original cryptpandas package. cryptnumpy will not offer you any additional functionalities in that case.

Requirements

• pandas
• numpy
• cryptography >= 41.0.4
• pyarrow >= 14.0.1