Password Strength

Password strength and validation.

Tutorial

Uppercase, Numbers, Special Characters

You test your passwords using the Policy object that controls what kind of password is acceptable in your system.

First, create the Policy object and define the rules that apply to passwords in your system:

from password_strength import PasswordPolicy

policy = PasswordPolicy.from_names(
    length=8,  # min length: 8
    uppercase=2,  # need min. 2 uppercase letters
    numbers=2,  # need min. 2 digits
    special=2,  # need min. 2 special characters
    nonletters=2,  # need min. 2 non-letter characters (digits, specials, anything)
)

Now, when you have the PasswordPolicy object, you can use it to test your passwords, and it will tell you which tests have failed:

policy.test('ABcd12!')
# -> [Length(8), Special(2)]

This tells us that 2 tests have failed: password is not long enough, and it does not have enough special characters. You can use this information to tell the user what precisely is wrong with their password.

policy.test('ABcd12!@')
# -> []

Empty list tells us that this password is alright.

This test, however, enabled uses to use passwords that have a lot of repetition.

So-Called Entropy Bits

Here's a test that's even better. You don't really need to define complex rules with special characters and stuff. All you actually need is a password that's long enough, complex enough, and easy to remember (see xkcd and Article: Everything We've Been Told About Passwords Is Wrong).

So, instead of defining all these rules, let's just require the password to be complex enough. Entropy bits is something that defines how much variety does your password have. '01111010010011' is long enough, but has only 2 entropy bits: that's how many bits you need to store its alphabet. However, a password that uses plenty of characters has more entropy.

policy = PasswordPolicy.from_names(
    entropybits=30  # need a password that has minimum 30 entropy bits (the power of its alphabet)
)

print(policy.test('0123456789'))
# -> []

This password is not long enough, or secure enough, but has enough entropy: its vocabulary has 10 different characters. Put this test together with other requirements to make sure there's no repetition in your passwords.

Complexity

Entropy bits are important, but difficult to understand. An even better, more intuitive test, is to require the password to be "complex enough". Complexity is a number in the range of 0.00..0.99. Good, strong passwords start at 0.66.

Let's first see how different passwords score:

from password_strength import PasswordStats

stats = PasswordStats('qwerty123')
print(stats.strength())  #-> Its strength is 0.316

stats = PasswordStats('G00dPassw0rd?!')
print(stats.strength())  #-> Its strength is 0.585

stats = PasswordStats('V3ryG00dPassw0rd?!')
print(stats.strength())  #-> Its strength is 0.767

So, 0.66 will be a very good indication of a good password. Let's implement our policy:

policy = PasswordPolicy.from_names(
    strength=0.66  # need a password that scores at least 0.5 with its strength
)

print(policy.test('V3ryG00dPassw0rd?!'))
# -> []  -- empty list means a good password

One good thing about using strength is that it allows users to use national aplhabets with passwords, which are most secure:

tested_pass = policy.password('Mixed-汉堡包/漢堡包, 汉堡/漢堡')
print(tested_pass.strength())  # -> 0.812 -- very good!
print(tested_pass.test())
#-> []  - good password; it actually scored 0.812

Notice how in the last example we use a different approach: policy.password() analyzes the password, and then we can both get its .strength(), and .test() it according to the current policy.

PasswordPolicy

Perform tests on a password.

Init Policy

PasswordPolicy(*tests)

Init password policy with a list of tests

Alternatively:

PasswordPolicy.from_names(**tests)

Init password policy from a dictionary of test definitions.

A test definition is simply:

{ test-name: argument } or { test-name: [arguments] }

Test name is just a lowercased class name.

Example:

PasswordPolicy.from_names(
    length=8,
    strength=(0.33, 30),
)

Bundled Tests

These objects perform individual tests on a password, and report True of False.

tests.EntropyBits(bits)

Test whether the password has >= bits entropy bits.

Entropy bits is the number of bits that is required to store the alphabet that's used in a password. It's a measure of how long is the alphabet.

tests.Length(length)

Tests whether password length >= length

tests.NonLetters(count)

Test whether the password has >= count non-letter characters

tests.NonLettersLc(count)

Test whether the password has >= count non-lowercase characters

tests.Numbers(count)

Test whether the password has >= count numeric characters

tests.Special(count)

Test whether the password has >= count special characters

tests.Strength(strength, weak_bits=30)

Test whether the password has >= strength strength.

A password is evaluated to the strength of 0.333 when it has weak_bits entropy bits, which is considered to be a weak password. Strong passwords start at 0.666.

tests.Uppercase(count)

Test whether the password has >= count uppercase characters

Testing

After the PasswordPolicy is initialized, there are two methods to test:

PasswordPolicy.password

password(password)

Get password stats bound to the tests declared in this policy.

If in addition to tests you need to get statistics (e.g. strength) -- use this object to double calculations.

See PasswordStats for more details.

PasswordPolicy.test

test(password)

Perform tests on a password.

Shortcut for: PasswordPolicy.password(password).test().

Custom Tests

ATest is a base class for password tests.

To create a custom test, just subclass it and implement the following methods:

• init() that takes configuration arguments
• test(ps) that tests a password, where ps is a PasswordStats object.

PasswordStats

PasswordStats allows to calculate statistics on a password.

It considers a password as a unicode string, and all statistics are unicode-based.

Constructor:

from password_strength import PasswordStats
PasswordStats(password)

PasswordStats.alphabet

Get alphabet: set of used characters

PasswordStats.alphabet_cardinality

Get alphabet cardinality: alphabet length

PasswordStats.char_categories

Character count per top-level category

The following top-level categories are defined:

• L: letter
• M: Mark
• N: Number
• P: Punctuation
• S: Symbol
• Z: Separator
• C: Other

PasswordStats.char_categories_detailed

Character count per unicode category, detailed format.

See: http://www.unicode.org/reports/tr44/#GC_Values_Table

PasswordStats.combinations

The number of possible combinations with the current alphabet

PasswordStats.count(*categories)

Count characters of the specified classes only

PasswordStats.count_except(*categories)

Count characters of all classes except the specified ones

PasswordStats.entropy_bits

Get information entropy bits: log2 of the number of possible passwords

https://en.wikipedia.org/wiki/Password_strength

PasswordStats.entropy_density

Get information entropy density factor, ranged {0 .. 1}.

This is ratio of entropy_bits() to max bits a password of this length could have. E.g. if all characters are unique -- then it's 1.0. If half of the characters are reused once -- then it's 0.5.

PasswordStats.length

Get password length

PasswordStats.letters

Count all letters

PasswordStats.letters_lowercase

Count lowercase letters

PasswordStats.letters_uppercase

Count uppercase letters

PasswordStats.numbers

Count numbers

PasswordStats.repeated_patterns_length

Detect and return the length of repeated patterns.

You will probably be comparing it with the length of the password itself and ban if it's longer than 10%

PasswordStats.sequences_length

Detect and return the length of used sequences:

• Alphabet letters: abcd...
• Keyboard letters: qwerty, etc
• Keyboard special characters in the top row: ~!@#$%^&*()_+
• Numbers: 0123456

PasswordStats.special_characters

Count special characters

Special characters is everything that's not a letter or a number

PasswordStats.strength(weak_bits=30)

Get password strength as a number normalized to range {0 .. 1}.

Normalization is done in the following fashion:

1. If entropy_bits <= weak_bits -- linear in range{0.0 .. 0.33} (weak)
2. If entropy_bits <= weak_bits*2 -- almost linear in range{0.33 .. 0.66} (medium)
3. If entropy_bits > weak_bits*3 -- asymptotic towards 1.0 (strong)

PasswordStats.test(tests)

Test the password against a list of tests

PasswordStats.weakness_factor

Get weakness factor as a float in range {0 .. 1}

This detects the portion of the string that contains:

• repeated patterns
• sequences

E.g. a value of 1.0 means the whole string is weak, and 0.5 means half of the string is weak.

Typical usage:

password_strength = (1 - weakness_factor) * strength