# NAME Dancer2::Plugin::Passphrase - Passphrases and Passwords as objects for Dancer2 # SYNOPSIS This plugin manages the hashing of passwords for Dancer2 apps, allowing developers to follow cryptography best practices without having to become a cryptography expert. It uses the bcrypt algorithm as the default, while also supporting any hashing function provided by [Digest](https://metacpan.org/pod/Digest) # MORE INFORMATION ## Purpose The aim of this module is to help you store new passwords in a secure manner, whilst still being able to verify and upgrade older passwords. Cryptography is a vast and complex field. Many people try to roll their own methods for securing user data, but succeed only in coming up with a system that has little real security. This plugin provides a simple way of managing that complexity, allowing developers to follow crypto best practice without having to become an expert. ## Rationale The module defaults to hashing passwords using the bcrypt algorithm, returning them in RFC 2307 format. RFC 2307 describes an encoding system for passphrase hashes, as used in the "userPassword" attribute in LDAP databases. It encodes hashes as ASCII text, and supports several passphrase schemes by starting the encoding with an alphanumeric scheme identifier enclosed in braces. RFC 2307 only specifies the `MD5`, and `SHA` schemes - however in real-world usage, schemes that are salted are widely supported, and are thus provided by this module. Bcrypt is an adaptive hashing algorithm that is designed to resist brute force attacks by including a cost (aka work factor). This cost increases the computational effort it takes to compute the hash. SHA and MD5 are designed to be fast, and modern machines compute a billion hashes a second. With computers getting faster every day, brute forcing SHA hashes is a very real problem that cannot be easily solved. Increasing the cost of generating a bcrypt hash is a trivial way to make brute forcing ineffective. With a low cost setting, bcrypt is just as secure as a more traditional SHA+salt scheme, and just as fast. Increasing the cost as computers become more powerful keeps you one step ahead For a more detailed description of why bcrypt is preferred, see this article: [http://codahale.com/how-to-safely-store-a-password/](http://codahale.com/how-to-safely-store-a-password/) ## Common Mistakes Common mistakes people make when creating their own solution. If any of these seem familiar, you should probably be using this module - Passwords are stored as plain text for a reason There is never a valid reason to store a password as plain text. Passwords should be reset and not emailed to customers when they forget. Support people should be able to login as a user without knowing the users password. No-one except the user should know the password - that is the point of authentication. - No-one will ever guess our super secret algorithm! Unless you're a cryptography expert with many years spent studying super-complex maths, your algorithm is almost certainly not as secure as you think. Just because it's hard for you to break doesn't mean it's difficult for a computer. - Our application-wide salt is "Sup3r\_S3cret\_L0ng\_Word" - No-one will ever guess that. This is common misunderstanding of what a salt is meant to do. The purpose of a salt is to make sure the same password doesn't always generate the same hash. A fresh salt needs to be created each time you hash a password. It isn't meant to be a secret key. - We generate our random salt using `rand`. `rand` isn't actually random, it's a non-unform pseudo-random number generator, and not suitable for cryptographic applications. Whilst this module also defaults to a PRNG, it is better than the one provided by `rand`. Using a true RNG is a config option away, but is not the default as it it could potentially block output if the system does not have enough entropy to generate a truly random number - We use `md5(pass.salt)`, and the salt is from `/dev/random` MD5 has been broken for many years. Commodity hardware can find a hash collision in seconds, meaning an attacker can easily generate the correct MD5 hash without using the correct password. - We use `sha(pass.salt)`, and the salt is from `/dev/random` SHA isn't quite as broken as MD5, but it shares the same theoretical weaknesses. Even without hash collisions, it is vulnerable to brute forcing. Modern hardware is so powerful it can try around a billion hashes a second. That means every 7 chracter password in the range \[A-Za-z0-9\] can be cracked in one hour on your average desktop computer. - If the only way to break the hash is to brute-force it, it's secure enough It is unlikely that your database will be hacked and your hashes brute forced. However, in the event that it does happen, or SHA512 is broken, using this module gives you an easy way to change to a different algorithm, while still allowing you to validate old passphrases