A simple symmetric encryption plugin for individual fields. The goal of this plugin is to encrypt data but still allow searching over fields with string values. This plugin relies on the Node crypto
module. Encryption and decryption happen transparently during save and find.
While this plugin works on individual fields of any type, note that for non-string fields, the original value is set to undefined after encryption. This is because if the schema has defined a field as an array, it would not be possible to replace it with a string value.
As of the stable 2.3.0 release, this plugin requires provision of a custom salt generation function (which would always provide a constant salt given the secret) in order to retain symmetric decryption capability.
Also consider mongoose-encryption if you are looking to encrypt the entire document.
Encryption is performed using AES-256-CBC
. To encrypt, the relevant fields are encrypted with the provided secret + random salt (or a custom salt via the provided saltGenerator
function). The generated salt and the resulting encrypted value is concatenated together using a :
character and the final string is put in place of the actual value for string
values. An extra boolean
field with the prefix __enc_
is added to the document which indicates if the provided field is encrypted or not.
Fields which are either objects or of a different type are converted to strings using JSON.stringify
and the value stored in an extra marker field of type string
with a naming scheme of __enc_
as prefix and _d
as suffix on the original field name. The original field is then set to undefined
. Please note that this might break any custom validation and application of this plugin on non-string fields needs to be done with care.
- Node
>=6
(Use2.3.4
for Node>=4.4.7 && <=6.x.x
) - MongoDB
>=2.6.10
- Mongoose
>=4.0.0
npm install mongoose-field-encryption --save-exact
- Always store your keys and secrets outside of version control and separate from your database. An environment variable on your application server works well for this.
- Additionally, store your encryption key offline somewhere safe. If you lose it, there is no way to retrieve your encrypted data.
- Encrypting passwords is no substitute for appropriately hashing them.
bcrypt
is one great option. You can also encrypt the password afer hashing it although it is not necessary. - If an attacker gains access to your application server, they likely have access to both the database and the key. At that point, neither encryption nor authentication do you any good.
For example, given a schema as follows:
const mongoose = require("mongoose");
const mongooseFieldEncryption = require("mongoose-field-encryption").fieldEncryption;
const Schema = mongoose.Schema;
const PostSchema = new Schema({
title: String,
message: String,
references: {
author: String,
date: Date
}
});
PostSchema.plugin(mongooseFieldEncryption, { fields: ["message", "references"], secret: "some secret key" });
const Post = mongoose.model("Post", PostSchema);
const post = new Post({ title: "some text", message: "hello all" });
post.save(function(err) {
console.log(post.title); // some text (only the message field was set to be encrypted via options)
console.log(post.message); // a9ad74603a91a2e97a803a367ab4e04d:93c64bf4c279d282deeaf738fabebe89
console.log(post.__enc_message); // true
});
The resulting documents will have the following format:
{
_id: ObjectId,
title: String,
message: String, // encrypted salt and hex value as string, e.g. 9d6a0ca4ac2c80fc84df0a06de36b548:cee57185fed78c055ed31ca6a8be9bf20d303283200a280d0f4fc8a92902e0c1
__enc_message: true, // boolean marking if the field is encrypted or not
references: undefined, // encrypted object set to undefined
__enc_references: true, // boolean marking if the field is encrypted or not
__enc_references_d: String // encrypted salt and hex object value as string, e.g. 6df2171f25fd1d32adc4a4059f867a82:5909152856cf9cdb7dc32c6af321c8fe69390c359c6b19d967eaa6e7a0a97216
}
find
works transparently and you can make new documents as normal, but you should not use the lean
option on a find if you want the fields of the document to be decrypted. findOne
, findById
and save
also all work as normal. update
works only for string fields and you would also need to manually set the __enc_
field value to false if you're updating an encrypted field.
From the mongoose package documentation: Note that findAndUpdate/Remove do not execute any hooks or validation before making the change in the database. If you need hooks and validation, first query for the document and then save it.
Note that as of 1.2.0
release, support for findOneAndUpdate
has also been added. Note that you would need to specifically set the encryption field marker for it to be encrypted. For example:
Post.findOneAndUpdate({ _id: postId }, { $set: { message: "snoop", __enc_message: false } });
The above also works for non-string fields. See changelog for more details.
Also note that if you manually set the value __enc_
prefix field to true then the encryption is not run on the corresponding field and this may result in the plain value being stored in the db.
Note that in order to use this option a fixed salt generator must be provided. See example as follows:
const messageSchema = new Schema({
title: String,
message: String,
name: String
});
messageSchema.plugin(mongooseFieldEncryption, {
fields: ["message", "name"],
secret: "some secret key",
saltGenerator: function(secret) {
return "1234567890123456"; // should ideally use the secret to return a string of length 16
}
});
const title = "some text";
const name = "victor";
const message = "hello all";
const Message = mongoose.model("Message", messageSchema);
const messageToSave = new Message({ title, message, name });
await messageToSave.save();
// note that we are only providing the field we would like to search with
const messageToSearchWith = new Message({ name });
messageToSearchWith.encryptFieldsSync();
// `messageToSearchWith.name` contains the encrypted string text
const results = await Message.find({ name: messageToSearchWith.name });
// results is an array of length 1 (assuming that there is only 1 message with the name "victor" in the collection)
// and the message in the results array corresponds to the one saved previously
fields
(required): an array list of the required fieldssecret
(required): a string cipher which is used to encrypt the data (don't lose this!)useAes256Ctr
(optional, defaultfalse
): a boolean indicating whether the olderaes-256-ctr
algorithm should be used. Note that this is strictly a backwards compatibility feature and for new installations it is recommended to leave this at default.saltGenerator
(optional, defaultconst defaultSaltGenerator = secret => crypto.randomBytes(16);
): a function that should return either autf-8
encoded string that is 16 characters in length or aBuffer
of length 16. This function is also passed the secret as shown in the default function example.
For performance reasons, once the document has been encrypted, it remains so. The following methods are thus added to the schema:
encryptFieldsSync()
: synchronous call that encrypts all fields as given by the plugin optionsdecryptFieldsSync()
: synchronous call that decrypts encrypted fields as given by the plugin optionsstripEncryptionFieldMarkers()
: synchronous call that removes the encryption field markers (useful for returning documents without letting the user know that something was encrypted)
Multiple calls to the above methods have no effect, i.e. once a field is encrypted and the __enc_
marker field value is set to true then the ecrypt operation is ignored. Same for the decrypt operation. Of course if the field markers have been removed via the stripEncryptionFieldMarkers()
call, then the encryption will be executed if invoked.
To enable searching over the encrypted fields the encrypt
and decrypt
methods have also been exposed.
const fieldEncryption = require('mongoose-field-encryption')
const encrypted = fieldEncryption.encrypt('some text', 'secret'));
const decrypted = fieldEncryption.decrypt(encrypted, 'secret')); // decrypted = 'some text'
As of version 3.0.5, one can setup a local development mongodb instance using docker:
- copy
development/docker-compose-dev.yml
todevelopment/docker-compose.yml
- copy
development/init-mongo-dev.js
todevelopment/init-mongo.js
- run
docker-compose up
in thedevelopment
folder
Feel free to make changes to the default docker configuration as required.
- Install dependencies with
npm install
and install mongo if you don't have it yet. - Start mongo via
docker-compose up
under thedevelopment
folder. - Run tests with
npm run test:auth
. Additionally you can pass your own mongodb uri as an environment variable if you would like to test against your own database, for e.g.URI='mongodb://username:password@127.0.0.1:27017/mongoose-field-encryption-test' npm test
release-it patch,minor,major
npm version patch,minor,major
npm publish
- Update development dependencies, fix unit tests, no functionality affected
- Add development db via docker (3.0.5)
- BREAKING: Drop Node 4 support
- Update development dependencies, no functionality affected
- Update documentation, no functionality affected
- Update documentation, no functionality affected
- FEATURE: Add provision for a custom salt generator, PR #27. Note that by using a custom salt, fixed search capability is now restored.
- Update dependencies, no functionality affected
- FIX: Fix bug where decryption fails when the field in question is not retrieved, PR #26.
- FIX: Fix bug where data was not getting decrypted on a
find()
, #23.
-
BREAKING: Use
cipheriv
instead of plaincipher
, #17.Note that this might break any fixed search capability as the encrypted values are now based on a random salt.
Also note that while this version maintains backward compatibility, i.e. decryption will automatically fall back to using the
aes-256-ctr
algorithm, any further updates will lead to the value being encrypted with the salt. In order to fully maintain backwards compatibilty, an new optionuseAes256Ctr
has been introduced (defaultfalse
), which can be set totrue
to continue using the plugin as before. It is highly recommended to start using the newer algorithm however, see issue for more details.
- FEATURE: Added support for
findOneAndUpdate
wheresvic#20
- FEATURE: Added support for mongoose 5 wheresvic#16.
- FIX: Removed mongoose dependency, moved to
peerDependencies
. - Formatted source code using prettier.