js-encryptedfs v0.1.0

Overview

This library provides an Encrypted File System (EFS)

Chunks

Chunks consist of a an acutal data 'block' that is encrypted. It is also prepended with the salt, initialization vector and authorisation tag used to encrypt the data.

Below is a diagram showing the layout of the encrypted chunks.

Blocks

This is a constant sized amount (optionally user-specified) of business data. A large file is split into several block of block_size (generall 4k). This is to to allow random access reads and writies. For example to read a small section of a file, the entire file does not need to be decrpted Only the block(s) that contain the section you want to read. This does mean however, that there needs to be an IV for each block. This is because reusing IVs, or having predictable IVs is a security threat. It can lead to the ___ attack. TODO: which attack again? Perhaps for large executables, where you need to always read the file in its entirely, We can get rid of the block and IVs. But consider if it's really worth it because you're only saving kilobytes here.

Segments

Some amount of data equal or smaller than a block.

Encryption scheme

EFS uses AES-GCM symmetric encryption, which requires the derivation of a symmetric encryption key using pbkdf and a randomly generated salt and init vector. Sizes for these parameters follow NIST recommendations: https://csrc.nist.gov/publications/detail/sp/800-38d/final.

An authorisation tag based on chunk encryption is stored along side the salt and init vector. This provides a basic chunk level integrity gurantee that can be verified upon decryption in accordance with the AES-GCM algorithm.

Functionality

• Keys are never decrypted on disk, they are in fact decrypted in an in-memory file system
• Uses Symmetric Encryption
• Encryption keys are transparent to the user

Getting Started

TODO: Write instructions

Installation

Building from source

TODO: Write instructions

Development

TODO: Write instructions

Testing

This library uses Jest for testing.

Operator warnings

TODO

• Move EFS to using fdIndex and an internal file descriptor index
• Decide if we want to use process.chdir(...) for lowerDir cwd or just have a convenience function to turn a relative path (passed to EFS) into an absolute path that we can give to lowerDir
• There is an issue with setUid and getUid on _upperDir (vfs) currently its not behaving predictably, if you are uid and gid 1000 and then set both of these (should be redundantly) to 1000 again, you lose access permissions
• Make sure metadata is written/read correctly (i.e. same position for all operations)