eDEX-UI Remote Monitoring Server (IN DEVELOPMENT/NOT WORKING YET)

Everything needed to setup a Linux server for remote monitoring with eDEX-UI. The basic idea is that SSH'ing into a server from eDEX would allow you to see all the UI information as if you were running from said server.

This is a work-in-progress, not-entirely-thought-of, research pending:tm: project that I decided to make public so users interested could watch my progress. Issues and PRs are disabled until a release comes out of this. Any questions, recommendations, ideas can be sent directly to my email: gabriel@saillard.dev.

Server security is ensured by:

Outright denying access to clients who do not already have an active connection pipe with the server
Reusing SSH keys to authentify remote monitoring requests
Transmitting over SSL

Current specs:

Prerequisite

Server must have eDEX remote backend running as root
Server must have users public keys stored somewhere (tbd)
Client must have server's IP and corresponding SSL CA certificate in storage (also tbd)
Client and Server must use SSH keys auth

Trigger

Client connects via ssh to server (in terminal, out of the eDEX UI scope)
Client detects connection and sees that the IP corresponds to an available key for monitoring
Client prompts user whether or not to try and initiate connection to remote monitoring, if SSH pipe is still open
If user confirms, connection attempt starts

Connection

Client reaches server's remote monitoring websocket
Server looks up client's IP address in its active remote connections lists
If a pipe is already opened with client; proceed, otherwise, connection is terminated
Server sends cryptographic challenge to client (encrypt random string with SSH pubkey)
Client decrypts the challenge and sends it back
If challenge is OK, proceed to linking, otherwise, connection terminated

Linking

Server forks a monitoring relay process with the permission level of the user (using setuid)
Once process is ready, server starts accepting monitoring requests from client
Client pipes all it's modules information queries to remote monitoring websocket
Queries are then piped by server to the correct relay process, and output is sent back encrypted with user's public key
Client decrypts incoming data and forwards it to modules
Server re-instates relay process if it ever crashes (unlikely, but failure recovery is always nice)

Breaking up

(tbd - how to monitor SSH pipe?)
- Possible solution: Monitor # of active connections with each IP, if after succesful pipe setup this number drops back to <2, close the pipe

Already implemented:

Websocket server with SSH support
Configuration file at ~/.config/eDEX-UI/RemoteServer/config.json
Deny connections that don't already have an active pipe (auto-whitelist)
Parse SSH public keys and create Q-A challenges
Managed monitoring relay processes (workers), launched with os-user perm level via setuid
- Serialized requests, organic processing flow via IPC (similar to eDEX multithread proxy but single-threaded)
  - Current tests show remote server overhead approx. 1ms/req
  - Performance on large-scale deployments serving monitoring to lots of eDEX clients currently untested, expected memory footprint +70M per remote connection
  - One connection, one relay process
Single-binary, ~0 deps build output

Currently the server is only capable of processing systeminformation requests. Linking client to server has been tested succesfully but there's still some work to do on connection trigger, authentication, breaking up and error handling.

ETA before 2020 Should be bundled with eDEX-UI v3 release.

bipolarmorgan / edex-ui-server