dotnetrussell/ensemble

A Bug Bounty Platform that allows hunters to issue commands over a geo-distributed cluster. The ideal user is someone who is attempting to scan multiple bug bounty programs simultaneously, on a recurring basis.

ensemble's Issues

Feature: Push Notifications

Feature Summary: Users should be able to subscribe to push notifications for director notifications. Users should be able to choose what type of notifications they wish to receive and these notifications should either pop up as a toast in windows or as a push notification on the phone.

Acceptance Criteria:
Given a user wants to setup push notifications
When they navigate to the settings page
Then they will have the ability to activate push notifications and select what notifications they want to receive such as job status, and agent status

Given a user has enabled push notifications
When a notification is received by the browser
Then a push notification will be sent to the user

Redeploy Docker Images On Github Repo Update

Create a new application that shuts down and restarts a director or agent docker image when the main branch for Ensemble repo is updated.

More details on the Dashboard page

More details on the Dashboard page about jobs, workspaces, and agents

Use a WSGI enabled server instead of flask

Do not use flask as the primary server. It's not appropriate for production. Instead use a web server able to communicate with Flask through a WSGI protocol such as waitress. Waitress doesn't currently support TLS natively so another approach will need to be created for that also.

Agent Restore State

Currently if an agent dies mid job, the agent will not resume the job when the agent restarts. This is going to take some investigation to figure out a way to do this.

• Agents start a subprocess, wait for the process to finish, and then dump standard out to a temp file which is then processed and returned when requested. The issue with state is that if the agent dies mid job, the job still continues but now no longer belongs to the agent. So if the agent restarts there's no way for the agent to retrieve the output.

One possible solution to this could be running all commands with "tee" and then dumping the results to a results file named with the job id as it runs. Then when the agent restarts it could check some flat file that contains running job ids and retrieve the job results

Another possible solution (untested) is to track the pid of the process and then continuing to read from STDOUT by tailing /proc/pid/fd/1 but I believe this won't continue to capture the output between the restart and the tail. This is probably a backup solution

Separate the Webclient and Director Processes

The webclient process is currently started by the director process. Instead we should have a watchdog process that controls starting both webclient and director as well as restarting them if they crash.

Feature: Better Scheduled Job Result Diffing

Feature Summary: The job diffing that's currently enabled doesn't fully work as intended. The output should be sorted in some repeatable way prior to diffing and the diff output should have colors to help identify changes more easily.

Acceptance Criteria:
Given a user has run a scheduled job more than once
When the user views the scheduled job results
Then the user will see a diff of the results that are color coded (green for new red for removed) and sorted

Feature: Easy way to install new tools on agents

Feature Summary: Users should have a nice interface for installing new tools on agents, seeing what tools are installed, and uninstalling tools from agents. Currently users can install tools by creating a job but this is not the correct use of that interface.

Acceptance Criteria:
Given a user wants to install a new tool on the cluster or modify an existing one
When the user navigates to the settings page
Then the user will be presented with an interface that allows the user to install or delete existing apps in the cluster

Create Watchdog Process for Agents

Create a watchdog process for agents that communicates with the director.

It will

• Notify director when agent goes down and up
• Restart agent when agent dies automatically
• Restart agent at when director issues command

Looking for some data / stats on the geo distributed results

Hello Russell,

Just saw your defcon talk, really nice tool you're building here.

I guess you did it at first as a POC but would you mind sharing some compared results between regions?

It could be simple metrics on the same list of domains, like :

• with httpx, how many alived domains per regions?
• where the domain is alived, is there a content difference?
• with nmap, how many open ports per regions per domain, etc

Regards!

Get client and server docker files working.