FIC Forensic CTF Platform
This is a CTF server for distributing and validating challenges. It is design to be robust, so it uses some uncommon technics like client certificate for authentication, lots of state of the art cryptographic methods and aims to be deployed in a DMZ network architecture.
This is a monorepo, containing several micro-services :
admin
is the web interface and API used to control the challenge and doing synchronization.checker
is an inotify reacting service that handles submissions checking.dashboard
is a public interface to explain and follow the conquest, aims to animate the challenge for visitors.evdist
is an inotify reacting service that handles settings changes during the challenge (eg. a 30 minutes event where hints are free, ...).generator
takes care of global and team's files generation.qa
is an interface dedicated to challenge development, it stores reports to be treated by challenges creators.receiver
is only responsible for receiving submissions. It is the only dynamic part accessibe to players, so it's codebase is reduce to the minimum. It does not parse or try to understand players submissions, it just write it down to a file in the file system. Parsing and treatment is made by thechecker
.remote/challenge-sync-airbus
is an inotify reacting service that allows us to synchronize scores and exercice validations with the Airbus scoring platform.remote/scores-sync-zqds
is an inotify reacting service that allows us to synchronize scores with the ZQDS scoring platform.repochecker
is a side project to check offline for synchronization issues.
Here is how thoses services speak to each others:
In the production setup, each micro-service runs in a dedicated container, isolated from each other. Moreover, two physical machines should be used:
phobos
communicates with players: displaying the web interface, authenticate teams and players, storing contest files and handling submissions retrieval without understanding them. It can't accessdeimos
so its job stops after writing requests on the filesystem.deimos
is hidden from players, isolated from the network. It can only accessphobos
via a restricted ssh connection, to retrieve requests fromphobos
filesystem and pushing to it newly generated static files.
Concretely, the L2 looks like this:
So, the general filesystem is organized this way:
DASHBOARD
contains files structuring the content of the dashboard screen(s).FILES
stores the contest file to be downloaded by players. To be accessible without authentication and to avoid bruteforce, each file is placed into a directory with a hashed name (the original file name is preserved). It's rsynced as is todeimos
.GENERATOR
contains a socket to allow other services to communicate with thegenerator
.PKI
takes care of the PKI used for the client certiciate authorization process, and more generaly, all authentication related files (htpasswd, dexidp config, ...). Only theshared
subdirectory is shared withdeimos
, private key and teams P12 don't go out.SETTINGS
stores the challenge config as wanted by admins. It's not always the config in use: it uses can be delayed waiting for a trigger.SETTINGSDIST
is the challenge configuration in use. It is the one shared with players.startingblock
keep thestarted
state of the challenge. This helpsnginx
to know when it can start distributing exercices related files.TEAMS
stores the static files generated by thegenerator
, there is one subdirectory by team (id of the team), plus some files at the root, which are common to all teams. There is also symlink pointing to team directory, each symlink represent an authentication association (certificate ID, OpenID username, htpasswd user, ...).submissions
is the directory where thereceiver
writes requests. It creates subdirectories at the name of the authentication association, as seen inTEAMS
,checker
then resolve the association regardingTEAMS
directory. There is also a special directory to handle team registration.
Here is a diagram showing how each micro-service uses directories it has access to (blue for read access, red for write access):
Local developer setup
Using Docker
Use docker-compose build
, then docker-compose up
to launch the infrastructure.
After booting, you'll be able to reach the main interface at: http://localhost:8042/ and the admin one at: http://localhost:8081/ (or at http://localhost:8042/admin/). The dashboard is available at http://localhost:8042/dashboard/ and the QA service at http://localhost:8042/qa/.
In this setup, there is no authentication. You are identfied as a team. On first use you'll need to register.
Import folder
Local import folder
The following changes is only required if your are trying to change the local import folder ~/fic
location.
Make the following changes inside this file docker-compose.yml
:
23 volumes:
24 - - ~/fic:/mnt/fic:ro
24 + - <custom-path-to-import-folder>/fic:/mnt/fic:ro
Git import
A git repository can be used:
29 - command: --baseurl /admin/ -localimport /mnt/fic -localimportsymlink
29 + command: --baseurl /admin/ -localimport /mnt/fic -localimportsymlink -git-import-remote git@gitlab.cri.epita.fr:ing/majeures/srs/fic/2042/challenges.git
Owncloud import folder
If your are trying to use the folder available with the Owncloud service, make the following changes inside this file docker-compose.yml
:
29 - command: --baseurl /admin/ -localimport /mnt/fic -localimportsymlink
29 + command: --baseurl /admin/ -clouddav=https://owncloud.srs.epita.fr/remote.php/webdav/FIC%202019/ -clouduser <login_x> -cloudpass '<passwd>'
Manual builds
Running this project requires a web server (configuration is given for nginx),
a database (currently supporting only MySQL/MariaDB), a Go compiler for the
revision 1.18 at least and a inotify
-aware system. You'll also need NodeJS to
compile some user interfaces.
-
Above all, you need to build Node projects:
cd frontend/fic; npm install && npm run build cd qa/ui; npm install && npm run build
-
First, you'll need to retrieve the dependencies:
go mod vendor
-
Then, build the three Go projects:
go build -o fic-admin ./admin go build -o fic-checker ./checker go build -o fic-dashboard ./dashboard go build -o fic-generator ./generator go build -o fic-qa ./qa go build -o fic-receiver ./receiver go build -o fic-repochecker ./repochecker ...
-
Before launching anything, you need to create a database:
mysql -u root -p <<EOF CREATE DATABASE fic; CREATE USER fic@localhost IDENTIFIED BY 'fic'; GRANT ALL ON fic.* TO fic@localhost; EOF
By default, expected credentials for development purpose is
fic
, for both username, password and database name. If you want to use other credentials, define the corresponding environment variable:MYSQL_HOST
,MYSQL_USER
,MYSQL_PASSWORD
andMYSQL_DATABASE
. Those variables are the one used by themysql
docker image, so just link them together if you use containers. -
Launch it!
./fic-admin &
After initializing the database, the server will listen on http://localhost:8081/: this is the administration part.
./fic-generator &
This daemon generates static and team related files and then waits another process to tell it to regenerate some files.
./fic-receiver &
This one exposes an API that gives time synchronization to clients and handle submission reception (but without treating them).
./fic-checker &
This service waits for new submissions (expected in
submissions
directory). It only watchs modifications on the file system, it has no web interface../fic-dashboard &
This last server runs the public dashboard. It serves all file, without the need of a webserver. It listens on port 8082 by default.
./fic-qa &
If you need it, this will launch a web interface on the port 8083 by default, to perform quality control.
For the moment, a web server is mandatory to serve static files, look
at the samples given in the configs/nginx
directory. You need to
pick one base configation flavor in the configs/nginx/base
directory, and associated with an authentication mechanism in
configs/nginx/auth
(named the file fic-auth.conf
in /etc/nginx
),
and also pick the corresponding configs/nginx/get-team
file, you
named fic-get-team.conf
.