cgcserv exposes each service's stdin and stdout on a tcp socket. Each new tcp connection spawns a new cgc process. The process is killed and the connection is closed only when the the client closes the socket (Results in process getting SIGPIPE). If the process exits successfully (some cgc applications may be "one-off" some may exit when given an appropriate command) cgcserv will re-spawn the process and keep the connection to the fuzzer open.

cgcserv is provided in a docker environment for convenience. The docker container takes care of building the cgc binaries and starting the cgcserver.

From the root directory,

make

This will execute the command:

docker build -t cgcserv .

Add proxies to ~/docker/config.json

See: https://docs.docker.com/network/proxy/

The server requires that ports are exposed. Each new session will start attached to a random port in the given range. From the root directory you can simply execute:

run.sh

which will execute the command:

docker run -it -p 5000-5100:5000-5100 cgcserv

To stop the server, type Ctl-C.

Ports can be mapped on the host if needed.

See: https://docs.docker.com/engine/reference/run/

This starts the main server which runs on port 5000 and waits for http requests to start services.

A web interface is provided to manage the server and provide coverage data. The interface is accessed by directing a browser to http://localhost:5000/. The server can also be managed by making direct GET requests to the server.

A session is a single service attached to a unique port and allows multiple clients to attach to it. Each new client spawns a new instance of the service. Coverage data is aggregated between all clients on the same session.

A session can be started by sending a GET request to the main server on port 5000.

'curl -X GET http://localhost:5000/session/start?service=SERVICE'

Where SERVICE is the name of the service to start. Arguments can be passed to the service by using '+' instead of spaces.

'curl -X GET http://localhost:5000/session/start?service=PTaaS+unused'

This returns a json struct as the response:

'{ sid:<SESSION_ID> port:<PORT_NO> }'

Where SESSION_ID is a unique ID for the new session and PORT_NO is the tcp port that the service was attached to.

A session can be stopped by sending a GET request to the main server on port 5000.

'curl -X GET http://localhost:5000/session/SESSION_ID/stop

Get details about a running session

'curl -X GET http://localhost:5000/session/SESSION_ID/details

Assuming that "fuzzer" generates tests on stdout:

fuzzer | nc localhost PORT_NO

Or more generally:

fuzzer | nc <ip_address> PORT_NO

Any output from the CGC server will appear on stdout.

To send service output back to the fuzzer:

socat EXEC:"fuzzer" TCP4:<ip_address>:PORT_NO

After starting the server, go to the /demos/PTaaS directory and execute:

start.sh

This will connect a fuzzer to the PTaaS service provided by the CGC server.

You can see live coverage results for this session by pointing your browser to:

http://localhost:5000