Getting started with Netzob

The goal of this tutorial is to present the usage of each main component of Netzob (inference of message format, construction of the state machine and generation of traffic) through an undocumented protocol.

You can download the protocol material here :

You can follow the tutorial with only the PCAP file. But, you will need the implementation if you want to generate traffic and allow Netzob to discuss with a real implementation.

Setting the Workspace

Just after installing Netzob, when you start it, you have to set the workspace directory (as in Eclipe).

Your first project

To create a project, navigate to the menu File > New project. Here, you can choose a project name which should be unique in the workspace.

Side note: by default, Netzob chooses a location inside a dedicated directory located in the “projects” directory of your current workspace. The newly created project is automatically selected which allow you to start working on it.

You can switch to another project at anytime through the use of the menu File > Open project from workspace. Do not forget to save your project before!

Capture traces

The first step in the inferring process of a protocol in Netzob is to capture and to import messages as samples. There are different methods to retrieve messages depending of the communication channel used (files, network, IPC, USB, etc.) and the format (PCAP, hex, raw binary flows, etc.).

For this tutorial, you can import network messages with the provided PCAP file. But we recommand to use the provided implementation to generate samples of traffic and capture them with Netzob. You can to this with the Netwok Capturer plugin, which is accessible is the menu File > Capture messages > Capture network traffic.

As shown on the picture, you have to launch the capture at the Layer 4 on the localhost lo interface. As the targeted protocol works over UDP, you’ll be able to capture only the UDP payloads. Then launch the server of the targeted protocol and then the client. This one will send different commands to the server and wait for the response.

Once you have captured one session, you have to select the messages you want to import (you should import everything) and click the Import messages button. A popup will ask you if you want to allow duplicate messages. It’s better to not do so, to avoid unnecessary messages. We recommend to repeat this import process 4 times, in order to have enough variation between messages.

Infer vocabulary

Let’s now start the inference of the message format (vocabulary).

The next picture shows the whole vocabulary inference interface and the intended meaning of each component.

The main window shows each message in raw hexadecimal format. You can play with visualization attributes : right click on the symbol, then select Visualization and the attribute you want to change (hex, decimal or even string format, the unit size and potentially the sign and endianess).

The following picture shows the rendering of the messages in hex format (on the left) and string format (on the right). You can then see that messages contain some interesting strings (api_identify, api_encrypt, api_decrypt, etc.).

You can use the filter functionality to display messages that contain a specific pattern. Here, we filter with the api_identify pattern.

This filter permits to easily retrieve the messages associated with a potential identification command.

You can see that a ‘#‘ character is present in each messages. You can try to split the messages by forcing their partitioning with a specific delimiter. To do so, use the Force partitioning functionality available in the symbol list (either with a right click on a symbol, or by selecting a symbol with its checkbox and then clicking on the Force partitioning button right above).

Using the ‘#‘ string delimiter, you’ll have the following result:

You may also want to play with Sequence alignment. This partitionment allows to align messages according to their common patterns.

After playing with the different partitionment available, you are able to retrieve the different commands associated with the targeted protocol, as shown on the following picture.

According to the name of the commands, you can see that a api_encrypt command is available. Let’s have a look at its message format, which looks like:

[command]#[dataToEncrypt][padding]

Netzob allows you to indicate that a specific field has a mutable content, which means its data is not fixed (such as the ‘#’ delimiter) nor part of a set of fixed elements (such as the command string).To specify the structure of a field and its attributes, right click on a field and select Edit Variable. A popup dialog displays a rooted tree that corresponds to the inferred structure of the field. For example, you should have all the observed values of the field (materialized through DataVariable leafs) under an AlternateVariable node variable.

Regarding the targeted protocol, as we want to allow any data for the current field, we first have to delete the AlternateVariableNode and modify the root node to a DataVariable that has a mutable behavior, as shown on the following picture.

You can visualize the associated message format on bottom-left corner. Its should display something like this:

Now that we have refined the api_encrypt command message, we have to do the same for other commands that also take as parameter a user data: api_identify, api_authentify and api_decrypt, but also for some response messages such as resp_decrypt and resp_encrypt.

At this time, you have a satisfactory approximation the vocabulary. You can now start to construct the state machine of the protocol.

Infer Grammar

In this tutorial, we won’t explain the automatic inference (learning) of the state machine. As the targeted protocol has a basic state machine, we will simply show how to model it in Netzob.

A basic state machine contains states and transitions. In Netzob, we use a complex structure to model the grammar of a protocol. This model allows to specify information such as the response time between an input symbol and an output symbol, or even the probability of the different output messages given an uniq input message. This model is called an SMMDT (Stochastic Mealy Machine with Deterministic Transitions).

The grammar perspective interface of Netzob allows you to create:

  • states (initial or not);
  • semi-stochastic transitions (i.e. “normal” transitions);
  • open channel transitions;
  • close channel transitions.

Regarding our targeted protocol, we construct the associated model with the following information:

  • 1 open channel transition and an initial state;
  • 1 close channel transition and a final state;
  • 4 main states: init, identified, authenticated, closed;
  • depending on the current state, we are able or not to launch certain commands;
  • some commands will trigger transitions (api_identify, api_authentify and api_bye).

Once modeled, this looks like:

Now that Netzob knows both the vocabulary and the grammar of the targeted protocol, we are able to generate traffic that respect the protocol model.

Generate traffic

Let’s go to the Simulator perspective of Netzob.

The simulator allows you to create either a client, a server or both. You can tell Netzob to talk with a real client or server implementation, or you can just launch a client and a server inside Netzob and let them talk together.

Let’s now create a client. We have to specify the following information:

  • client name;
  • initiator or not (i.e. who opens the communication channel ?): it will usally be yes for a client;
  • client or server side: client;
  • protocol: UDP for te targeted protocol;
  • bind IP: nothing here, as the client finds its own interface;
  • bind port: nothing here, as the client finds its own port;
  • target IP: 127.0.0.1;
  • target port: 4242.

Now start the real server implementation, select the client in Netzob and click the Start button on the top-right corner. This will generate and send commands to the real server, and you’ll be able to see the exchanged messages in the interface, as shown on the following picture.

After this introductive tutorial, we’ll be glade to have feedbacks and to help you (see our mailing list user@lists.netzob.org or ou IRC channel #netzob on Freenode).

If you want to go further and start contributing to Netzob, that’s perfect. There are many simple or complex tasks everyone can do: translation, documentation, bug fix, feature proposal or implementation.