nprintml 0.1.0

Fully automated traffic analysis with nPrint

nprintML

nprintML bridges the gap between nPrint, which generates standard fingerprints for packets, and AutoML, which allows for optimized model training and traffic analysis. nprintML enables users with network traffic and labels to perform optimized packet-level traffic analysis without writing any code.

Getting It

Dependencies

Python versions 3.6 through 3.8 are supported.

You might check what versions of Python are installed on your system, e.g.:

ls -1 /usr/bin/python*

As needed, consult your package manager or python.org.

Depending on your situation, consider pyenv for easy installation and management of arbitrary versions of Python.

nprintML further requires nPrint (see below).

Installation

nprintML itself is available for download from the Python Package Index (PyPI) and via pip:

python -m pip install nprintml

This downloads, builds and installs the nprintml console command. If you're happy to manage your Python (virtual) environment, you're all set with the above.

That said, installation of this command via a tool such as pipx is strongly encouraged. pipx will ensure that nprintML is installed into its own virtual environment, such that its third-party libraries do not conflict with any others installed on your system.

(Note that nPrint and nprintML are unrelated to the PyPI distribution named "nprint.")

Post-installation

nprintML depends on the nPrint command, which may be installed separately, (with reference to the nPrint documentation).

For quick-and-easy satisfaction of this requirement, nprintML supplies the bootstrapping command nprint-install, which is made available to your environment with nprintML installed. This command will inspect its execution environment and attempt to retrieve, compile and install nPrint with the most appropriate defaults:

nprint-install

nPrint may thereby be installed system-globally, to the user environment, to the (virtual) environment to which nprintML was installed, or to a specified path prefix. Consult the command's --help for more information.

nprint-install is identically available through its Python module (no different from pip above):

python -m nprintml.net.install

Using It

nprintML supplies the top-level shell command nprintml –

nprintml ...

– as well as its terse alias nml –

nml ...

In case of command path ambiguity and in support of debugging, the nprintml command is also available through its Python module:

python -m nprintml ...

The nPrintML traffic analysis pipeline is customizable. Traffic ingestion leverages nPrint, and as such supports its inputs. In addition, beyond a single PCAP file, nprintML may ingest multiple PCAP files and recursive directories of files, as outlined in the wiki.

A simple example involves per-packet machine learning given a single PCAP and IP address labels:

nprintml --ipv4 --pcap-file test.pcap --label-file labels.txt --aggregator index

The above instructs nprintML to execute a traffic analysis pipeline considering each packet in the file test.pcap as a sample, and to attach labels to each source IP address (nPrint's default index) as specified in labels.txt.

The label file should be formatted as follows:

Item,Label  # (optional header line)
IP1,label1
IP2,label2
IP3,label3
...

Through this labeling scheme we can attach labels to ports, ip addresses, and entire flows with nPrintML. For more information and advanced usage see the wiki.

Another example of using nPrintML is running a machine learning pipeline where every PCAP is considered to contain a single sample. The following command – (this time using terse aliases) – will create a machine learning pipeline using every PCAP file in the directory pcaps/ and the labels in labels.txt with IPv4 nPrints:

nml -4 --pcap-dir pcaps/ -L labels.txt -a pcap

The label file for the above follows the same format as in single PCAP usage, with only the Item column changing to specify file names as opposed to IP addresses:

item,label  # (optional header line)
path/name1.pcap,label1
path/name2.pcap,label2
path/name3.pcap,label3
...

Note that the path/ in the above example is the path relative to the directory specified by --pcap-dir, that is relative to the directory pcaps/.

Development

Development requirements may be installed via the dev extra (below assuming a source checkout):

pip install --editable .[dev]

(Note: the installation flag --editable is also used above to instruct pip to place the source checkout directory itself onto the Python path, to ensure that any changes to the source are reflected in Python imports.)

Development tasks are then managed via argcmdr sub-commands of manage …, (as defined by the repository module manage.py), e.g.:

manage version patch -m "initial release of nprintml" \
       --build                                        \
       --release