Downloads directory: http://digitalcorpora.org/downloads/tcpflow/

Most common GNU/Linux distributions ship tcpflow in their repositories. So on Debian/Ubuntu/etc you can say

sudo apt-get install tcpflow

and on Fedora/RedHat/CentOS/etc you can say

sudo dnf install tcpflow

And that's it. If this isn't good-enough for whatever reason, you can build from source:

To compile for Linux

Be sure you have the necessary precursors. There are files in the root directory that will do this for you, depending on your host operating system:

CONFIGURE_ARCH_17_8.sh CONFIGURE_FEDORA_18.sh CONFIGURE_FEDORA_26.sh CONFIGURE_UBUNTU_16_04.sh

Depending on your OS, just:

# sudo bash CONFIGURE_<YOUROS>.sh

Once you have configured your OS, compile and install with:

./configure
make
sudo make install

If you want do download the development tree with git, be sure to do a complete checkout with --recursive and then run bootstrap.sh, configure and make:

git clone --recursive https://github.com/simsong/tcpflow.git
cd tcpflow
bash bootstrap.sh
./configure
make
sudo make install  

To download and compile for Amazon AMI:

ssh ec2-user@<your ec2 instance>
sudo bash yum -y install git make gcc-c++ automake autoconf boost-devel cairo-devel libpcap-devel openssl-devel zlib-devel
git clone --recursive https://github.com/simsong/tcpflow.git
sh bootstrap.sh

To Compile for Windows with mingw on Fedora Core:

yum -y install mingw64-gcc mingw64-gcc-c++ mingw64-boost mingw64-cairo mingw64-zlib
mingw64-configure
make

To use CMake, see detailed instructions: cmake/README.md

From a clean repository as normal user (not root):

./bootstrap.sh     # Generates the file ./configure
./configure        # Generates the file tcpflow.spec
rpmbuild -bb tcpflow.spec --build-in-place

Check the specfile and resulted RPM:

rpmlint tcpflow.spec
rpmlint ~/rpmbuild/RPMS/x86_64/tcpflow-....rpm

Install:

sudo dnf install ~/rpmbuild/RPMS/x86_64/tcpflow-....rpm

tcpflow is a program that captures data transmitted as part of TCP connections (flows), and stores the data in a way that is convenient for protocol analysis and debugging. Each TCP flow is stored in its own file. Thus, the typical TCP flow will be stored in two files, one for each direction. tcpflow can also process stored 'tcpdump' packet flows.

tcpflow stores all captured data in files that have names of the form:

   [timestampT]sourceip.sourceport-destip.destport[--VLAN][cNNNN]

where: timestamp is an optional timestamp of the time that the first packet was seen T is a delimiter that indicates a timestamp was provided sourceip is the source IP address sourceport is the source port destip is the destination ip address destport is the destination port VLAN is the VLAN port c is a delimiter indicating that multiple connections are present NNNN is a connection counter, when there are multiple connections with the same [time]/sourceip/sourceport/destip/destport combination.
Note that connection counting rarely happens when timestamp prefixing is performed.

HERE are some examples:

   128.129.130.131.02345-010.011.012.013.45103

The contents of the above file would be data transmitted from host 128.129.131.131 port 2345, to host 10.11.12.13 port 45103.

   128.129.130.131.02345-010.011.012.013.45103c0005

The sixth connection from 128.129.131.131 port 2345, to host 10.11.12.13 port 45103.

   1325542703T128.129.130.131.02345-010.011.012.013.45103

A connection from 128.129.131.131 port 2345, to host 10.11.12.13 port 45103, that started on at 5:19pm (-0500) on January 2, 2012

   128.129.130.131.02345-010.011.012.013.45103--3

A connection from 128.129.131.131 port 2345, to host 10.11.12.13 port 45103 that was seen on VLAN port 3.

You can change the template that is used to create filenames with the -F and -T options. If a directory appears in the template the directory will be automatically created.

If you use the -a option, tcpflow will automatically interpret HTTP responses.

   If the output file is
      208.111.153.175.00080-192.168.001.064.37314,

   Then the post-processing will create the files:
      208.111.153.175.00080-192.168.001.064.37314-HTTP
      208.111.153.175.00080-192.168.001.064.37314-HTTPBODY

   If the HTTPBODY was compressed with GZIP, you may get a 
   third file as well:

      208.111.153.175.00080-192.168.001.064.37314-HTTPBODY-GZIP

   Additional information about these streams, such as their MD5
   hash value, is also written to the DFXML file

tcpflow is similar to 'tcpdump', in that both process packets from the wire or from a stored file. But it's different in that it reconstructs the actual data streams and stores each flow in a separate file for later analysis.

tcpflow understands sequence numbers and will correctly reconstruct data streams regardless of retransmissions or out-of-order delivery. However, tcpflow currently does not understand IP fragments; flows containing IP fragments will not be recorded properly.

tcpflow can output a summary report file in DFXML format. This file includes information about the system on which the tcpflow program was compiled, where it was run, and every TCP flow, including source and destination IP addresses and ports, number of bytes, number of packets, and (optionally) the MD5 hash of every bytestream.

tcpflow uses the LBL Packet Capture Library (available at ftp://ftp.ee.lbl.gov/libpcap.tar.Z) and therefore supports the same rich filtering expressions that programs like 'tcpdump' support. It should compile under most popular versions of UNIX; see the INSTALL file for details.

tcpflow is a useful tool for understanding network packet flows and performing network forensics. Unlike programs such as WireShark, which show lots of packets or a single TCP connection, tcpflow can show hundreds, thousands, or hundreds of thousands of TCP connections in context.

A common use of tcpflow is to reveal the contents of HTTP sessions. Using tcpflow you can reconstruct web pages downloaded over HTTP. You can even extract malware delivered as 'drive-by downloads.'

Jeremy Elson originally wrote this program to capture the data being sent by various programs that use undocumented network protocols in an attempt to reverse engineer those protocols. RealPlayer (and most other streaming media players), ICQ, and AOL IM are good examples of this type of application. It was later used for HTTP protocol analysis.

Simson Garfinkel founded Sandstorm Enterprises in 1998. Sandstorm created a program similar to tcpflow called TCPDEMUX and another version of the program called NetIntercept. Those programs are commercial. After Simson left Sandstorm he had need for a tcp flow reassembling program. He found tcpflow and took over its maintenance.

Please enter bugs on the github issue tracker

tcpflow currently does not understand IP fragments. Flows containing IP fragments will not be recorded correctly. IP fragmentation is increasingly a rare event, so this does not seem to be a significant problem.

If you are writing an article about tcpflow, please cite our technical report:

• Passive TCP Reconstruction and Forensic Analysis with tcpflow, Simson Garfinkel and Michael Shick, Naval Postgraduate School Technical Report NPS-CS-13-003, September 2013. https://calhoun.nps.edu/handle/10945/36026

Simson L. Garfinkel simsong@acm.org

I continue to port bulk_extractor, tcpflow, be13_api and dfxml to modern C++. After surveying the standards I’ve decided to go with C++17 and not C++14, as support for 17 is now widespread. (I probably don’t need 20). I am sticking with autotools, although there seems a strong reason to move to CMake. I am keeping be13_api and dfxml as a modules that are included, python-style, rather than making them stand-alone libraries that are linked against. I’m not 100% sure that’s the correct decision, though.

The project is taking longer than anticipated because I am also doing a general code refactoring. The main thing that is taking time is figuring out how to detangle all of the C++ objects having to do with parser options and configuration.

Given that tcpflow and bulk_extractor both use be13_api, my attention has shifted to using tcpflow to get be13_api operational, as it is a simpler program. I’m about three quarters of the way through now. I anticipate having something finished before the end of 2020.

--- Simson Garfinkel, October 18, 2020

Thanks to:

• Jeffrey Pang, for the radiotap implementation
• Doug Madory, for the Wifi parser
• Jeremy Elson, for the original idea and initial tcp/ip implementation