super_mediator Tutorial

Overview

What is super_mediator?

super_mediator is an IPFIX mediator that ingests IPFIX data from YAF (via TCP, UDP, file(s), Spread, etc.) and exports to one or more collectors such as SiLK, CSV files, JSON files, or a MySQL database. Like a traditional IPFIX mediator, super_mediator can filter, modify, and aggregate the data as it passes through.

Why do you need a mediator?

You don't, if you just want to traditional flow data and don't need to modify, filter, or aggregate it any way. The standard tool chain works just fine. You will typically run YAF exporting IPFIX over TCP to rwflowpack which either writes flow data locally to your repository or sends it back to your data repository hosted on a separate server. See the SiLK Install Guide

super_mediator is used to collect the Deep Packet Inspection (DPI) data elements that YAF exports and write them in a format that can be used for bulk loading into a database or rotating CSV files.

super_mediator can be used to export to Orcus, a passive DNS analysis system. You don't NEED super_mediator to use Orcus. YAF can export directly to Orcus if your flow sensor is dedicated to monitoring DNS traffic. However, if you have single sensor that you hope to collect traditional flow data AND create a passive DNS repository, super_mediator can be used to forward the flow data to SiLK and the DNS-enriched flow data to Orcus. Alternatively, if you're concerned with PII or have limited bandwidth, super_mediator can perform de-duplication of DNS resource records. The deduplication removes attribution of DNS queries but can reduce the data signficantly.

Goal for this tutorial

This tutorial will provide examples of different exporters. It will show multiple configurations and the data that each one produces. The super_mediator.conf man page provides additional examples and more detailed documentation. This tutorial shows multiple methods of structuring your MySQL database and a few methods of import. The example only uses an IPFIX file as input. super_mediator has multiple methods of ingest. It can listen on a TCP or UDP port for connections from yaf (the preferred method) or poll a file directory for IPFIX files. Spread is also supported. See the super_mediator.conf man page for examples of different COLLECTOR configurations.

Dependencies

Core dependencies for yaf and super_mediator:

Tools:

Optional:

MySQL is needed if you want to store your data in a MySQL database. We will show a few examples of how to setup your MySQL database if you choose to store the DPI data in MySQL.

SiLK IPset is only used if you want to filter on IP addresses. This may be due to privacy concerns or policy or you simply just want to limit the amount of data to keep around.

SiLK is the ideal tool for flow analysis. A MySQL database is not the appropriate tool for analyzing a lot of flow data. One of the examples will show how to export flow data from super_mediator to SiLK.

If you want some cool tools for analyzing DNS that work on top of an Oracle or PostgreSQL database, download and install Orcus.

How to install

The following tools can be downloaded from [https://tools.netsa.cert.org].

tar -xvzf libfixbuf-1.7.0.tar.gz
cd libfixbuf-1.7.0
./configure
make
make install

tar -xvzf yaf-2.7.0.tar.gz
cd yaf-2.7.0
./configure --enable-applabel --enable-plugins
make
make install

tar -xvzf super_mediator-1.1.0.tar.gz
cd super_mediator-1.1.0
./configure
make
make install

tar -xvzf silk-3.10.0.tar.gz
cd silk-3.10.0
./configure --with-libfixbuf=/usr/local/lib/pkgconfig --enable-ipv6
make
make install

Getting Started

We need data to get started. You can sniff your network and get the data that way, or if you already have a large pcap or a bunch of small pcaps, we can run yaf with Deep Packet Inspection (DPI) enabled on the PCAP file and and use super_mediator to analyze the flow and DPI data.
Let's start with a PCAP example just to experiment. It is a good idea to start with a small amount of data in order to determine the best configuration for your analysis needs.

Starting yaf:

$ yaf --in some_big_pcap.pcap \
    --out ipfix_file.yaf \
    --applabel \
    --max-payload=2048 \
    --plugin-name=/usr/local/lib/dpacketplugin.la \
    --verbose

In the unlikely event you run into an error message running yaf, here are a few common warning and error messages, and solutions on how to silence them.

Common warning and error messages:

Problem:

yaf: error while loading libraries: libairframe-2.7.0.so.4:
cannot open share object file: No such file or directory

Solution: Most likely YAF libraries were installed in a nonstandard location. Try running ldconfig or setting LD_LIBRARY_PATH to the location of libairframe.

Problem:

Couldn't open library "dnsplugin": file not found

Solution: Most likely YAF application labeling libraries were installed in a nonstandard location. Set LTDL_LIBRARY_PATH to the location of those libraries.

Interpreting yaf output

If you don't run into any of the above issues, you should see something like this after running yaf with the above options:

[2014-01-16 14:05:04] yaf starting
[2014-01-16 14:05:04] Initializing Rules From File: /usr/local/etc/yafApplabelRules.conf
[2014-01-16 14:05:04] Application Labeler accepted 36 rules.
[2014-01-16 14:05:04] Application Labeler accepted 0 signatures.
[2014-01-16 14:05:04] DPI Running for ALL Protocols
[2014-01-16 14:05:04] Reading packets from some_big_pcap.pcap
[2014-01-16 14:05:04] Initializing Rules from DPI File /usr/local/etc/yafDPIRules.conf
[2014-01-16 14:05:04] DPI rule scanner accepted 63 rules from the DPI Rule File
[2014-01-16 14:05:07] Processed 5921725 packets into 42096 flows:
[2014-01-16 14:05:07]   Mean flow rate 15599.44/s.
[2014-01-16 14:05:07]   Mean packet rate 2194402.64/s.
[2014-01-16 14:05:07]   Virtual bandwidth 7851.6178 Mbps.
[2014-01-16 14:05:07]   Maximum flow table size 10742.
[2014-01-16 14:05:07]   181 flush events.
[2014-01-16 14:05:07]   19580 asymmetric/unidirectional flows detected (46.51%)
[2014-01-16 14:05:07] Assembled 33328 fragments into 15414 packets:
[2014-01-16 14:05:07]   Expired 552 incomplete fragmented packets. (0.01%)
[2014-01-16 14:05:07]   Maximum fragment table size 41.
[2014-01-16 14:05:07] Rejected 201232 packets during decode: (3.29%)
[2014-01-16 14:05:07]   201232 due to unsupported/rejected packet type: (3.29%)
[2014-01-16 14:05:07]     201232 unsupported/rejected Layer 3 headers. (3.29%)
[2014-01-16 14:05:07]     196465 ARP packets. (3.21%)
[2014-01-16 14:05:07] yaf Exported 1 stats records.
[2014-01-16 14:05:07] yaf terminating

yaf tells you which application labeling rules file it used to grab its rules and signatures. 36 rules were loaded into YAF at startup time, 0 signatures. What's the difference between rules and signatures? Rules are first executed against flows that have a src/dst port matching the rule ID first. If there are no matches, yaf starts with the first rule listed in the file and continues down the list until it finds a match. A signature, on the other hand, is performed on every flow irregardless of the ports. yaf does not load any signatures by default. Signatures may be useful if you're looking for command and control strings in an HTTP request/response so that yaf looks for the string of interest instead of jumping to the HTTP regular expression and labeling it as 80.

You should also see the following line which indicates successful loading of the DPI plugin:

[2014-01-16 14:05:04] DPI Running for ALL Protocols

If that line is absent, or yaf couldn't find one of the application labeling plugins you will see an error such as:

Problem:

couldn't load requested plugin: missing function "ypGetMetaData" in \
/usr/local/lib/yaf/dpacketplugin.la plugin

Solution: You probably didn't run configure with --enable-applabel. Scroll to the top of this page, reread the directions for installing yaf and try again.

Problem:

couldn't load requested plugin: missing function "ypGetMetaData" in aolplugin plugin.

Solution: You didn't give the correct argument to --plugin-name. The only acceptable plugins to this command line option are dpacketplugin.la and dhcp_fp_plugin.la, unless you have written your own. Please scroll up, reread the directions for running yaf and try again.

Next, you'll see the PCAP file that you specified on the command line, the location of the yafDPIRules.conf that yaf is reading from, and how many accepted DPI rules yaf collected from it.

The remaining 16 lines provide statistics about what yaf found in the PCAP file and what it exported. This information was also exported in an IPFIX Options Record in the output file you provided when you ran yaf. You'll see that yaf ignored 201,232 packets. A majority of these packets were ARP (196,465) packets.

Using yafscii

A quick way to view the data is to use yafscii. yafscii takes IPFIX flow data files and prints them in an ASCII format that looks similar to tcpdump, with one flow per line. It can also write it in pipe-delimited format. yafscii ONLY PRINTS FLOW RECORDS. It will not print the DPI data that yaf exported. This is how the data should look:

$ yafscii --in ipfix_file.ipfix --out - | less
2011-01-28 22:00:22.230 - 22:00:25.763 (3.533 sec) tcp 74.125.113.109:993 => 10.
10.0.183:63216 89bcf80b AP/AP vlan 384 (10/850 ->) eof
2011-01-28 22:00:25.635 - 22:00:25.776 (0.141 sec) tcp 10.10.1.16:49356 => 64.23
3.169.193:443 5b17dc2d:097ab094 S/APS:AS/APS vlan 384:384 (14/1418 <-> 8/3876) r
tt 6 ms eof applabel: 443
2011-01-28 22:00:25.779 tcp 10.10.0.41:44520 => 184.168.85.77:443 9f849956 S/S v
lan 384 (2/120 ->) eof
2011-01-28 21:52:05.070 - 22:00:25.779 (500.709 sec) udp 172.16.0.1:1867 => 172.
16.0.160:9996 vlan 1f4 (1898/2772880 ->) eof
2011-01-28 21:45:25.899 - 22:00:25.779 (899.880 sec) udp 172.16.6.1:62499 => 172
.16.6.10:12000 vlan 1fa (3943/5764000 ->) eof

If you are not familiar with yafscii, check out the yafscii man page.

Super_mediator Command Line

Now let's see what kind of information super_mediator will produce. You can run super_mediator from the command line with a limited set of features or use the super_mediator.conf file and get the full set of capabilities. By default, super_mediator run from the command line will read and write an IPFIX file. In text mode (-m TEXT) it will decode the IPFIX and write to a text file in pipe-delimited format.

$ super_mediator -i ipfix_file.ipfix -o - -m TEXT

2011-01-28 21:45:28.636|2011-01-28 21:45:29.131|   0.495|   0.080|  6|
                  172.16.0.163|58367|      14|    3502|00|00:00:00:00:00:00|
                     128.121.146.100|   80|      13|    6916|00|00:00:00:00:00:0
0|       S|    APSF|      AS|    APSF|65d93ec9|508634fa|1f4|   80|000|000|
http|111|Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.6; en-US; rv:1.9.2.13) \  
         Gecko/20101203 Firefox/3.6.13
http|112|GET /account/available_features
http|115|http://twitter.com/
http|117|twitter.com
http|118|2094
http|123|203 Firefox
http|123|200 OK
http|123|200 OK
http|121|en-us,en;q=0.5
http|120|application/json, text
http|122|text/javascript
http|114|HTTP/1.1

Above is one particular HTTP flow. The first line is the flow information, similar to what yafscii outputs in --tabular mode. The following lines are the DPI data fields that yaf collected and exported. The first column identifies the protocol, the second column identifies the information element ID, and the third column displays the data that was stored in the information element. You could easily grep for all http flows in the file or user-agent strings (information element 111). A list of information element IDs can be found here along with a description of what each information element should contain.

Perhaps, we only want to see the IP addresses, ports, application label, and DPI data. From the command line:

$ super_mediator -i ipfix_file.ipfix -o - -m TEXT -f 0,1,4,5,7,73

172.16.0.163|128.121.146.100|58367|80|80|111|Mozilla/5.0 (Macintosh; U; \       
Intel Mac OS X 10.6; en-US; rv:1.9.2.13) Gecko/20101203 Firefox/3.6.13
172.16.0.163|128.121.146.100|58367|80|80|112|GET /account/available_features
172.16.0.163|128.121.146.100|58367|80|80|115|http://twitter.com/
172.16.0.163|128.121.146.100|58367|80|80|117|twitter.com
172.16.0.163|128.121.146.100|58367|80|80|118|2094
172.16.0.163|128.121.146.100|58367|80|80|123|203 Firefox
172.16.0.163|128.121.146.100|58367|80|80|123|200 OK
172.16.0.163|128.121.146.100|58367|80|80|123|200 OK
172.16.0.163|128.121.146.100|58367|80|80|121|en-us,en;q=0.5
172.16.0.163|128.121.146.100|58367|80|80|120|application/json, text
172.16.0.163|128.121.146.100|58367|80|80|122|text/javascript
172.16.0.163|128.121.146.100|58367|80|80|114|HTTP/1.1

As you can see it exports one line for each DPI data element yaf exported.

There are a few other basic command line options available, mainly for input/output, daemonizing, and logging. The super_mediator man page will give you all of the options available.

Configuration File

The vast majority of configuration options are only available through the the configuration file. The super_mediator.conf allows you to do all the basic configuration the command line options provide and more. You can configure automatic uploads into a MySQL database, filtering, multiple collectors and exporters, de-duplication, and more.

By default the super_mediator.conf is installed in /usr/local/etc. Let's walk through the default file:

The first block is the COLLECTOR block. This block configures how super_mediator should be listening for connections from yaf. You can configure multiple COLLECTOR blocks. As you can see this is listening for TCP connections on port 18000. This must match the yaf command line arguments --ipfix and --ipfix-port.

COLLECTOR TCP
   PORT 18000
COLLECTOR END

The second block is the first of 4 EXPORTER blocks. This particular EXPORTER uses the keyword FLOW_ONLY so that only flow is sent to the collecting process. This particular EXPORTER is meant for an rwflowpack process running on localhost listening for TCP connections on port 18001. This EXPORTER is given the name "silk" which will help identify log messages pertaining to this exporter in the log file. The name is optional. This should match the SiLK sensor.conf probe protocol and listen-on-port values. Exporter 1 Output

# rwflowpack - exporter 1
EXPORTER TCP "silk"
   PORT 18001
   HOST localhost
   FLOW_ONLY
EXPORTER END

The third block configures super_mediator to perform de-duplication of DNS resource records. It's writing the records to pipe-delimited text files to /data/dns and each file will have the prefix "yaf2dns" and the timestamp that the file was opened. The files will rotate every 1200 seconds (20 minutes). The files will be "locked" until super_mediator has finished writing to them. This means that the file will be prepended with a "." until the time limit expires. See below for more detail on how to further configure DNS deduplication.
Exporter 2 Output

#dedup process - exporter 2
EXPORTER TEXT "dns-dedup"
   PATH "/data/dns/yaf2dns"
   DELIMITER "|"
   ROTATE 1200
   DNS_DEDUP_ONLY
   LOCK
EXPORTER END

The third exporter block configures super_mediator to write only DPI data to protocol-specific, pipe-delimited files. The files will rotate every 20 minutes and they will be locked until super_mediator has finished writing to them. The files will be written to the /data/dpi directory. When using the MULTI_FILES keyword, PATH expects a file directory. The filenames will the protocol names, dns.txt0, http.txt1, smtp.txt2, etc. This EXPORTER could be used for external script or daemon that is loading the files into a database or simply for quick searching per protocol. Note that if the keyword TIMESTAMP_FILES is not present, the files will rollover approximately every 3.5 days.
Exporter 3 Output

#dpi 2 database - exporter 3
EXPORTER TEXT "dpi"
   PATH "/data/dpi"
   ROTATE 1200
   MULTI_FILES
   DPI_ONLY
   LOCK
EXPORTER END

The fourth exporter is meant for a custom collector. Perhaps, you have your own database schema and you didn't use the one that super_table_creator creates for you. You can specify the matching fields using the FIELDS keyword. This exporter writes to one particular file. It does not rollover. Exporter 4 Output

#custom field lists - exporter 4
EXPORTER TEXT "custom"
   PATH "/data/flow/custom.txt"
   FIELDS stime,etime,sip,dip,sport,dport,protocol,vlanint,iflags,\
          uflags,riflags,ruflags,application,DPI
EXPORTER END

The following block is commented out by default, but if uncommented it would effect the second EXPORTER block. An EXPORTER that uses the keyword MULTI_FILES, can further specify a DPI_CONFIG block to configure the filenames and which elements to write to that file. By default, super_mediator will write all the HTTP elements to the http file, all of the DNS resource records to the dns file, and so on. If you use the following DPI_CONFIG block, super_mediator will only write user-agent strings (111) and HTTP Get strings (112) to a file prefixed with "http." For DNS, super_mediator will only write A, NS, SOA, MX, and AAAA records to a file prefixed with "dns."

#DPI_CONFIG
#  TABLE http [111, 112]
#  TABLE dns [1, 2, 6, 12, 28]
#DPI_CONFIG END

The following block further configures the DNS de-duplication capability in super_mediator. By default, super_mediator will write a DNS record the first time it is seen and keep the record in memory for five minutes or until it has been seen 500 times. At that time, the record will be flushed from memory and the next time that record is seen it will be exported. You can change the default behavior using the DNS_DEDUP block. The below block would change the hit count from 500 to 10000. You could also use the FLUSH_TIME keyword to change the flush time from 5 minutes to something longer (or shorter). Additionally, there is a keyword, LAST_SEEN, which will change the format of the data and how the records are exported [see below].

DNS_DEDUP
   MAX_HIT_COUNT 10000
DNS_DEDUP END

This statement changes the default logging behavior to DEBUG.

LOGLEVEL DEBUG

This statement defines the location of the log file.

LOG "/var/log/super_mediator.log"

To configure super_mediator to rollover the log file each day, and to compress old log files, add the following statement to the configuration file (it must be a valid file directory):

LOGDIR "/var/log/super_mediator"

IPFIX File Example

To demonstrate the exporters defined in the configuration file, we will run an IPFIX file created by yaf. In order to use the IPFIX file created by yaf from the above example, we will need to change the COLLECTOR block to:

COLLECTOR FILEHANDLER
   PATH "ipfixfile.ipfix"
COLLECTOR END

Since we're just running super_mediator over a file, we may want to just log to the terminal, so comment out the following line:

#LOG "/var/log/super_mediator.log"

Create the file directories:

mkdir /data/flow
mkdir /data/dns
mkdir /data/dpi

We also need to start rwflowpack if we want to use this configuration file. This example will use the same sensor.conf created in this example. rwflowpack must be started prior to starting super_mediator:

/usr/local/sbin/rwflowpack --sensor-conf=/data/sensor.conf \
                   --root-dir=/data \
                   --log-destination-/var/log/rwflowpack.log \
                   --site-config=/data/silk.conf

If rwflowpack is not running properly, an error message such as the one below will be written in the log file:

[2014-01-18 19:07:43] Fatal: couldn't create connected TCP socket to localhost:18001 \
Connection refused

Otherwise, you'll see that super_mediator successfully read the given file. super_mediator will display log messages of every file it opens and closes, the YAF statistics records, and statistics about it's own process such as how many flows it read and wrote, filtered, and de-duplicated.

Running super_mediator with a configuration file:

$ /usr/local/bin/super_mediator -c /usr/local/etc/super_mediator.conf
[2015-06-29 14:48:16] super_mediator starting
[2015-06-29 14:48:16] custom: Opening Text File: /data/flow/custom.txt
[2015-06-29 14:48:16] custom: Exporter Active.
[2015-06-29 14:48:16] dpi: Exporter Active.
[2015-06-29 14:48:16] dns-dedup: Exporter Active.
[2015-06-29 14:48:16] silk: Exporter Active.
[2014-02-11 15:18:33] Initialization Successful, starting...
[2015-06-29 14:48:16] C1: Opening file: ipfixfile.ipfix
[2015-06-29 14:48:16] dpi: Opening Text File: /data/dpi/.http.txt0
[2015-06-29 14:48:16] dpi: Opening Text File: /data/dpi/.flow.txt0
[2015-06-29 14:48:16] dpi: Opening Text File: /data/dpi/.tls.txt0
[2015-06-29 14:48:16] dpi: Opening Text File: /data/dpi/.ftp.txt0
[2015-06-29 14:48:16] dpi: Opening Text File: /data/dpi/.ssh.txt0
[2015-06-29 14:48:16] dpi: Opening Text File: /data/dpi/.dns.txt0
[2015-06-29 14:48:16] dns-dedup: Opening Text File: /data/dns/.yaf2dns.20110128215027.txt
[2015-06-29 14:48:16] dpi: Opening Text File: /data/dpi/.rtp.txt0
[2015-06-29 14:48:17] dpi: Opening Text File: /data/dpi/.imap.txt0
[2015-06-29 14:48:17] dpi: Opening Text File: /data/dpi/.smtp.txt0
[2015-06-29 14:48:17] dpi: Opening Text File: /data/dpi/.pop3.txt0
[2015-06-29 14:48:17] dpi: Opening Text File: /data/dpi/.irc.txt0
[2015-06-29 14:48:17] dpi: Opening Text File: /data/dpi/.tftp.txt0
[2015-06-29 14:48:17] C1: YAF ID: 0 IP: 192.168.1.5 Uptime: 0d:0h:22m:29s
[2015-06-29 14:48:17] C1: YAF Flows: 42096 Packets: 5921725 Dropped: 0 Ignored: 201232 Out of Sequence: 0 Expired Frags: 552 Assembled Frags: 15414
[2015-06-29 14:48:17] C1: Closing Connection: End of file
[2015-06-29 14:48:17] INACTIVE Collector C1: 42096 flows, 0 other flows, 1 stats, 0 filtered, 0 files
[2015-06-29 14:48:17] Exporter custom: 42096 records, 0 stats, 117.7622 Mbps, 349.68 bytes per record
[2015-06-29 14:48:17] Exporter dpi: 42096 records, 0 stats, 65.3756 Mbps, 194.13 bytes per record
[2015-06-29 14:48:17] Exporter dns-dedup: 5899 records, 0 stats, 3.0931 Mbps, 65.54 bytes per record
[2015-06-29 14:48:17] Exporter silk: 42096 records, 0 stats, 67.3536 Mbps, 200.00 bytes per record
[2015-06-29 14:48:17] custom: Closing File /data/flow/custom.txt
[2015-06-29 14:48:17] Unlocking File /data/dpi/ftp.txt0
[2015-06-29 14:48:17] Unlocking File /data/dpi/ssh.txt0
[2015-06-29 14:48:17] Unlocking File /data/dpi/smtp.txt0
[2015-06-29 14:48:17] Unlocking File /data/dpi/dns.txt0
[2015-06-29 14:48:17] Unlocking File /data/dpi/tftp.txt0
[2015-06-29 14:48:17] Unlocking File /data/dpi/http.txt0
[2015-06-29 14:48:17] Unlocking File /data/dpi/imap.txt0
[2015-06-29 14:48:17] Unlocking File /data/dpi/irc.txt0
[2015-06-29 14:48:17] Unlocking File /data/dpi/pop3.txt0
[2015-06-29 14:48:17] Unlocking File /data/dpi/tls.txt0
[2015-06-29 14:48:17] Unlocking File /data/dpi/flow.txt0
[2015-06-29 14:48:17] Unlocking File /data/dpi/rtp.txt0
[2015-06-29 14:48:17] Exporter dns-dedup: 22285 DNS records, 3459 filtered, 5899 flushed (73.53% compression)
[2015-06-29 14:48:17] dns-dedup: Closing File /data/dns/yaf2dns.20110128215027.txt
[2015-06-29 14:48:17] Unlocking File /data/dns/yaf2dns.20110128215027.txt
[2015-06-29 14:48:17] SM: Uptime: 0d:0h:0m:0s, Total Flows: 42096, Filtered: 0, Stats: 1, DNS: 14345
[2015-06-29 14:48:17] super_mediator Terminating

As you can see, super_mediator provides a lot of log information in DEBUG mode. It logs every file open, close, lock, and unlock procedure. It provides statistics about how many flows it exported to each exporter. In addition, it writes the yaf statistics record to the log file. (To disable this, use the NO_STATS keyword in the super_mediator.conf file). Finally, it writes overall statistics about the flows, DNS flows, and the DNS records it has received, as well as how many DNS resource records that have been aggregated.

Exporter 1 Explained - SiLK

The first exporter sent flow data to SiLK. We can do some simple queries to determine how many flows it received. The following query uses rwfilter to query all the flows (the PCAP is from 2011). rwstats is used to calculate the protocols from most to least active:

$ export SILK_DATA_ROOTDIR=/data

$ rwfilter --start-date=2010/01/01 --end-date=2014/01/01 \
           --proto=0- --type=all --pass=stdout \
       | rwstats --fields=protocol --bottom --count=10
INPUT: 64501 Records for 5 Bins and 64501 Total Records
OUTPUT: Top 10 Bins by Records
pro|   Records|  %Records|   cumul_%|
  6|     32005| 49.619386| 49.619386|
 17|     31697| 49.141874| 98.761260|
  2|       399|  0.618595| 99.379855|
  1|       368|  0.570534| 99.950388|
 47|        32|  0.049612|100.000000|

Exporter 2 Explained - DNS Deduplication

The second exporter was the DNS de-duplication exporter.

$ head -n 6 /data/dns/yaf2dns.20110128215025.txt
2011-01-28 21:50:25.880|1|spreadsheets.l.google.com.|72.14.204.102
2011-01-28 21:50:25.880|1|spreadsheets.l.google.com.|72.14.204.100
2011-01-28 21:50:25.880|1|spreadsheets.l.google.com.|72.14.204.113
2011-01-28 21:50:25.880|1|spreadsheets.l.google.com.|72.14.204.101
2011-01-28 21:50:25.880|6|l.google.com.|ns2.google.com.
2011-01-28 21:50:36.205|6|l.google.com.|ns4.google.com.

The above lines are in the format:

first seen time | resource record type | rrname | rrval

As you can see the first 4 rows are A records, while the last 2 rows are SOA records.

What if we would like to see how many times a particular resource record appeared in the given time period? You would need to modify the super_mediator.conf DNS_DEDUP block. Add the LAST_SEEN keyword:

DNS_DEDUP
   MAX_HIT_COUNT 5000
   LAST_SEEN
   FLUSH_TIME 900
DNS_DEDUP END

The LAST_SEEN keyword instructs super_mediator to write records when they are flushed, rather than when they are first seen. When this option is used, the output format will also include the time the record was last seen and a count of the number of times it was seen. This particular PCAP is data collected over a 15 minute time period, so adjusting the timeout to 900 seconds will give us the hit count for the whole PCAP (unless of course there are more than 5000 records).

Rerunning super_mediator with the new configuration file yields the following results:

$ head -n 6 /data/dns/yaf2dns.20110128215106.txt 
2011-01-28 21:50:25.880|2011-01-28 21:50:25.880|1|spreadsheets.l.google.com.|1|72.14.204.102
2011-01-28 21:50:25.880|2011-01-28 21:50:25.880|1|spreadsheets.l.google.com.|1|72.14.204.100
2011-01-28 21:50:25.880|2011-01-28 21:50:25.880|1|spreadsheets.l.google.com.|1|72.14.204.113
2011-01-28 21:50:25.880|2011-01-28 21:50:25.880|1|spreadsheets.l.google.com.|1|72.14.204.101
2011-01-28 21:50:25.880|2011-01-28 21:50:25.880|6|l.google.com.|1|ns2.google.com.
2011-01-28 21:50:36.205|2011-01-28 21:50:36.205|6|l.google.com.|1|ns4.google.com.

The format of the above lines is:

first seen time | last seen time | rrtype | rrname | hitcount | rrval

For the above records, each record was only seen once in the 15 minute time period.

When performing DNS analysis with yaf and super_mediator, it is a good idea to add the following option to the yaf command line:

--udp-uniflow=53

The --udp-uniflow option instructs yaf to create one flow for each DNS packet. By default yaf only captures the payload from the first packet (in each direction) for a UDP flow. If you use the --udp-uniflow option, it ensures that you get every DNS request/response yaf processes.

See below for instructions on how to manually load this data into the MySQL database or here for instructions on how to setup automatic upload.

Exporter 3 Explained - MULTIFILES

The third exporter wrote a file for each protocol yaf captured information for. The MULTI_FILES option is ideal for database import. super_mediator writes the flow information to the flow.txt file and creates one file for each protocol. The flow information and the DPI information can be correlated using the flow key hash, the start time, and the observation ID, which together make up a primary key for the flow. The flow key hash is a hash of the 5-tuple and vlan. The 5-tuple is the source IP address, destination IP address, protocol, source transport port, and destination transport port.

For the most part, the protocol files have the same format:

flow key hash | start time | obID | Info Element ID | data

Field Description
flow key hash The hash of the 5-tuple and vlan
start time The start time (in milliseconds since EPOCH) of the flow
obID The observation ID of the exporting process (by default YAF uses 0)
Info Element ID The ID of the IPFIX Information Element that contained the data.
data The data that YAF collected.

The flow file will have the format:

flow key hash | start time | src IP | dst IP | protocol | src port | dst port | VLAN | obID

Field Description
flow key hash The hash of the 5-tuple and vlan
start time The start time (in milliseconds since EPOCH) of the flow
src IP source IP Address
dst IP destination IP Address
protocol transport protocol of the flow
src port source transport port
dst port destination transport port
vlan The VLAN tag for the flow
obID The observation ID of the exporting process (by default yaf uses 0)

DNS and TLS/SSL are the exceptions to the above format. DNS output has the format:

flow key hash | start time | obid | QR | dnsID | section | nx | authoritative | rrtype | ttl | rrname | rrval

Field Description
flow key hash The hash of the 5-tuple and vlan
start time The start time (in milliseconds since EPOCH) of the flow
obID The observation ID of the exporting process (by default YAF uses 0)
QR Q for Query, R for Response
dnsID The DNS transaction ID used to match Queries and Responses
section The section of the DNS record that this resource record was taken from.
0 for Question, 1 for Answer, 2 for Name Server, 3 for Additonal section.
nx The DNS NXDomain or Response Code. This corresponds with
the DNS RCODE header field. The field will be set to 3
for a Name Error, 2 for a Server Failure, 1 for a Format Error,
and 0 for No Error.
authoritative The Authoritative Header field, this bit is only
valid in responses and specifies that the name server is
an authority for the domain name in the question section.
rrtype The Type of Resource Record. See a list in super_mediator.conf man page.
ttl The DNS Time to Live.
rrname The QNAME from the DNS Question Section or NAME in the DNS Resource Record Section
rrval This depends on the type of the DNS Resource Record Type.

TLS/SSL files have the format:

flow key hash | start time | obID | Info Element ID | SI | cert number | data

Field Description
flow key hash The hash of the 5-tuple and vlan
start time The start time (in milliseconds since EPOCH) of the flow
obID The observation ID of the exporting process (by default yaf uses 0)
Info Element ID The object identifier from the X.509 RelativeDistinguishedName Sequence
SI S if the field was taken from the Subject Section of the Certificate,
I for the Issuer Section
cert number There may be multiple certificates in the certificate chain,
this lists the order in the sequence
data The X.509 object identifier value.

Here are some examples:

$ ls /data/dpi
dns.txt0   ftp.txt0    imap.txt0    pop3.txt0   ssh.txt0
flow.txt0  http.txt0   irc.txt0     smtp.txt0   tls.txt0

$ head -n5 /data/dpi/dns.txt0
1361713272|1296251349702|0|R|51786|1|0|0|5|69883|spreadsheets.google.com.|spreadsheets.l.google.com.
1361713272|1296251349702|0|R|51786|1|0|0|1|77|spreadsheets.l.google.com.|72.14.204.102
1361713272|1296251349702|0|R|51786|1|0|0|1|77|spreadsheets.l.google.com.|72.14.204.100
1361713272|1296251349702|0|R|51786|1|0|0|1|77|spreadsheets.l.google.com.|72.14.204.113
1361713272|1296251349702|0|R|51786|1|0|0|1|77|spreadsheets.l.google.com.|72.14.204.101

To find the corresponding DNS flow:

$ cat /data/dpi/flow.txt0 | grep 1361713272
1361713272|1296251349702|172.16.0.5|8.8.8.8|17|60019|53|500|0
    
$ cat /data/dpi/tls.txt0
2993670442|1296252025635|0|6|I|0|ZA
2993670442|1296252025635|0|10|I|0|Thawte Consulting (Pty) Ltd.
2993670442|1296252025635|0|3|I|0|Thawte SGC CA
2993670442|1296252025635|0|6|S|0|US
2993670442|1296252025635|0|8|S|0|California
2993670442|1296252025635|0|7|S|0|Mountain View
2993670442|1296252025635|0|10|S|0|Google Inc
2993670442|1296252025635|0|3|S|0|m.google.com
2993670442|1296252025635|0|189|I|0|2
2993670442|1296252025635|0|244|I|0|1a f7 d3 de 2a 67 9c 85 77 2c 93 f8 5d de 8b b4
2993670442|1296252025635|0|247|I|0|091218000000Z
2993670442|1296252025635|0|248|I|0|111218235959Z
2993670442|1296252025635|0|250|I|0|141
2993670442|1296252025635|0|187|I|0|0x0005
2993670442|1296252025635|0|186|I|0|3
2993670442|1296252025635|0|288|I|0|0x0301

To find the corresponding TLS flow:

$ cat /data/dpi/flow.txt0 | grep 2993670442
2993670442|1296252025635|10.10.1.16|64.233.169.193|6|49356|443|900|0

$ cat /data/dpi/http.txt0
3069422788|1296252022064|0|111|Apple iPhone v8A306 Maps v4.0.1
3069422788|1296252022064|0|111|Apple iPhone v8A306 Maps v4.0.1
3069422788|1296252022064|0|111|Apple iPhone v8A306 Maps v4.0.1
3069422788|1296252022064|0|112|POST /glm/mmap
3069422788|1296252022064|0|117|www.google.com
3069422788|1296252022064|0|117|www.google.com
3069422788|1296252022064|0|117|www.google.com
3069422788|1296252022064|0|118|88
3069422788|1296252022064|0|118|145
3069422788|1296252022064|0|118|221
3069422788|1296252022064|0|118|8
3069422788|1296252022064|0|118|25611
3069422788|1296252022064|0|123|306 Maps v
3069422788|1296252022064|0|123|306 Maps v
3069422788|1296252022064|0|123|306 Maps v
3069422788|1296252022064|0|123|200 OK
3069422788|1296252022064|0|123|200 OK
3069422788|1296252022064|0|121|en-us
3069422788|1296252022064|0|121|en-us
3069422788|1296252022064|0|121|en-us
3069422788|1296252022064|0|120|*/*
3069422788|1296252022064|0|120|*/*
3069422788|1296252022064|0|120|*/*
3069422788|1296252022064|0|122|application/x-www-form-urlencoded
3069422788|1296252022064|0|122|application/x-www-form-urlencoded
3069422788|1296252022064|0|122|application/x-www-form-urlencoded
3069422788|1296252022064|0|122|application/binary
3069422788|1296252022064|0|122|application/binary
3069422788|1296252022064|0|114|HTTP/1.1
3069422788|1296252022064|0|114|HTTP/1.1
3069422788|1296252022064|0|114|HTTP/1.1
3069422788|1296252022064|0|114|HTTP/1.1
3069422788|1296252022064|0|114|HTTP/1.1

To find the corresponding http flow:

$ cat /data/dpi/flow.txt | grep 3069422788
3069422788|1296252022064|10.10.0.188|72.14.204.104|6|52407|80|900|0

Manual Import into MySQL Database

Now let's upload this data to a MySQL database:

First we need to build the database and tables. super_mediator provides a program, super_table_creator, that will do this for you. If super_table_creator was not installed with super_mediator, it was probably because the MySQL client library is not installed, or that it couldn't find it. You may need to recompile super_mediator using:

$ ./configure --with-mysql=[path to mysql_config utility]
$ make
$ make install

To use super_table_creator to create the appropriate, super_table_creator needs the credentials to access the MySQL database:

$ super_table_creator --name root --password password --database super_flows
Flow Index Table Created Successfully
DNS Table Created Successfully
HTTP Table Created Successfully
TLS Table Created Successfully
SLP Table Created Successfully
IMAP Table Created Successfully
SMTP Table Created Successfully
POP3 Table Created Successfully
IRC Table Created Successfully
FTP Table Created Successfully
TFTP Table Created Successfully
SIP Table Created Successfully
RTSP Table Created Successfully
MYSQL Table Created Successfully
P0F Table Created Successfully
DHCP Table Created Successfully
SSH Table Created Successfully
NNTP Table Created Successfully

If you check your database, you should see all the above tables were created in the "super_flows" database. Check out the super_table_creator man page for a description of the schemas it creates for each table.

To manually load each file into the database individually:

$ mysqlimport -u root -p --fields-terminated-by="|" super_flows /data/dpi/http.txt0
super_flows.http: Records: 38731  Deleted: 0  Skipped: 0  Warnings: 567

This shows that all the lines in the file were uploaded into the database. Unfortunately, mysqlimport will not show the warnings. If you're interested in seeing the warnings, you can use the mysql -e command:

$ mysql -u root -p -e "LOAD DATA INFILE '/data/dpi/http.txt0' \
into table http FIELDS TERMINATED by '|'; SHOW WARNINGS;" super_flows

You could also write a script that uploads all the files to the database:

$ for file in /data/dpi/*; \
do mysqlimport -u root --password=password \
--fields-terminated-by="|" super_flows ${file}; done

We could also load the dns dedup records from Exporter 2 to the database:

$ super_table_creator --name root --password password --database super_flows --dns-dedup
Successfully created DNS dedup table.

If you used the LAST_SEEN keyword in the DNS_DEDUP configuration block:

$ super_table_creator --name root --password password --database super_flows --dns-dedup
Successfully created DNS dedup last seen table.

To load the data in the database:

$ for file in /data/dns/*; \
do mysql -u root --password=password -e \
"load data infile '${file}' into table dns_dedup \
FIELDS TERMINATED BY '|'" super_flows; done

Automating MySQL Import

How to setup super_mediator with MySQL:

It is possible to have super_mediator automatically import the files into the database. For each exporter, super_mediator needs the MySQL credentials and database/table name. To do this using the above configuration file:

EXPORTER TEXT
   PATH "/data/dpi"
   ROTATE 1200
   MULTI_FILES
   DPI_ONLY
   LOCK
   MYSQL_USER "root"
   MYSQL_PASSWORD "secretPassword"
   MYSQL_DATABASE "super_flows"
   REMOVE_UPLOADED
EXPORTER END

Since we're using the MULTI_FILES keyword, we don't have to specify a table using the MYSQL_TABLE keyword. The filename is used to determine which table the file should be loaded to. Note that if you're using the DPI_CONFIG block and renaming the files to different names, the filename prefix should match the MySQL table you intend to upload the file to. The REMOVE_UPLOADED file will remove the file onces it's been successfully uploaded. If there is an error, the file will remain in the file PATH.

Similarly, we can have it upload our deduplicated DNS records:

EXPORTER TEXT
   PATH "/data/dns/yaf2dns"
   DELIMITER "|"
   ROTATE 1200
   DNS_DEDUP_ONLY
   LOCK
   MYSQL_USER "root"
   MYSQL_PASSWORD "secretPassword"
   MYSQL_DATABASE "super_flows"
   MYSQL_TABLE "dns_dedup"
   REMOVE_UPLOADED
EXPORTER END

super_mediator will display messages in the log file such as:

[2014-01-22 14:40:23] Successfully imported file /data/dpi/ftp.txt0 to table 'ftp'
[2014-01-22 14:40:23] Removed Imported File '/data/dpi/ftp.txt0'

With the above setup, you would have to join the flow table and the dns, http, etc. when you want to find the flow information for a particular DPI string. For example, you may want to find out who queried for "news.google.com":

mysql> select f.sip from flow f, dns d where d.qr='Q' and \
d.name='news.google.com.' and f.flow_key = d.flow_key and f.stime = d.stime;

Depending on the amount of data, this query may take a while. By adding the NO_INDEX keyword to the second EXPORTER block, super_mediator will not create the flow index file/table and instead write the 5-tuple with the DPI data.

EXPORTER TEXT
   PATH "/data/dpi"
   ROTATE 1200
   MULTI_FILES
   DPI_ONLY
   LOCK
   NO_INDEX
EXPORTER END

$ cat /data/dpi/http.txt0
2011-01-28 22:00:22.064|10.10.0.188|72.14.204.104|6|52407|80|900|0|111|Apple iPhone \
v8A306 Maps v4.0.1
2011-01-28 22:00:22.064|10.10.0.188|72.14.204.104|6|52407|80|900|0|112|POST /glm/mmap
2011-01-28 22:00:22.064|10.10.0.188|72.14.204.104|6|52407|80|900|0|117|www.google.com
2011-01-28 22:00:22.064|10.10.0.188|72.14.204.104|6|52407|80|900|0|118|88
2011-01-28 22:00:22.064|10.10.0.188|72.14.204.104|6|52407|80|900|0|118|145
2011-01-28 22:00:22.064|10.10.0.188|72.14.204.104|6|52407|80|900|0|118|221
2011-01-28 22:00:22.064|10.10.0.188|72.14.204.104|6|52407|80|900|0|123|306 Maps v
2011-01-28 22:00:22.064|10.10.0.188|72.14.204.104|6|52407|80|900|0|123|200 OK
2011-01-28 22:00:22.064|10.10.0.188|72.14.204.104|6|52407|80|900|0|121|en-us
2011-01-28 22:00:22.064|10.10.0.188|72.14.204.104|6|52407|80|900|0|122|application/\
x-www-form-urlencoded
2011-01-28 22:00:22.064|10.10.0.188|72.14.204.104|6|52407|80|900|0|122|application/binary
2011-01-28 22:00:22.064|10.10.0.188|72.14.204.104|6|52407|80|900|0|114|HTTP/1.1

$ cat /data/dpi/dns.txt0
2011-01-28 21:49:09.702|172.16.0.5|8.8.8.8|17|60019|53|500|0|Q|51786|0|0|0|1|0|\
spreadsheets.google.com.|
2011-01-28 21:49:09.702|172.16.0.5|8.8.8.8|17|60019|53|500|0|R|51786|1|0|0|5|69883|\
spreadsheets.google.com.|spreadsheets.l.google.com.
2011-01-28 21:49:09.702|172.16.0.5|8.8.8.8|17|60019|53|500|0|R|51786|1|0|0|1|77|\
spreadsheets.l.google.com.|72.14.204.102
2011-01-28 21:49:09.702|172.16.0.5|8.8.8.8|17|60019|53|500|0|R|51786|1|0|0|1|77|\
spreadsheets.l.google.com.|72.14.204.100
2011-01-28 21:49:09.702|172.16.0.5|8.8.8.8|17|60019|53|500|0|R|51786|1|0|0|1|77|\
spreadsheets.l.google.com.|72.14.204.113
2011-01-28 21:49:09.702|172.16.0.5|8.8.8.8|17|60019|53|500|0|R|51786|1|0|0|1|77|\
spreadsheets.l.google.com.|72.14.204.101

$ cat /data/dpi/tls.txt0
2011-01-28 22:00:25.635|10.10.1.16|64.233.169.193|6|49356|443|900|0|6|I|0|ZA
2011-01-28 22:00:25.635|10.10.1.16|64.233.169.193|6|49356|443|900|0|10|I|0|Thawte \
Consulting (Pty) Ltd.
2011-01-28 22:00:25.635|10.10.1.16|64.233.169.193|6|49356|443|900|0|3|I|0|Thawte SGC CA
2011-01-28 22:00:25.635|10.10.1.16|64.233.169.193|6|49356|443|900|0|6|S|0|US
2011-01-28 22:00:25.635|10.10.1.16|64.233.169.193|6|49356|443|900|0|8|S|0|California
2011-01-28 22:00:25.635|10.10.1.16|64.233.169.193|6|49356|443|900|0|7|S|0|Mountain View
2011-01-28 22:00:25.635|10.10.1.16|64.233.169.193|6|49356|443|900|0|10|S|0|Google Inc
2011-01-28 22:00:25.635|10.10.1.16|64.233.169.193|6|49356|443|900|0|3|S|0|m.google.com
2011-01-28 22:00:25.635|10.10.1.16|64.233.169.193|6|49356|443|900|0|289|I|0|2
2011-01-28 22:00:25.635|10.10.1.16|64.233.169.193|6|49356|443|900|0|244|I|0|\
1af7d3de2a679c85772c93f85dde8bb4
2011-01-28 22:00:25.635|10.10.1.16|64.233.169.193|6|49356|443|900|0|247|I|0|091218000000Z
2011-01-28 22:00:25.635|10.10.1.16|64.233.169.193|6|49356|443|900|0|248|I|0|111218235959Z
2011-01-28 22:00:25.635|10.10.1.16|64.233.169.193|6|49356|443|900|0|250|I|0|141
2011-01-28 22:00:25.635|10.10.1.16|64.233.169.193|6|49356|443|900|0|6|I|1|US
2011-01-28 22:00:25.635|10.10.1.16|64.233.169.193|6|49356|443|900|0|10|I|1|VeriSign, Inc.
2011-01-28 22:00:25.635|10.10.1.16|64.233.169.193|6|49356|443|900|0|11|I|1|Class 3 \
Public Primary Certification Authority
2011-01-28 22:00:25.635|10.10.1.16|64.233.169.193|6|49356|443|900|0|6|S|1|ZA
2011-01-28 22:00:25.635|10.10.1.16|64.233.169.193|6|49356|443|900|0|10|S|1|Thawte \
Consulting (Pty) Ltd.
2011-01-28 22:00:25.635|10.10.1.16|64.233.169.193|6|49356|443|900|0|3|S|1|Thawte SGC CA
2011-01-28 22:00:25.635|10.10.1.16|64.233.169.193|6|49356|443|900|0|289|I|1|2
2011-01-28 22:00:25.635|10.10.1.16|64.233.169.193|6|49356|443|900|0|244|I|1|30000002
2011-01-28 22:00:25.635|10.10.1.16|64.233.169.193|6|49356|443|900|0|247|I|1|040513000000Z
2011-01-28 22:00:25.635|10.10.1.16|64.233.169.193|6|49356|443|900|0|248|I|1|140512235959Z
2011-01-28 22:00:25.635|10.10.1.16|64.233.169.193|6|49356|443|900|0|250|I|1|141

The NO_INDEX keyword adds the following 8 fields to every row in the file:

start time | src IP | dst IP | protocol | src port | dst port | vlan | obid

To create MySQL tables with the above NO_INDEX schema:

$ super_table_creator --name root --password password --database super_flows --no-index

De-duplication per flow

As you may have noticed in the HTTP data above, most of the fields appear with repeated data. As of super_mediator version 1.1.0, you can add the DEDUP_PER_FLOW keyword within the EXPORTER block to consolidate repeated field values into one row. If I modify the third exporter block to the following and re-run super_mediator:

EXPORTER TEXT "dpi"
   PATH "/data/dpi"
   ROTATE 1200
   MULTI_FILES
   DPI_ONLY
   DEDUP_PER_FLOW
   LOCK
EXPORTER END

$ cat /data/dpi/http.txt0
3069422788|1296252022064|0|111|3|Apple iPhone v8A306 Maps v4.0.1
3069422788|1296252022064|0|112|1|POST /glm/mmap
3069422788|1296252022064|0|117|3|www.google.com
3069422788|1296252022064|0|118|1|88
3069422788|1296252022064|0|118|1|145
3069422788|1296252022064|0|118|1|221
3069422788|1296252022064|0|118|1|8
3069422788|1296252022064|0|118|1|25611
3069422788|1296252022064|0|123|3|306 Maps v
3069422788|1296252022064|0|123|2|200 OK
3069422788|1296252022064|0|121|3|en-us
3069422788|1296252022064|0|120|3|*/*
3069422788|1296252022064|0|122|3|application/x-www-form-urlencoded
3069422788|1296252022064|0|122|2|application/binary
3069422788|1296252022064|0|114|5|HTTP/1.1

You can see that it changes the format of the CSV file slightly by adding an additional column, count. Using the DEDUP_PER_FLOW option with TEXT exporters, super_mediator will de-duplicate repeated data values and add the count column between the Information Element ID and the data field. The DEDUP_PER_FLOW option only affects certain protocols (HTTP, RTSP, SMTP, POP3, IRC, SSH, IMAP, SIP, SLP, FTP, POP3, RTSP, MODBUS, ENIP). super_table_creator will create the count column for the tables if ran with the --dedupflow option.

Exporter 4 Explained - Custom Exporters

The fourth exporter was our "custom list" exporter. It defined which fields we wanted to export for each flow.

$ head -4 /data/dpi/custom.txt
2011-01-28 21:45:28.636|2011-01-28 21:45:29.131|172.16.0.163|128.121.146.100|\
58367|80|6|500|S|APSF|AS|APSF|80|111|Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.6;\
 en-US; rv:1.9.2.13) Gecko/20101203 Firefox/3.6.13
2011-01-28 21:45:28.636|2011-01-28 21:45:29.131|172.16.0.163|128.121.146.100|\
58367|80|6|500|S|APSF|AS|APSF|80|112|GET /account/available_features
2011-01-28 21:45:28.636|2011-01-28 21:45:29.131|172.16.0.163|128.121.146.100|\
58367|80|6|500|S|APSF|AS|APSF|80|115|http://twitter.com/
2011-01-28 21:45:28.636|2011-01-28 21:45:29.131|172.16.0.163|128.121.146.100|\
58367|80|6|500|S|APSF|AS|APSF|80|117|twitter.com

MySQL Bonus - Correlating Exporters.

The following configuration will allow you to correlate the DPI data with all of the flow information in order to view packet/byte count, TCP flags, etc.

You should create two exporters, one MULTI_FILES exporter and one custom exporter:

EXPORTER TEXT
  PATH "/data/mediator/dpi"
  ROTATE 600
  MULTI_FILES
  LOCK
  DPI_ONLY
  TIMESTAMP_FILES
  REMOVE_UPLOADED
  MYSQL_USER "mediator"
  MYSQL_PASSWORD "AnExtremelySecretPassword"
  MYSQL_DATABASE "super_flows"
EXPORTER END

EXPORTER TEXT
  PATH "/data/mediator/dpi/fflow"
  FIELDS hash,stimems,etimems,sipint,dipint,sport,dport,protocol,application,vlanint,\
  OBDOMAIN,pkts,rpkts,bytes,rbytes,iflags,riflags,uflags,ruflags
  ROTATE 600
  REMOVE_UPLOADED
  TIMESTAMP_FILES
  MYSQL_USER "mediator"
  MYSQL_PASSWORD "AnExtremelySecretPassword"
  MYSQL_DATABASE "super_flows"
  MYSQL_TABLE "custom_flow"
EXPORTER END

For the second exporter, you will need to create a new table in the database that matches the fields you defined. As a bonus, I have included the MySQL create table command to create the corresponding table for the above exporter:

CREATE TABLE `custom_flow` (
 `flow_key` int unsigned NOT NULL,
 `stime` bingint unsigned DEFAULT NULL,
 `etime` bigint unsigned DEFAULT NULL,
 `sip` int unsigned DEFAULT NULL,
 `dip` int unsigned DEFAULT NULL,
 `sport` mediumint unsigned DEFAULT NULL,
 `dport` mediumint unsigned DEFAULT NULL,
 `protocol` tinyint unsigned DEFAULT NULL,
 `application` mediumint unsigned DEFAULT NULL,
 `vlan` int unsigned DEFAULT NULL,
 `obid` int unsigned  DEFAULT NULL,
 `pkts` bigint unsigned  DEFAULT NULL,
 `rpkts` bigint unsigned DEFAULT NULL,
 `bytes` bigint  unsigned DEFAULT NULL,
 `rbytes` bigint unsigned DEFAULT NULL,
 `iflags` VARCHAR(10) DEFAULT NULL,
 `riflags` VARCHAR(10) DEFAULT NULL,
 `uflags` VARCHAR(10) DEFAULT NULL,
 `ruflags` VARCHAR(10) DEFAULT NULL
);

So if something interesting is identified in the DPI data, it can easily be joined with the custom_flow table to determine packet and byte counts.