This trace set is available for public download. See the above download links.

This is a 600 day long packet header trace captured at the border of the University of Waikato network. The traces were captured using a single DAG3 card and the WDCap trace capture software. The version of WDCap used was version 1.0.0 and the Libtrace version was 1.0.6.

The capture point was located between the University's network infrastructure and the commodity Internet. This allowed access to all the traffic that was coming into and exiting the University. However, no internal traffic would have been observed and captured by our capture point. At one point during the capture of this traceset, the hub that the capture point was connected to was upgraded from 10 Mbit to 100 Mbit. This resulted in a small period where no packets were captured and was also possibly the cause of one of the connection resets that occurred during the capture of this traceset.

The capture machine performed all the anonymization and truncation before exporting the packets via the network to a second machine. That machine was tasked with writing the packets to disk and rotating the trace files as required.

Each trace file is named using the following format: yyyymmdd-HHMMSS-[code].gz. The time and date refers to the time in UTC when the first packet in the file was captured. The code refers to the event which caused the previous file to be closed and this new file to be created. Note that the codes used for this traceset are very different to those used in subsequent Waikato tracesets.

Codes of interest for this traceset are as follows:

• 0 - Rotation boundary was reached
• 64 - Connection to collector was restarted
• 128 - Encryption key was changed

Regular file rotation (code 0) occured daily at Midnight (UTC).

There were a few occasions where the connection between the collector and the disk exporter was lost. Due to the reliable nature of our collection process, no packets should have been lost as a result of such failures. However, it may be possible that a few duplicate packets may appear in the traces both immediately before and after the connection was lost.

Additionally, the capture interface was accidentally disconnected on at least two occasions which resulted in gaps in the capture. The capture itself continued without any problem - just no packets were captured during the time period where the interface was disconnected.

The packets have been truncated at the end of the transport header. No user payload is included in any of the packets. However, ICMP packets are truncated after 8 bytes of ICMP header - the IP header for the original datagram is not included.

The IP addresses contained within the packets have been anonymised using Crypto-Pan AES encryption, which is a prefix-preserving anonymisation method. This means that unanonymised IP addresses that were on the same subnet will also be identifiable as on the same subnet when the addresses are anonymized. We change the encryption key once a week on Sunday midnight (local time). This key change causes a file rotation, with a rotation code of 128.

The IP checksums have been validated and anonymized. If the checksum was correct, it has been replaced with 0. If the checksum was incorrect, it has been replaced with 1. TCP checksums have not been anonymized, however.

The recommended method for processing these traces is to use Libtrace, which we have developed. There are a number of tools included with libtrace such as a packet dumping utility, a trace format converter (for example, to convert to pcap), a trace splitting/filtering tool and a few statistic generators. We suggest you examine the Libtrace Wiki for more details on the Libtrace tools and the library itself.