Skip to content
Tutorials Center
My learning

Topics

Lesson 28 of 37 ยท Statistics and Automation

Statistics: Conversations, Endpoints and I/O Graphs

Use the Wireshark Statistics menu to summarize a capture: Protocol Hierarchy, Conversations, Endpoints, I/O Graphs and Expert Information, plus how to script the same numbers.

Scrolling through packets answers detailed questions. For the big picture, such as "who talks most?" or "when did traffic spike?", use the Statistics menu. It turns a capture into tables and graphs in seconds, and it is usually the right place to start with an unfamiliar file.

This lesson covers the most useful windows and shows how to recreate their numbers in code. Wireshark is a desktop program that cannot run in the browser, so window contents appear as read-only text. The Python blocks are stand-ins using sample records.

The windows you will use most

Statistics at a glance

Protocol Hierarchy Share of traffic per protocol.
Conversations Who talks to whom, how much.
Endpoints Totals per single address.
I/O Graphs Traffic over time.

Open them from the Statistics menu.

1 / 2

Protocol Hierarchy

Statistics, then Protocol Hierarchy, shows a tree of protocols with percentages of packets and bytes. It answers "what is in this capture?" and exposes surprises: 40 percent Telnet, tunnels, or a protocol nobody expected.

TEXT
Protocol            % Packets   Packets   % Bytes     Bytes
Ethernet               100.0      52 310    100.0   41 070 112
  IPv4                  97.6      51 054     98.1   40 280 004
    TCP                 88.3      46 190     94.5   38 810 330
      TLS               61.2      32 011     83.0   34 090 112
      HTTP               9.8       5 120      7.2    2 957 000
    UDP                  9.2       4 810      3.4    1 395 200
      DNS                8.7       4 550      0.4      160 400

Right-click a protocol and choose "Apply as Filter" to jump to its packets.

Conversations and Endpoints

An endpoint is one address (Ethernet, IPv4, IPv6, TCP port, UDP port). A conversation is a pair of endpoints. Both windows show packets and bytes in each direction, and the first and last time seen (Rel Start, Duration). Click a column header to sort. Tick "Limit to display filter" to analyze only filtered packets.

TEXT
Address A      Address B         Packets  Bytes     Duration
10.0.0.20      151.101.0.10       12 440  14 207 884   301.2
10.0.0.42      203.0.113.80        1 890     912 004  3 590.0
10.0.0.20      10.0.0.1            4 550     388 000   312.0

Use it to find the top talkers, one-way flows (many packets A to B, none back), and conversations with unusual durations, as in Malware Traffic and Beaconing.

Python Build a conversations table

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

I/O Graphs

Statistics, then I/O Graph, plots a value per time interval, for example packets per second. Its real power is graphs with display filters: add a second line for tcp.analysis.retransmission and you can see whether retransmissions align with a drop in throughput. Add dns.flags.rcode != 0 to see error bursts. You can also plot a field's value, such as tcp.analysis.ack_rtt, using the Y field option with an aggregation such as AVG or MAX.

Choose the interval wisely. One-second intervals show bursts. Ten-minute intervals show daily patterns in long captures.

Python Bucket packets per second and mark spikes

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Expert Information and Capture File Properties

Analyze, then Expert Information groups findings by severity. Statistics, then Capture File Properties records the capture file's name, size, time span, and the interface used. These details belong in every report.

  • TCP Stream Graphs: Time Sequence (Stevens), Throughput and Round Trip Time views of one connection.
  • Service Response Time: how long DNS, HTTP, SMB and other services take to answer; a table of min, max and average times.
  • HTTP, Requests/Responses, Load Distribution: which hosts, paths and status codes dominate. Compare error codes such as 404 and 500.
  • Packet Lengths: a histogram of packet sizes. A sudden pile of maximum-size packets suggests a bulk transfer; many tiny packets suggests scans or chatty protocols.

Common mistakes

  • Reading statistics from a filtered view without noticing the "Displayed" versus "Captured" columns.
  • Forgetting that bytes in the Conversations window include headers.
  • Trusting one graph. Confirm any spike by looking at the packets inside it.

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. Which window shows the share of a capture taken by each protocol?

    Protocol Hierarchy is a tree with packet and byte percentages per protocol.
  2. 2. What is the difference between an endpoint and a conversation?

    Endpoints list individual addresses; conversations list the pairs that exchanged traffic.
  3. 3. Why add a second I/O Graph line with the filter tcp.analysis.retransmission?

    Overlaying a filtered series shows whether problems line up with throughput drops.
  4. 4. Which tool is best for a quick list of top talkers by bytes?

    Sort the Conversations table by the Bytes column to find the heaviest pairs.

Summary

  • Statistics windows summarize a capture so you know where to look.
  • Protocol Hierarchy, Conversations and Endpoints answer "what" and "who"; I/O Graphs answer "when".
  • Overlay filtered lines in I/O Graphs to correlate problems with traffic.
  • The same numbers are easy to script, which the next lessons do with command-line tools.

Next: tshark: Wireshark on the Command Line