Learning Modbus TCP with ControlThings.IO

-
Modbus TCP is a layer 7 communications protocol and a de facto standard in the field of industrial control systems. Having recently moved into the utilities sector I wanted to learn more about it. In this post I'll be taking a look at the Modbus TCP challenge included in the ControlThings.io ICS pentesting distro and hopefully learning a little about Modbus along the way. So let's get started!

The challenge itself is located at ~/Samples/Protocols/ModbusTCP. We have some pcaps, some reference material and some instructions as follows:

Plant1-ModbusTCP.pcap
This is an export of the ModbusTCP packets from the Plant1.pcap file
Source: https://www.cloudshark.org/captures/76038eaa4a3b

Modbus Challenges:
- Which IP address is the master on?
- How many slaves is the master talking to?
- Is the master writing any data to the slaves?
- Does the traffic spike in the middle related [SIC] to modbus?

So first of all, we need to understand what are these masters and slaves being referred to? I found an excellent write up with some diagrams on the Schneider Electronics Website (Schneider in fact invented Modbus in 1979, as it turns out). As they note:

 "The Modbus protocol exchanges information using a request-reply mechanism between a master (client) and a slave (server). The master-slave principle is a model for a communication protocol in which one device (the master) controls one or more other devices (the slaves). "

Additionally, we learn that: 
  • Only 1 master is connected to the network at a time.
  • Only the master can initiate communication and send requests to the slaves.
  • The master can address each slave individually using its specific address or all slaves simultaneously using address 0.
  • The slaves can only send replies to the master.
  • The slaves cannot initiate communication, either to the master or to other slaves.

Okay, this seems like enough information to get started, let's open the Plant1-ModbusTCP.pcap file in Wireshark and take a look:

So we can immediately see that we have a .10 host issuing queries and a whole bunch of others responding. Let's confirm our suspicion that 148.81.0.10 is the master by applying some display filters. First, let's see all the packets where .10 is the source:


Notice how everything it's sending is a Query. Now let's see the packets where it's the destination:


And as suspected, everything else is responding to it. So there's a definite answer to question 1 of the challenge. Remember that only the master can can initiate communications, and that all the slaves can do is reply to the master. What we can clearly see is 141.81.0.10 is the master and is issuing queries to multiple slaves that are responding to it. Okay next question, how many slaves are there? We can quickly find this using the Statistics->Endpoints options in Wireshark and looking at the IPv4 traffic:

So in addition to the master we have a total of 13 slaves it's talking to. (Notice the huge difference in the volume of traffic sent from the master compared to each individual slave). I found another great description of Modbus TCP here where it's stated that:

"Modbus will support up to 247 slaves from addresses 1 to 247 (JBUS 1 to 255) - address 0 is reserved for broadcast messages. In practice, the number of slave addresses that can be used is determined by the communications link that is chosen. For example, RS485 is limited to a total of 31 slaves."

So that's question 2 answered. The next question asks us whether the master is writing any data to the slaves. Let's have a look!

Well that seems pretty clear, we seem to have several different 'Read' operations being performed but also at least one 'Write' operation ('Write Multiple Coils'). But lets have a quick look at the Modbus specification to confirm this:

Okay, so here we see that Modbus supports many different read and write operations and the specification provides full details of each. Notice how the function number in the spec also corresponds to what is actually displayed in the packet itself: 'Write Multiple Coils' is function 15, 'Read Discrete Inputs' is function 2 etc. Nice!

On closer inspection, the majority of traffic in the capture appears to consist of the following 4 functions:

  • Function 1: Read Coils
  • Function 2: Read Discrete Inputs
  • Function 4: Read Input Registers
  • Function 15: Write Multiple Coils

So next, the final part of the challenge asks us to identify whether the spike in traffic in the middle of the capture is related to Modbus. Let's take a look at that now. First of all where is this spike? We can use the I/O Graph in Wireshark to analyse the traffic in this capture and this is what we get:

Hmmm ... no obvious spike there that I could see anyway. I reached out to ControlThings.io on their Discord and queried this. Shortly after I got the following reply from Justin Searle:

"Correct. There is no Modbus spike. And sorry, that question originates from the ~/Samples/Protocols/Combined/Plant1.pcap capture which is a superset of that Modbus capture and several other captures. If you check out that Plant1.pcap file, you will see there is a traffic spike in another protocol."

Okay, so sounds like we're done with Modbus for now, but let's check out that other pcap that Justin is talking about:

So immediately we can see multiple protocols are in use here, not just Modbus. Let's retry that traffic analysis and see if we can find the spike:

And there we have it! So now let's try and figure out what's causing it. By clicking directly at the top of the spike we can see that the traffic spikes at packet no. 28254, so let's take a look at that:

As we can see, this is indeed not a Modbus spike but instead appears to be caused by 148.81.0.10 sending CIP (Common Industrial Protocol) Multiple Service Packets to a range of other hosts that are then responding with Success messages.

After checking back in with Justin, I was reminded of the "Try Harder" days of OSCP. There is more to this part of the challenge than meets the eye as it turns out and CIP is not what's actually causing the spike! Using Justin's hints I modified my approach as follows:

First up, I changed the I/O graph to display bytes instead of the default, which is packets. This resulted in a much more prominent visualisation of the traffic spike as shown below:

To change the display to bytes you need to double-click on the the word "Packets" in the "Y Axis" column and this then brings up a drop-down control that lets you choose what you want shown on that axis.

Another improvement that lets us see the spike in even more detail is to change the default time-interval down from 1 second to 10ms or 100ms. The screenshot below shows the spike displayed with a 100ms interval:

So if it's not CIP, what is actually causing this spike? 

The first thing I wanted to do was to reduce the size of the problem space. I identified the packet numbers of the approximate start and end of the spike and used editcap to create a new pcap file containing only those packets I was interested in:

With only around 3000 packets, my new file (Spike.pcap) was much more manageable to work with. I took a look at the I/O graph of the new file, showing all the traffic:

Next, I created a bunch of traffic filters, one for each protocol in the new PCAP file and applied them:

As can be seen, the TDS protocol traffic matches the profile of the spike, accounting for significantly more traffic than the other protocols at this timestamp. So let's dig a little deeper and look at the traffic stats for the capture:


Interestingly, in my new capture file there were only 12 more TDS packets than Modbus packets during the period of the spike and the TDS packet count was only 2.5% higher than Modbus. Looking at this, I began to wonder whether TDS was really the source of the spike. However, returning to the I/O graph and viewing the size of the TDS traffic (red) compared to the Modbus traffic (black) it was pretty clear what's going on:

Although the packet count was similar, the overall TDS traffic size is much bigger than the Modbus traffic, with the Modbus peaking at just under 5000 bytes/100ms compared to over 25000 bytes/100ms for TDS.

I really enjoyed this exercise and it was a great way to dip a toe in the water with some of these industrial protocols for the first time. It was also great to learn some new features of Wireshark and to see how easy it is to fall into the trap of making incorrect assumptions about network traffic. 

Many thanks to Justin and the team at ControlThings.IO! Check out their Discord server and download ControlThings.IO here. 


Comments

Post a Comment

Popular posts from this blog

Buffer Overflow Fun with Brainpan 1

-
Image
Brainpan 1 is a vulnerable VM by  @superkojiman  and was posted to Vulnhub back in 2013. I picked it at random while browsing through some of the older entries, looking for my next target. There is minimal information provided in advance, so it's a really black-box challenge. Great! Let's see what we can do with this ... Nmap revealed an interesting result - just two, unusual ports open and something weird going on with one of them: Checking them both out with a web browser, I got the following back: 9999 looked interesting so I fired up netcat and spent a bit of time interacting with the service: Ok so it was clear that this was an important service but without a password I wasn't going to get very far. I switched attention to the Python SimpleHTTP server on port 10000 and, after a quick look at view-source, decided that brute forcing some further files and/or directories was going to be required. I quickly discovered a /bin/ directory and wit
Read more

Real World Web Application Security Testing

-
Introduction Web application security testing in the real world rarely provides opportunities for getting a reverse shell and ultimately “rooting the box” as is our usual goal in CTF scenarios. Nevertheless, it is very rare to find web applications with no significant vulnerabilities, and in my experience it’s nearly always possible to gain access to information and functions that are supposed to be restricted, often by getting access to the administrator account or equivalent.  In this post, I’ll walk you through the tried and tested techniques I’ve employed over many years to achieve consistently good results.  And before we get started, if you’ve ever read a web application pentest report you’ll most likely be very familiar with the many low-rated vulnerabilities that seemingly every site suffers from: Overly permissive CORS policy, secure flag not set on cookie, no anti-CSRF tokens, weak SSL ciphers supported etc. etc., the list goes on. I’m not going to talk about these types of i
Read more

Modelling Security Concepts with Archimate

-
Image
In this post I will be using Archimate to model some fundamental information security concepts. If you're a solution or enterprise architect and you need to get an understanding of information security and how security concerns can be modelled, I hope you'll find this to be a useful starting point. So let's get straight to it and start modelling some basic information security concepts, starting with threat agents. Unlike some areas, there is a straightforward mapping for us in Archimate with the  Business Actor  entity. A business actor is an active entity defined in Archimate as an "organizational entity that is capable of performing behavior" and may exist outside of our own organisation. Whereas in our normal business layer models we are typically concerned with actors such as customers, sales and back-office staff, in security we need to think in terms of the malicious entities that wish to cause harm to our organisation and its systems. Unlike normal actors
Read more