In Part 7 of Team82's OPC UA Deep Dive Series, we will share examples of different denial-of-service attacks against OPC UA implementations. Successful exploits of implementation vulnerabilities can result in production stoppages, financial losses, and safety risks.
Throughout our extensive OPC UA Deep Dive Series, we researched the OPC UA protocol and its different functions and importance within operational technology (OT) environments. The centerpiece tool of our work is an advanced OPC UA Exploit Framework we built and used to execute many unique attacks against OPC UA implementations. We coordinated disclosures with affected vendors for all the vulnerabilities and attack concepts we discovered and developed. In this blog we will share some details of those vulnerabilities and their impact.
Denial of service (DoS) attacks on industrial networks utilizing OPC UA are particularly perilous due to their potential to disrupt critical operations. OPC UA-based systems play a vital role in industrial automation, and a DoS attack can overload network resources or flood communication channels, rendering the system unavailable.
This downtime can result in production stoppages, financial losses, and safety risks, because the affected industrial processes may not respond correctly in emergency situations. Furthermore, a prolonged DoS attack can exhaust resources needed for monitoring and recovery, making it challenging to mitigate the impact and might cause physical damage.
Short Description: Sending many OPC UA message chunks without providing the Final chunk leads to uncontrolled memory allocations and eventually to a crash.
The OPC UA specification determines that OPC UA messages are constructed of chunks (“message chunks”). Each message chunk has an OPC UA header with the following attributes:
Message Type (MSG, OPN, CLO)
Is Final Flag (F = Final, C = Chunk)
Message Size (32 bit)
In most cases, OPC UA protocol stacks will limit the allowed message size of each individual OPC UA message. If a large message needs to be sent, the message will be divided into multiple chunks each marked as C (chunk). When the final chunk arrives (marked as F Final) the full message will get decoded and handled.
However, we found that the application did not limit the amount of received chunks per a single session or in total for all concurrent sessions. Therefore, an attacker can abuse this to send an unlimited number of huge chunks (2 GB each) without ever sending the Final closing chunk. This will lead to an unlimited memory consumption because an increasing amount of heap allocation will be made until the entire RAM memory is consumed. Finally the OPC UA server will crash due to uncontrolled resource exhaustion.
In order to reproduce the issue:
Open a valid OPC UA session
Send many MSG chunks (IsFinal = Chunk)
Never send the Final chunk
Keep the session open
We would like to thank both ZDI and Prosys for their swift response and fixing all of our findings.
OPC Unified Architecture .NET Standard Protocol Stack implementation, and the ConditionRefresh method
Short Description: Invoking many method calls (e.g. ConditionRefresh) leads to uncontrolled memory allocations and eventually to a crash.
The OPC UA specifications recommend servers to implement the ConditionRefresh Method. ConditionRefresh allows a Client to request a Refresh of all Condition instances that currently are important for the client to know. This includes previous states of a Condition instance for which the Server maintains Branches. A Client would typically invoke this Method when it initially connects to a Server and following any situation, such as communication disruptions, in which it would require resynchronization with the Server. This Method is only available on the ConditionType or its subtypes. To invoke this Method, the call shall pass the well-known Method Id of the Method on the ConditionType and the ObjectId shall be the well-known ObjectId of the ConditionType Object.
We found that the OPC UA stack handles this method asynchronously and does not limit the number of times that the client can call this method. The result is that a malicious client can call this procedure an unlimited number of times, forcing the server to reach an extreme memory and resource exhaustion which results in a crash due to insufficient resources.
The ConditionRefresh handling starts with a call back
Inside, there is an async call to DoConditionReresh. This call will be executed in the background by starting a new thread.
By calling this method multiple times, we are forcing the server to start many new threads that will do some work.
This work includes iterating through the current OPC UA objects and copying them while allocating huge amounts of memory.
Eventually, the server will consume all existing heap memory.
And finally, it will crash due to an “out-of-memory” error or hitting a heap guard.
We would like to thank both ZDI and OPC Foundation and the maintainer of UA-.NETStandard for their swift response and fixing all of our findings.
In this blog we’ve discussed three different attack concepts against OPC UA servers that cause denial of service conditions. All of these vulnerabilities were responsibly disclosed and fixed long ago.
In upcoming parts of our OPC UA Deep Dive Series, we will present other findings such as remote code execution vulnerabilities against OPC UA servers and even clients.
CWE-285: IMPROPER AUTHORIZATION
In Emerson Rosemount GC370XA, GC700XA, and GC1500XA products, an unauthenticated user with network access could obtain access to sensitive information or cause a denial-of-service condition.
CVSS v3: 6.9
In Emerson Rosemount GC370XA, GC700XA, and GC1500XA products, an unauthenticated user with network access could bypass authentication and acquire admin capabilities.
CVSS v3: 8.3
CWE-77: IMPROPER NEUTRALIZATION OF SPECIAL ELEMENTS USED IN A COMMAND ('COMMAND INJECTION')
In Emerson Rosemount GC370XA, GC700XA, and GC1500XA products, an unauthenticated user with network access could execute arbitrary commands in root context from a remote computer.
CVSS v3: 9.8
CWE-20: Improper Input Validation
MachineSense FeverWarn Raspberry Pi-based devices lack input sanitization, which could allow an attacker on an adjacent network to send a message running commands or could overflow the stack.
CVSS v3: 8.1
CWE-284: Improper Access Control
MachineSense FeverWarn devices are configured as Wi-Fi hosts in a way that attackers within range could connect to the device's web services and compromise the device.
CVSS v3: 8.8