When assessing an application for weaknesses in a linux environment, we won’t always have the luxury of freely available source code or documentation. As a result, these situations require more of a black box approach where much of the information about the application will be revealed by attempting to monitor things such as network communications, calls to cryptographic functions, and file I/O.
One method of monitoring applications to extract information is to attach a debugger, such as GDB, to the process and to dump register or stack values as breakpoints are hit for the desired function calls. While this has the advantage of giving fine grained control over things such as code flow and register contents, it is also a cumbersome process compared to hooking the function calls of interest to modify their behavior.
Function call hooking refers to a range of techniques used to intercept calls to pre-existing functions and wrap around them to modify the function’s behavior at runtime. In this article we’ll be focusing on function hooking in linux using the dynamic loader API, which allows us to dynamically load and execute calls from shared libraries on the system at runtime, and allows us to wrap around existing functions by making use of the LD_PRELOAD environment variable.
The LD_PRELOAD environment variable is used to specify a shared library that is to be loaded first by the loader. Loading our shared library first enables us to intercept function calls and using the dynamic loader API we can bind the originally intended function to a function pointer and pass the original arguments through it, effectively wrapping the function call.
Let’s use the ubiquitous “hello world” demonstration as an example. In this example we’ll intercept the puts function and change the output. Here’s our helloworld.c file:
int puts(const char *message)
int (*new_puts)(const char *message);
new_puts = dlsym(RTLD_NEXT, "puts");
if(strcmp(message, "Hello world!n") == 0)
result = new_puts("Goodbye, cruel world!n");
result = new_puts(message);
Let’s take a moment to examine what’s going on here in our libexample.c file:
Line 5 contains our puts function declaration. To intercept the original puts we define a function with the exact same name and function signature as the original libc puts function.
Line 7 declares the function pointer new_puts that will point to the originally intended puts function. As before with the intercepting function declaration this pointer’s function signature must match the function signature of puts.
Line 10 initializes our function pointer using the dlsym() function. The RTLD_NEXT enum tells the dynamic loader API that we want to return the next instance of the function associated with the second argument (in this case puts) in the load order.
We compare the argument passed to our puts hook against “Hello world!n” on line 12 and if it matches, we replace it with “Goodbye, cruel world!n”. If the two strings do not match we simply pass the original message on to puts on line 14.
First we compile helloworld.c as one normally would. Next we compile libexample.c into a shared library by specifying the -shared and -fPIC compile flags and link against libdl using the -ldl flag. The -D_GNU_SOURCE flag is specified to satisfy #ifdef conditions that allow us to use the RTLD_NEXT enum. Optionally this flag can be replaced by adding “#define _GNU_SOURCE” somewhere near the top of our libexample.c file. After compiling our source files, we set the LD_PRELOAD environment variable to point to the location of our newly created shared library.
As expected, when our helloworld binary is executed the puts function is intercepted and “Goodbye, cruel world!” rather than the original “Hello world!” string is displayed. Now that we’re familiar with the process of hooking function calls let’s apply it towards a bit more practical example. Let’s pretend for a moment that we have an application that we are assessing and that this application uses OpenSSL to encrypt communications of sensitive data. Let’s also assume that attempts to man-in-the-middle these communications at the network level have been fruitless. To get at this sensitive data we will intercept calls to SSL_write, the function responsible for encrypting then sending data over a socket. Intercepting SSL_write will allow us to log the string sent to the function and pass the original parameters along, effectively bypassing the encryption protections while allowing the application to run normally. To get started let’s take a look at the SSL_write function definition:
int SSL_write(SSL *ssl, const void *buf, int num);
Here’s the code I’ve written to intercept SSL_write in hook.c:
As we can see our function definition needs to return an integer and take three arguments: a pointer to an SSL context, a pointer to a buffer containing the string to encrypt, and the number of bytes to write. In addition to our intercepting function definition we define a matching function pointer that will point to the originally intended SSL_write function and initialize it with the dlsym function. After pointing our pointer to the original function, we log the process ID of the process calling SSL_write, and the string sent to it. Next we compile our source to a shared library:
The only difference between this compilation and last is the -lssl flag, which we specify in order to link our code against the OpenSSL library. Now let’s go ahead and set LD_PRELOAD to point to our newly created libhook library:
Now that LD_PRELOAD is set we’re ready to start intercepting calls to SSL_write on processes executed from here onward. To test this let’s go ahead and use the curl utility over HTTPS and intercept the HTTPS request.
sigma@ubuntu:~/code$ curl https://www.netspi.com > /dev/null
% Total % Received % Xferd Average Speed Time Time Time Current Dload Upload Total Spent Left Speed
100 19086 0 19086 0 0 37437 0 --:--:-- --:--:-- --:--:-- 60590
After successful completion of the command there should be a log file that we can examine:
As we can see the request has been logged in plaintext, while the application was allowed to function normally. Had this been a scenario where data integrity relied heavily upon SSL encryption and the assumption that man-in-the-middle attacks would be occurring only at the network level, any such integrity would have been compromised. These are really just a few examples of what’s possible using the dynamic loader API and LD_PRELOAD. Since the shared library we create will be loaded into the running process’ memory space we could do things like dump the memory of the process to examine the memory at runtime or tamper with runtime variables. Other uses for this method of function call hooking and loading generally fall under the use case of user-land rootkits and malware, which will be the focus on the next article in this series.
PTaaS is NetSPI’s delivery model for penetration testing. It enables customers to simplify the scoping of new engagements, view their testing results in real time, orchestrate faster remediation, perform always-on continuous testing, and more - all through the Resolve™ vulnerability management and orchestration platform.
We help organizations defend against adversaries by being the best at simulating real-world, sophisticated adversaries with the products, services, and training we provide. We know how attackers think and operate, allowing us to help our customers better defend against the threats they face daily.
At NetSPI, we believe that there is simply no replacement for human-led manual deep dive testing. Our Resolve platform delivers automation to ensure our people spend time looking for the critical vulnerabilities that tools miss. We provide automated and manual testing of all aspects of an organization’s entire attack surface, including external and internal network, application, cloud, and physical security.
Our proven methodology ensures that the client experience and our findings aren’t only as good as the latest tester assigned to your project. That consistency gives our customers assurance that if vulnerabilities exist, we will find them.
Necessary cookies help make a website usable by enabling basic functions like page navigation and access to secure areas of the website. The website cannot function properly without these cookies.
YouTube session cookie.
Marketing cookies are used to track visitors across websites. The intention is to display ads that are relevant and engaging for the individual user and thereby more valuable for publishers and third party advertisers.
Analytics cookies help website owners to understand how visitors interact with websites by collecting and reporting information anonymously.
Preference cookies enable a website to remember information that changes the way the website behaves or looks, like your preferred language or the region that you are in.
Unclassified cookies are cookies that we are in the process of classifying, together with the providers of individual cookies.
Cookies are small text files that can be used by websites to make a user's experience more efficient. The law states that we can store cookies on your device if they are strictly necessary for the operation of this site. For all other types of cookies we need your permission. This site uses different types of cookies. Some cookies are placed by third party services that appear on our pages.