☑️Thick Client Applications
Thick client applications are the applications that are installed locally on our computers. Unlike thin client applications that run on a remote server and can be accessed through the web browser, these applications do not require internet access to run, and they perform better in processing power, memory, and storage capacity.
Thick client applications are usually applications used in enterprise environments created to serve specific purposes. Such applications include project management systems, customer relationship management systems, inventory management tools, and other productivity software. These applications are usually developed using Java, C++, .NET, or Microsoft Silverlight.
A critical security measure that, for example, Java has is a technology called sandbox. The sandbox is a virtual environment that allows untrusted code, such as code downloaded from the internet, to run safely on a user's system without posing a security risk. In addition, it isolates untrusted code, preventing it from accessing or modifying system resources and other applications without proper authorization. Besides that, there are also Java API restrictions and Code Signing that helps to create a more secure environment.
In a .NET environment, a thick client, also known as a rich client or fat client, refers to an application that performs a significant amount of processing on the client side rather than relying solely on the server for all processing tasks. As a result, thick clients can provide a better performance, more features, and improved user experiences compared to their thin client counterparts, which rely heavily on the server for processing and data storage.
Some examples of thick client applications are web browsers, media players, chatting software, and video games. Some thick client applications are usually available to purchase or download for free through their official website or third-party application stores, while other custom applications that have been created for a specific company, can be delivered directly from the IT department that has developed the software. Deploying and maintaining thick client applications can be more difficult than thin client applications since patches and updates must be done locally to the user's computer. Some characteristics of thick client applications are:
Independent software.
Working without internet access.
Storing data locally.
Less secure.
Consuming more resources.
More expensive.
Thick client applications can be categorized into two-tier and three-tier architecture. In two-tier architecture, the application is installed locally on the computer and communicates directly with the database. In the three-tier architecture, applications are also installed locally on the computer, but in order to interact with the databases, they first communicate with an application server, usually using the HTTP/HTTPS protocol. In this case, the application server and the database might be located on the same network or over the internet. This is something that makes three-tier architecture more secure since attackers won't be able to communicate directly with the database.
Since a large portion of thick client applications are downloaded from the internet, there is no sufficient way to ensure that users will download the official application, and that raises security concerns. Web-specific vulnerabilities like XSS, CSRF, and Clickjacking, do not apply to thick client applications. However, thick client applications are considered less secure than web applications with many attacks being applicable, including:
Improper Error Handling.
Hardcoded sensitive data.
DLL Hijacking.
Buffer Overflow.
SQL Injection.
Insecure Storage.
Session Management.
Penetration Testing Steps
Thick client applications are considered more complex than others, and the attacking surface can be large. Thick client application penetration testing can be done both using automated tools and manually. The following steps are usually followed when testing thick client applications.
Information Gathering
In this step, penetration testers have to identify the application architecture, the programming languages and frameworks that have been used, and understand how the application and the infrastructure work. They should also need to identify technologies that are used on the client and server sides and find entry points and user inputs. Testers should also look for identifying common vulnerabilities. The following tools will help us gather information.
Client Side attacks
Although thick clients perform significant processing and data storage on the client side, they still communicate with servers for various tasks, such as data synchronization or accessing shared resources. This interaction with servers and other external systems can expose thick clients to vulnerabilities similar to those found in web applications, including command injection, weak access control, and SQL injection.
Sensitive information like usernames and passwords, tokens, or strings for communication with other services, might be stored in the application's local files. Hardcoded credentials and other sensitive information can also be found in the application's source code, thus Static Analysis is a necessary step while testing the application. Using the proper tools, we can reverse-engineer and examine .NET and Java applications including EXE, DLL, JAR, CLASS, WAR, and other file formats. Dynamic analysis should also be performed in this step, as thick client applications store sensitive information in the memory as well.
Network Side Attacks
If the application is communicating with a local or remote server, network traffic analysis will help us capture sensitive information that might be transferred through HTTP/HTTPS or TCP/UDP connection, and give us a better understanding of how that application is working. Penetration testers that are performing traffic analysis on thick client applications should be familiar with tools like:
Server Side Attacks
Server-side attacks in thick client applications are similar to web application attacks, and penetration testers should pay attention to the most common ones including most of the OWASP Top Ten.
Retrieving hardcoded Credentials
(Post compromise) The following scenario walks us through enumerating and exploiting a thick client application, in order to move laterally inside a corporative network during penetration testing. The scenario starts after we have gained access to an exposed SMB service.
Exploring the NETLOGON share of the SMB service reveals RestartOracle-Service.exe among other files. Downloading the executable locally and running it through the command line, it seems like it does not run or it runs something hidden.
Downloading the tool ProcMon64 from SysInternals and monitoring the process reveals that the executable indeed creates a temp file in C:\Users\Matt\AppData\Local\Temp.
In order to capture the files, it is required to change the permissions of the Temp folder to disallow file deletions. To do this, we right-click the folder C:\Users\Matt\AppData\Local\Temp and under Properties -> Security -> Advanced -> cybervaca -> Disable inheritance -> Convert inherited permissions into explicit permissions on this object -> Edit -> Show advanced permissions, we deselect the Delete subfolders and files, and Delete checkboxes.
Finally, we click OK -> Apply -> OK -> OK on the open windows. Once the folder permissions have been applied we simply run again the Restart-OracleService.exe and check the temp folder. The file 6F39.bat is created under the C:\Users\cybervaca\AppData\Local\Temp\2. The names of the generated files are random every time the service is running.
Listing the content of the 6F39 batch file reveals the following.
Inspecting the content of the file reveals that two files are being dropped by the batch file and being deleted before anyone can get access to the leftovers. We can try to retrieve the content of the 2 files, by modifying the batch script and removing the deletion.
After executing the batch script by double-clicking on it, we wait a few minutes to spot the oracle.txt file which contains another file full of base64 lines, and the script monta.ps1 which contains the following content, under the directory c:\programdata\. Listing the content of the file monta.ps1 reveals the following code.
This script simply reads the contents of the oracle.txt file and decodes it to the restart-service.exe executable. Running this script gives us a final executable that we can further analyze.
Now when executing restart-service.exe we are presented with the banner Restart Oracle created by HelpDesk back in 2010.
Inspecting the execution of the executable through ProcMon64 shows that it is querying multiple things in the registry and does not show anything solid to go by.
Let's start x64dbg, navigate to Options -> Preferences, and uncheck everything except Exit Breakpoint:
By unchecking the other options, the debugging will start directly from the application's exit point, and we will avoid going through any dll files that are loaded before the app starts. Then, we can select file -> open and select the restart-service.exe to import it and start the debugging. Once imported, we right click inside the CPU view and Follow in Memory Map:
Checking the memory maps at this stage of the execution, of particular interest is the map with a size of 0000000000003000 with a type of MAP and protection set to -RW--.
Memory-mapped files allow applications to access large files without having to read or write the entire file into memory at once. Instead, the file is mapped to a region of memory that the application can read and write as if it were a regular buffer in memory. This could be a place to potentially look for hardcoded credentials.
If we double-click on it, we will see the magic bytes MZ in the ASCII column that indicates that the file is a DOS MZ executable.
Let's return to the Memory Map pane, then export the newly discovered mapped item from memory to a dump file by right-clicking on the address and selecting Dump Memory to File. Running strings on the exported file reveals some interesting information.
Reading the output reveals that the dump contains a .NET executable. We can use De4Dot to reverse .NET executables back to the source code by dragging the restart-service_00000000001E0000.bin onto the de4dot executable.
Now, we can read the source code of the exported application by dragging and dropping it onto the DnSpy executable.
With the source code disclosed, we can understand that this binary is a custom-made runas.exe with the sole purpose of restarting the Oracle service using hardcoded credentials.
Web Vulnerabilities
Thick client applications with a three-tier architecture have a security advantage over those with a two-tier architecture since it prevents the end-user from communicating directly with the database server. However, three-tier applications can be susceptible to web-specific attacks like SQL Injection and Path Traversal.
Suppose we find a .jar file fatty-client.jar and we found the creds to log in.
Foothold
Clicking on Profile -> Whoami reveals that the user qtc is assigned with the user role.
Clicking on the ServerStatus, we notice that we can't click on any options.
This implies that there might be another user with higher privileges that is allowed to use this feature. Clicking on the FileBrowser -> Notes.txt reveals the file security.txt. Clicking the Open option at the bottom of the window shows the following content.
This note informs us that a few critical issues in the application still need to be fixed. Navigating to the FileBrowser -> Mail option reveals the dave.txt file containing interesting information. We can read its content by clicking the Open option at the bottom of the window.
The message from dave says that all admin users are removed from the database. It also refers to a timeout implemented in the login procedure to mitigate time-based SQL injection attacks.
Path Traversal
Since we can read files, let's attempt a path traversal attack by giving the following payload in the field and clicking the Open button.
The server filters out the / character from the input. Let's decompile the application using JD-GUI, by dragging and dropping the fatty-client-new.jar onto the jd-gui.
Save the source code by pressing the Save All Sources option in jdgui. Decompress the fatty-client-new.jar.src.zip by right-clicking and selecting Extract files. The file fatty-client-new.jar.src/htb/fatty/client/methods/Invoker.java handles the application features. Reading its content reveals the following code.
The showFiles function takes in one argument for the folder name and then sends the data to the server using the sendAndRecv() call. The file fatty-client-new.jar.src/htb/fatty/client/gui/ClientGuiTest.java sets the folder option. Let's read its content.
We can replace the configs folder name with .. as follows.
Next, compile the ClientGuiTest.Java file.
This generates several class files. Let's create a new folder and extract the contents of fatty-client.jar into it.
Navigate to the raw directory and decompress fatty-client-new-2.jar by right-clicking and selecting Extract Here. Overwrite any existing htb/fatty/client/gui/*.class files with updated class files.
Finally, we build the new JAR file.
Let's log in to the application and navigate to FileBrowser -> Config option.
This is successful. We can now see the content of the directory configs/../. The files fatty-server.jar and start.sh look interesting. Listing the content of the start.sh file reveals that fatty-server.jar is running inside an Alpine Docker container.
We can modify the open function in fatty-client-new.jar.src/htb/fatty/client/methods/Invoker.java to download the file fatty-server.jar as follows.
Rebuild the JAR file by following the same steps and log in again to the application. Then, navigate to FileBrowser -> Config, add the fatty-server.jar name in the input field, and click the Open button.
The fatty-server.jar file is successfully downloaded onto our desktop, and we can start the examination.
SQL Injection
Decompiling the fatty-server.jar using JD-GUI reveals the file htb/fatty/server/database/FattyDbSession.class that contains a checkLogin() function that handles the login functionality. This function retrieves user details based on the provided username. It then compares the retrieved password with the provided password.
Let's check how the client application sends credentials to the server. The login button creates the new object ClientGuiTest.this.user for the User class. It then calls the setUsername() and setPassword() functions with the respective username and password values. The values that are returned from these functions are then sent to the server.
Let's check the setUsername() and setPassword() functions from htb/fatty/shared/resources/user.java.
The username is accepted without modification, but the password is changed to the format below.
We also notice that the username isn't sanitized and is directly used in the SQL query, making it vulnerable to SQL injection.
The checkLogin function in htb/fatty/server/database/FattyDbSession.class writes the SQL exception to a log file.
Login into the application using the username qtc' to validate the SQL injection vulnerability reveals a syntax error. To see the error, we need to edit the code in the fatty-client-new.jar.src/htb/fatty/client/gui/ClientGuiTest.java file as follows.
Listing the content of the error-log.txt file reveals the following message.
This confirms that the username field is vulnerable to SQL Injection. However, login attempts using payloads such as ' or '1'='1 in both fields fail. Assuming that the username in the login form is ' or '1'='1, the server will process the username as below.
The above query succeeds and returns the first record in the database. The server then creates a new user object with the obtained results.
It then compares the newly created user password with the user-supplied password.
The user-supplied password hash user.getPassword() is calculated as follows.
Although the hash sent to the server by the client doesn't match the one in the database, and the password comparison fails, the SQL injection is still possible using UNION queries. Let's consider the following example.
It is possible to create fake entries using the SELECT operator. Let's input an invalid username to create a new user entry.
Similarly, the injection in the username field can be leveraged to create a fake user entry.
This way, the password, and the assigned role can be controlled. The following snippet of code sends the plaintext password entered in the form. Let's modify the code in htb/fatty/shared/resources/User.java to submit the password as it is from the client application.
We can now rebuild the JAR file and attempt to log in using the payload abc' UNION SELECT 1,'abc','[email protected]','abc','admin in the username field and the random text abc in the password field.
The server will eventually process the following query.
Last updated