Login
In essence, the main idea came to use WAF + YARA (YARA right-to-left = ARAY) to detect malicious files at the WAF level before WAF can forward them to the backend e.g. files uploaded through web functions see: https://owasp.org/www-community/vulnerabilities/Unrestricted_File_Upload
When a web page allows uploading files, most of the WAFs are not inspecting files before sending them to the backend. Implementing WAF + YARA could provide malware detection before WAF forwards the files to the backend.
Yes, one solution is to use ModSecurity + Clamav, most of the pages call ClamAV as a process and not as a daemon, in this case, analysing a file could take more than 50 seconds per file. See this resource: https://kifarunix.com/intercept-malicious-file-upload-with-modsecurity-and-clamav/
:-( A few clues here Black Hat Asia 2019 please continue reading and see below our quick LAB deployment.
Basically, It is a quick deployment (1) with pre-compiled and ready-to-use YARA rules via ModSecurity (WAF) using a custom rule; (2) this custom rule will perform an inspection and detection of the files that might contain malicious code, (3) typically web functions (upload files) if the file is suspicious will reject them receiving a 403 code Forbidden by ModSecurity.
FreshRSS YaraCompile.py compiles all the yara rules. (Python3 code)test.conf is a virtual host that contains the mod security rules. (ModSecurity Code)modsec_yara.py in order to inspect the file that is trying to upload. (Python3 code)/YaraRules/Compiled
/YaraRules/rules
/YaraRules/YaraScripts
/etc/apache2/sites-enabled
/temporal Blueteamers: Rule enforcement, best alerting, malware detection on files uploaded through web functions.Redteamers/pentesters: GreyBox scope , upload and bypass with a malicious file, rule enforcement.Security Officers: Keep alerting, threat hunting.SOC: Best monitoring about malicious files.CERT: Malware Analysis, Determine new IOC.The Proof of Concept is based on Debian 11.3.0 (stable) x64 OS system, OWASP CRC v3.3.2 and Yara 4.0.5, you will find the automatic installation script here wafaray_install.sh and an optional manual installation guide can be found here: manual_instructions.txt also a PHP page has been created as a "mock" to observe the interaction and detection of malicious files using WAF + YARA.
alex@waf-labs:~$ su root
root@waf-labs:/home/alex#
# Remember to change YOUR_USER by your username (e.g waf)
root@waf-labs:/home/alex# sed -i 's/^\(# User privi.*\)/\1\nalex ALL=(ALL) NOPASSWD:ALL/g' /etc/sudoers
root@waf-labs:/home/alex# exit
alex@waf-labs:~$ sudo sed -i 's/^\(deb cdrom.*\)/#\1/g' /etc/apt/sources.list
alex@waf-labs:~$ sudo sed -i 's/^# \(deb\-src http.*\)/ \1/g' /etc/apt/sources.list
alex@waf-labs:~$ sudo sed -i 's/^# \(deb http.*\)/ \1/g' /etc/apt/sources.list
alex@waf-labs:~$ echo -ne "\n\ndeb http://deb.debian.org/debian/ bullseye main\ndeb-src http://deb.debian.org/debian/ bullseye main\n" | sudo tee -a /etc/apt/sources.list
alex@waf-labs:~$ sudo apt-get update
alex@waf-labs:~$ sudo apt-get install sudo -y
alex@waf-labs:~$ sudo apt-get install git vim dos2unix net-tools -y
alex@waf-labs:~$ git clone https://github.com/alt3kx/wafarayalex@waf-labs:~$ cd wafaray
alex@waf-labs:~$ dos2unix wafaray_install.sh
alex@waf-labs:~$ chmod +x wafaray_install.sh
alex@waf-labs:~$ sudo ./wafaray_install.sh >> log_install.log
# Test your LAB environment
alex@waf-labs:~$ firefox localhost:8080/upload.phpOnce the Yara Rules were downloaded and compiled.
It is similar to when you deploy ModSecurity, you need to customize what kind of rule you need to apply. The following log is an example of when the Web Application Firewall + Yara detected a malicious file, in this case, eicar was detected.
Message: Access denied with code 403 (phase 2). File "/temporal/20220812-184146-YvbXKilOKdNkDfySME10ywAAAAA-file-Wx1hQA" rejected by
the approver script "/YaraRules/YaraScripts/modsec_yara.py": 0 SUSPECTED [YaraSignature: eicar]
[file "/etc/apache2/sites-enabled/test.conf"] [line "56"] [id "500002"]
[msg "Suspected File Upload:eicar.com.txt -> /temporal/20220812-184146-YvbXKilOKdNkDfySME10ywAAAAA-file-Wx1hQA - URI: /upload.php"]$ sudo service apache2 stop
$ sudo service apache2 start
$ cd /var/log
$ sudo tail -f apache2/test_access.log apache2/test_audit.log apache2/test_error.log
A malicious file is uploaded, and the ModSecurity rules plus Yara denied uploading file to the backend if the file matched with at least one Yara Rule. (Example of Malware: https://secure.eicar.org/eicar.com.txt) NOT EXECUTE THE FILE.
For this demo, we disable the rule 933110 - PHP Inject Attack to validate Yara Rules. A malicious file is uploaded, and the ModSecurity rules plus Yara denied uploading file to the backend if the file matched with at least one Yara Rule. (Example of WebShell PHP: https://github.com/drag0s/php-webshell) NOT EXECUTE THE FILE.
A malicious file is uploaded, and the ModSecurity rules plus Yara denied uploading file to the backend if the file matched with at least one Yara Rule. (Example of Malware Bazaar (RecordBreaker): https://bazaar.abuse.ch/sample/94ffc1624939c5eaa4ed32d19f82c369333b45afbbd9d053fa82fe8f05d91ac2/) NOT EXECUTE THE FILE.
In case that you want to download more yara rules, you can see the following repositories:
Alex Hernandez aka (@_alt3kx_)
Jesus Huerta aka @mindhack03d
Israel Zeron Medina aka @spk085
sandfly-entropyscan is a utility to quickly scan files or running processes and report on their entropy (measure of randomness) and if they are a Linux/Unix ELF type executable. Some malware for Linux is packed or encrypted and shows very high entropy. This tool can quickly find high entropy executable files and processes which often are malicious.
Entropy is a measure of randomness. For binary data 0.0 is not-random and 8.0 is perfectly random. Good crypto looks like random white noise and will be near 8.0. Good compression removes redundant data making it appear more random than if it was uncompressed and usually will be 7.7 or above.
A lot of malware executables are packed to avoid detection and make reverse engineering harder. Most standard Linux binaries are not packed because they aren't trying to hide what they are. Searching for high entropy files is a good way to find programs that could be malicious just by having these two attributes of high entropy and executable.
Usage of sandfly-entropyscan:
-csv output results in CSV format (filename, path, entropy, elf_file [true|false], MD5, SHA1, SHA256, SHA512)
-delim change the default delimiter for CSV files of "," to one of your choosing ("|", etc.)
-dir string directory name to analyze
-file string full path to a single file to analyze
-proc check running processes (defaults to ELF only check)
-elf only check ELF executables
-entropy float show any file/process with entropy greater than or equal to this value (0.0 min - 8.0 max, defaults 0 to show all files)
-version show version and exit
Search for any file that is executable under /tmp:
sandfly-entropyscan -dir /tmp -elf
Search for high entropy (7.7 and higher) executables (often packed or encrypted) under /var/www:
sandfly-entropyscan -dir /var/www -elf -entropy 7.7
Generates entropy and cryptographic hashes of all running processes in CSV format:
sandfly-entropyscan -proc -csv
Search for any process with an entropy higher than 7.7 indicating it is likely packed or encrypted:
sandfly-entropyscan -proc -entropy 7.7
Generate entropy and cryptographic hash values of all files under /bin and output to CSV format (for instance to save and compare hashes):
sandfly-entropyscan -dir /bin -csv
Scan a directory for all files (ELF or not) with entropy greater than 7.7: (potentially large list of files that are compressed, png, jpg, object files, etc.)
sandfly-entropyscan -dir /path/to/dir -entropy 7.7
Quickly check a file and generate entropy, cryptographic hashes and show if it is executable:
sandfly-entropyscan -file /dev/shm/suspicious_file
Do spot checks on systems you think have a malware issue. Or you can automate the scan so you will get an output if we find something show up that is high entropy in a place you didn't expect. Or simply flag any executable ELF type file that is somewhere strange (e.g. hanging out in /tmp or under a user's HTML directory). For instance:
Did a high entropy binary show up under the system /var/www directory? Could be someone put a malware dropper on your website:
sandfly-entropyscan -dir /var/www -elf -entropy 7.7
Setup a cron task to scan your /tmp, /var/tmp, and /dev/shm directories for any kind of executable file whether it's high entropy or not. Executable files under tmp directories can frequently be a malware dropper.
sandfly-entropyscan -dir /tmp -elf
sandfly-entropyscan -dir /var/tmp -elf
sandfly-entropyscan -dir /dev/shm -elf
Setup another cron or automated security sweep to spot check your systems for highly compressed or encrypted binaries that are running:
sandfly-entropyscan -proc -entropy 7.7
git clone https://github.com/sandflysecurity/sandfly-entropyscan.git
go build
./sandfly-entropyscan
There are a some basic build scripts that build for various platforms. You can use these to build or modify to suit. For Incident Responders, it might be useful to keep pre-compiled binaries ready to go on your investigation box.
build.sh - Build for current OS you're running on when you execute it.
We use a simple method for seeing if a file may be an executable ELF type. We can spot ELF format files for multiple platforms. Even if malware has Intel/AMD, MIPS and Arm dropper binaries we will still be able to spot all of them.
It's possible to flag a legitimate binary that has a high entropy because of how it was compiled, or because it was packed for legitimate reasons. Other files like .zip, .gz, .png, .jpg and such also have very high entropy because they are compressed formats. Compression removes redundancy in a file which makes it appear to be more random and has higher entropy.
On Linux, you may find some kinds of libraries (.so files) get flagged if you scan library directories.
However, it is our experience that executable binaries that also have high entropy are often malicious. This is especially true if you find them in areas where executables normally shouldn't be (such as again tmp or html directories).
The entropy calculation requires reading in all the bytes of the file and tallying them up to get a final number. It can use a lot of CPU and disk I/O, especially on very large file systems or very large files. The program has an internal limit where it won't calculate entropy on any file over 2GB, nor will it try to calculate entropy on any file that is not a regular file type (e.g. won't try to calculate entropy on devices like /dev/zero).
Then we calculate MD5, SHA1, SHA256 and SHA512 hashes. Each of these requires going over the file as well. It's reasonable speed on modern systems, but if you are crawling a very large file system it can take some time to complete.
If you tell the program to only look at ELF files, then the entropy/hash calculations won't happen unless it is an ELF type and this will save a lot of time (e.g. it will ignore massive database files that aren't executable).
If you want to automate this program, it's best to not have it crawl the entire root file system unless you want that specifically. A targeted approach will be faster and more useful for spot checks. Also, use the ELF flag as that will drastically reduce search times by only processing executable file types.
For incident responders, running sandfly-entropyscan against the entire top-level "/" directory may be a good idea just to quickly get a list of likely packed candidates to investigate. This will spike CPU and disk I/O. However, you probably don't care at that point since the box has been mining cryptocurrency for 598 hours anyway by the time the admins noticed.
Again, use the ELF flag to get to the likely problem candidate executables and ignore the noise.
There is a script called scripts/testfiles.sh that will make two files. One will be full of random data and one will not be random at all. When you run the script it will make the files and run sandfly-entropyscan in executable detection mode. You should see two files. One with very high entropy (at or near 8.0) and one full of non-random data that should be at 0.00 for low entropy. Example:
./testfiles.sh
Creating high entropy random executable-like file in current directory.
Creating low entropy executable-like file in current directory.
high.entropy.test, entropy: 8.00, elf: true
low.entropy.test, entropy: 0.00, elf: true
You can also load up the upx utility and compress an executable and see what values it returns.
Sandfly Security produces an agentless endpoint detection and incident response platform (EDR) for Linux. Automated entropy checks are just one of thousands of things we search for to find intruders without loading any software on your Linux endpoints.
Get a free license and learn more below:
https://www.sandflysecurity.com @SandflySecurity
