In order to detect malware and other file-based threats, ClamAV relies on signatures to differentiate clean and malicious/unwanted files. ClamAV signatures are primarily text-based and conform to one of the ClamAV-specific signature formats associated with a given method of detection. These formats are explained in the [Signature formats](#signature-formats) section below. In addition, ClamAV 0.99 and above support signatures written in the YARA format. More information on this can be found in the [Using YARA rules in ClamAV](#using-yara-rules-in-clamav) section.
The ClamAV project distributes a collection of signatures in the form of CVD (ClamAV Virus Database) files. The CVD file format provides a digitally-signed container that encapsulates the signatures and ensures that they can't be modified by a malicious third-party. This signature set is actively maintained by [Cisco Talos](https://www.talosintelligence.com/) and can be downloaded using the `freshclam` application that ships with ClamAV. For more details on this, see the [CVD file](#inspecting-signatures-inside-a-CVD-file) section.
ClamAV CVD and CLD database archives may be unpacked to the current directory using `sigtool -u <database name>`. For more details on inspecting CVD and CLD files, see [Inspecting signatures inside a CVD file](#inspecting-signatures-inside-a-cvd-file). Once unpacked, you'll observe a large collection of database files with various extensions described below.
The CVD and CLD database archives may be supplemented with custom database files in the formats described to gain additional detection functionality. This is done simply by adding files of the following formats to the database directory, typically `/usr/local/share/clamav` or `"C:\Program Files\ClamAV\database"`. Alternatively, `clamd` and `clamscan` can be instructed to load the database from an alternative database file or database directory manually using the `clamd``DatabaseDirectory` config option or the `clamscan -d` command line option.
_Note_: Signature databases with an extension ending in `u` are only loaded when Potentially Unwanted Application (PUA) signatures are enabled (default: off).
Body-based signature content is a definition that matches not based on a hash but based on the specific sequences of bytes exhibited by the target file.
ClamAV body-based signature content has a [special format](BodySignatureFormat.md) to allow regex-like matching of data that is not entirely known. This format is used extensively in both Extended Signatures and Logical Signatures.
To test a new signature, first create a text file with the extension corresponding to the signature type (Ex: `.ldb` for logical signatures). Then, add the signature as it's own line within the file. This file can be passed to `clamscan` via the `-d` option, which tells ClamAV to load signatures from the file specified. If the signature is not formatted correctly, ClamAV will display an error - run `clamscan` with `--debug --verbose` to see additional information about the error message. Some common causes of errors include:
If the rule is formatted correctly, `clamscan` will load the signature(s) in and scan any files specified via the command line invocation (or all files in the current directory if none are specified). A successful detection will look like the following:
- The file may have exceeded one or more of the default scanning limits built-in to ClamAV. Try running `clamscan` with the following options to see if raising the limits addresses the issue: `--max-filesize=2000M --max-scansize=2000M --max-files=2000000 --max-recursion=2000000 --max-embeddedpe=2000M --max-htmlnormalize=2000000 --max-htmlnotags=2000000 --max-scriptnormalize=2000000 --max-ziptypercg=2000000 --max-partitions=2000000 --max-iconspe=2000000 --max-rechwp3=2000000 --pcre-match-limit=2000000 --pcre-recmatch-limit=2000000 --pcre-max-filesize=2000M`.
- If matching on HTML or text files, ClamAV might be performing normalization that causes the content of the scanned file to change. See the [HTML](#html) and [Text file](#text-file) sections for more details.
- libclamav may have been unable to unpack or otherwise process the file. See [Debug information from libclamav](#debug-information-from-libclamav) for more details.
In order to create efficient signatures for ClamAV it’s important to understand how the engine handles input files. The best way to see how it works is having a look at the debug information from libclamav. You can do it by calling `clamscan` with the `--debug` and `--leave-temps` flags. The first switch makes `clamscan` display all the interesting information from libclamav and the second one avoids deleting temporary files so they can be analyzed further.
LibClamAV debug: Bytecode executing hook id 257 (0 hooks)
attachment.exe: OK
[...]
```
No additional files get created by libclamav. By writing a signature for the decompressed file you have more chances that the engine will detect the target data when it gets compressed with another packer.
This method should be applied to all files for which you want to create signatures. By analyzing the debug information you can quickly see how the engine recognizes and preprocesses the data and what additional files get created. Signatures created for bottom-level temporary files are usually more generic and should help detecting the same malware in different forms.
ClamAV contains HTML normalization code which makes it easier to write signatures for HTML data that might differ based on white space, capitalization, and other insignificant differences. Running `sigtool --html-normalise` on a HTML file can be used to see what a file's contents will look like after normalization. This command should generate the following files:
The code automatically decodes JScript.encode parts and char ref’s (e.g. `f`). To create a successful signature for the input file type, the rule must match on the contents of one of the created files. Signatures matching on normalized HTML should have a target type of 3. For reference, see [Target Types](Signatures/FileTypes.md#Target-Types).
Similarly to HTML all ASCII text files get normalized (converted to lower-case, all superfluous white space and control characters removed, etc.) before scanning. Running `sigtool --ascii-normalise` on a text file will result in a normalized version being written to the file named 'normalised\_text'. Rules matching on normalized ASCII text should have a target type of 7. For reference, see [Target Types](Signatures/FileTypes.md#Target-Types).
If the file is compressed with UPX, FSG, Petite or another PE packer supported by libclamav, ClamAV will attempt to automatically unpack the executable and evaluate signatures against the unpacked executable. To inspect the executable that results from ClamAV's unpacking process, run `clamscan` with `--debug --leave-temps`. Example output for a FSG compressed file:
In the example above, `/tmp/clamav-f592b20f9329ac1c91f0e12137bcce6c` is the unpacked executable, and a signature can be written based off of this file.
`sigtool` pulls in libclamav and provides shortcuts to doing tasks that `clamscan` does behind the scenes. These can be really useful when writing a signature or trying to get information about a signature that might be causing FPs or performance problems.
-`--md5` / `--sha1` / `--sha256`: Generate the MD5/SHA1/SHA256 hash and calculate the file size, outputting both as a properly-formatted `.hdb`/`.hsb` signature
-`--decode`: Given a ClamAV signature from STDIN, show a more user-friendly representation of it. An example usage of this flag is `cat test.ldb | sigtool --decode`.
-`--hex-dump`: Given a sequence of bytes from STDIN, print the hex equivalent. An example usage of this flag is `echo -n "Match on this" | sigtool --hex-dump`.
-`--html-normalise`: Normalize the specified HTML file in the way that `clamscan` will before looking for rule matches. Writing signatures off of these files makes it easier to write rules for target type HTML (you'll know what white space, capitalization, etc. to expect). See the [HTML](#html) section for more details.
-`--ascii-normalise`: Normalize the specified ASCII text file in the way that `clamscan` will before looking for rule matches. Writing signatures off of this normalized file data makes it easier to write rules for target type Txt (you'll know what white space, capitalization, etc. to expect). See the [Text files](#text-files) sectino for more details.
-`--test-sigs`: Given a signature and a sample, determine whether the signature matches and, if so, display the offset into the file where the match occurred. This can be useful for investigating false positive matches in clean files.
CVD (ClamAV Virus Database) is a digitally signed container that includes signature databases in various text formats. The header of the container is a 512 bytes long string with colon separated fields:
```
ClamAV-VDB:build time:version:number of signatures:functionality level required:MD5 checksum:digital signature:builder name:build time (sec)
```
`sigtool --info` displays detailed information about a given CVD file:
To view the signature associated with a given detection name, the CVD files can be unpacked and the underlying text files searched for a rule definition using a tool like `grep`. To do this, use `sigtool`'s `--unpack` flag as follows:
- [CASC](https://github.com/Cisco-Talos/CASC) - CASC is a plugin for IDA Pro that allows the user to highlight sections of code and create a signature based on the underlying instructions (with options to ignore bytes associated with registers, addresses, and offsets). It also contains SigAlyzer, a tool to take an existing signature and locate the regions within the binary that match the subsignatures.
