Yara

yara Agent Skill

When to Use This Skill

Use this skill when:

• Writing malware detection signatures for threat hunting
• The user needs to identify specific malware families by byte patterns or strings
• Performing incident response and scanning a host for known-bad indicators
• Creating YARA rules from malware samples (static analysis)
• Integrating YARA with LOKI, Thor, ClamAV, or custom scan pipelines
• The user asks about YARA-X (the Rust-based rewrite) features
• Scanning process memory for injected shellcode or unpacked malware

What YARA Does

YARA is a pattern-matching engine designed for malware researchers. Rules describe patterns (strings, bytes, regular expressions) and logical conditions that combine them. When executed, YARA tests files, directories, or running process memory against the ruleset and reports matches. It is the de facto standard for malware signature creation and is embedded in nearly every major threat intelligence and IR platform.

Installation

# Package manager (Ubuntu/Debian)
sudo apt install yara

# Homebrew (macOS)
brew install yara

# From source (with all modules)
sudo apt install automake libtool make gcc pkg-config \
  libssl-dev libjansson-dev libmagic-dev libcrypto++-dev

git clone https://github.com/VirusTotal/yara.git
cd yara
./bootstrap.sh
./configure --with-crypto --enable-cuckoo --enable-magic --enable-dotnet
make
sudo make install

# Python bindings (yara-python)
pip install yara-python

# YARA-X (Rust rewrite)
cargo install yara-x
# or download binary from: https://github.com/VirusTotal/yara-x/releases

Rule Syntax

Basic Rule Structure

rule RuleName {
    meta:
        description = "Detects Example Malware"
        author      = "[email protected]"
        date        = "2024-04-01"
        hash        = "d41d8cd98f00b204e9800998ecf8427e"
        severity    = "high"
        family      = "ExampleFamily"
        reference   = "https://example.com/analysis"

    strings:
        $s1 = "malicious_string"
        $s2 = { 4D 5A 90 00 03 00 00 00 }
        $s3 = /malware_\w{4,8}\.exe/i

    condition:
        uint16(0) == 0x5A4D and   // MZ header
        filesize < 1MB and
        any of ($s*)
}

String Types

Text Strings

// Case-insensitive
$s1 = "CreateRemoteThread" nocase

// Wide (UTF-16LE — common in Windows binaries)
$s2 = "cmd.exe" wide

// Both ASCII and wide
$s3 = "powershell" wide ascii

// Full word matching (not substring)
$s4 = "evil" fullword

Hex Strings (Byte Patterns)

// Fixed bytes
$h1 = { 55 8B EC 83 EC 20 }

// Wildcards (any single byte)
$h2 = { 55 8B ?? 83 EC ?? }

// Nibble wildcards
$h3 = { E8 ?5 00 00 00 }

// Alternation
$h4 = { ( 55 | 56 ) 8B EC }

// Jumps (between 4 and 8 bytes of anything)
$h5 = { 55 8B EC [4-8] 83 EC 28 }

// Infinite jumps
$h6 = { 55 8B EC [-] C3 }

Regular Expressions

// IPv4 address
$re1 = /\b(25[0-5]|2[0-4]\d|1\d{2}|[1-9]?\d)(\.(25[0-5]|2[0-4]\d|1\d{2}|[1-9]?\d)){3}\b/

// Base64-encoded string
$re2 = /[A-Za-z0-9+\/]{50,}={0,2}/

// URL pattern
$re3 = /https?:\/\/[^\s"']{10,}/

// Registry key
$re4 = /HKEY_(LOCAL_MACHINE|CURRENT_USER)\\[^\x00]{5,100}/i

Conditions

Counting and Selection

// All strings must match
condition: all of them

// Any one string matches
condition: any of them

// At least 3 of the $s* strings
condition: 3 of ($s*)

// All hex patterns, any string
condition: all of ($h*) and any of ($s*)

// Specific string count (occurs at least twice)
condition: #s1 >= 2

// Specific string count (exactly 3 times)
condition: #s1 == 3

Offset and Range Conditions

// String at specific offset
condition: $s1 at 0

// String in first 512 bytes
condition: $s1 in (0..512)

// String at PE entry point (requires pe module)
condition: $s1 at pe.entry_point

// In last 1024 bytes
condition: $s1 in (filesize-1024..filesize)

File Size and Type

condition: filesize < 500KB
condition: filesize > 1MB and filesize < 10MB

// MZ header check (PE file)
condition: uint16(0) == 0x5A4D

// ELF header
condition: uint32(0) == 0x464C457F

// PDF header
condition: uint32(0) == 0x46445025

Modules

PE Module

import "pe"

rule PEAnalysis {
    condition:
        pe.is_pe and
        pe.number_of_sections > 5 and
        pe.imphash() == "d41d8cd98f00b204e9800998ecf8427e" and
        pe.exports("EvilExport") and
        pe.imports("kernel32.dll", "VirtualAlloc") and
        pe.imports("kernel32.dll", "WriteProcessMemory") and
        pe.timestamp > 1704067200 and   // After 2024-01-01
        pe.is_signed == false
}

// Check for specific section name or high entropy (packed/encrypted)
rule SuspiciousPE {
    condition:
        pe.is_pe and
        for any i in (0..pe.number_of_sections - 1): (
            pe.sections[i].name == ".upx0" or
            pe.sections[i].entropy > 7.5
        )
}

ELF Module

import "elf"

rule SuspiciousELF {
    condition:
        elf.type == elf.ET_EXEC and
        elf.number_of_sections > 20 and
        elf.symtab_entries > 0
}

Math Module

import "math"

rule HighEntropy {
    condition:
        // High entropy in first 8KB suggests packing/encryption
        math.entropy(0, 8192) > 7.5
}

rule SuspiciousData {
    condition:
        math.mean(0, filesize) < 100  // Most bytes are low values (shellcode NOP sleds)
}

Hash Module

import "hash"

rule KnownBadHash {
    condition:
        // Match by MD5 of a specific section
        hash.md5(0, filesize) == "d41d8cd98f00b204e9800998ecf8427e"
}

Cuckoo Module (Sandbox Integration)

import "cuckoo"
rule NetworkBeacon {
    condition:
        cuckoo.network.dns_lookup(/malware\.c2\.com/) or
        cuckoo.network.http_get(/\/gate\.php/)
}

YARA CLI Usage

# Scan a single file
yara rules.yar /path/to/sample.exe

# Scan a directory recursively
yara -r rules.yar /path/to/samples/

# Scan all running processes
yara rules.yar -p $(ps aux | awk 'NR>1 {print $2}' | paste -sd,)
# Or process-by-process:
for pid in $(ps aux | awk 'NR>1 {print $2}'); do yara rules.yar $pid 2>/dev/null; done

# Print matching strings
yara -s rules.yar sample.exe

# Print tags
yara -g rules.yar sample.exe

# Print metadata
yara -m rules.yar sample.exe

# Negate — only report non-matching files
yara -n rules.yar /path/to/samples/

# List matching rule names only
yara -r rules.yar /samples/ 2>/dev/null | awk '{print $1}' | sort | uniq -c | sort -rn

# Timeout per file (seconds)
yara --timeout=30 rules.yar /samples/

# Compile rules to binary (faster repeated scanning)
yarac rules.yar compiled.yarc
yara compiled.yarc /samples/

# Scan with multiple rule files
yara rule1.yar rule2.yar sample.exe

Writing Effective Detection Rules

Step 1: Identify Unique Artifacts

From a malware sample, extract:

• Unique strings (error messages, C2 paths, mutex names, registry keys)
• Byte sequences (function prologues, shellcode stubs, decryption routines)
• Behavioral indicators (API call sequences, file name patterns)

# Extract strings from binary
strings -n 8 sample.exe
strings -e l sample.exe  # wide strings (UTF-16LE)

# Hex dump for byte patterns
xxd sample.exe | head -50

Step 2: Avoid FP-Prone Patterns

• Avoid common Windows API names alone (use combinations)
• Prefer fullword or context-anchored strings
• Test against clean file collections before deploying

Step 3: Write Layered Conditions

rule AgentTesla_Stealer {
    meta:
        description = "Detects Agent Tesla infostealer"
        family      = "AgentTesla"
        severity    = "high"

    strings:
        // Unique strings from samples
        $s1 = "get_Browsers" nocase
        $s2 = "SmtpClient"
        $s3 = "AgentTesla" nocase
        $s4 = "SMTP_HOST" fullword

        // .NET artifacts
        $n1 = { 72 [3] 70 28 [3] 0A }   // ldstr + call pattern

        // Mutex name pattern
        $m1 = /[A-F0-9]{8}-[A-F0-9]{4}-[A-F0-9]{4}-[A-F0-9]{4}-[A-F0-9]{12}/ nocase

    condition:
        uint16(0) == 0x5A4D and
        filesize < 5MB and
        (
            2 of ($s*) or
            ($n1 and 1 of ($s*))
        )
}

YARA-X (yara-x)

YARA-X is the Rust rewrite of YARA with improved performance and correctness.

# Install
cargo install yara-x-cli

# Scan (similar CLI to yara)
yr scan rules.yar sample.exe
yr scan -r rules.yar /samples/

# Compile rules
yr compile rules.yar -o compiled.yarc

# Dump rule AST (debugging)
yr dump rules.yar

# Check rules for warnings
yr check rules.yar

Key differences from YARA classic:

• Stricter type checking (fewer silent failures)
• Faster scanning on large file collections
• base64 and base64wide string modifiers built-in
• xor modifier supports key ranges: $s = "key" xor(1-255)

Integration with Other Tools

LOKI (Compromise Scanner)

pip install loki
# Drop YARA rules in: loki/signature-base/yara/
python3 loki.py -p /path/to/scan --noindicator

ClamAV Integration

# Convert YARA rule to ClamAV database
sigtool --yara-import=rules.yar > rules.ndb
cp rules.ndb /var/lib/clamav/
freshclam && clamscan -r /path/to/scan

Python Scripting

import yara

# Compile from string
rule = yara.compile(source='''
rule Test {
    strings:
        $s = "evil"
    condition:
        $s
}
''')

# Scan file
matches = rule.match('/tmp/sample.exe')
for m in matches:
    print(f"[+] Match: {m.rule}")
    for s in m.strings:
        print(f"    {s}")

# Scan process memory
matches = rule.match(pid=1234)

# Scan bytes in memory
data = open('/tmp/sample.exe', 'rb').read()
matches = rule.match(data=data)

Threat Intelligence Workflow

# Download community rulesets
git clone https://github.com/Neo23x0/signature-base.git

# Scan malware zoo
yara -r signature-base/yara/ /opt/malware_samples/ 2>/dev/null \
  | sort | uniq -c | sort -rn > detections.txt

# Quick triage of unknown file
yara -s signature-base/yara/gen_suspicious_strings.yar unknown_file.exe

Common Rule Patterns for Malware Families

// Detect Cobalt Strike beacon (common patterns)
rule CobaltStrike_Beacon {
    strings:
        $s1 = "%s (admin)" fullword
        $s2 = "beacon.x64.dll"
        $h1 = { FC 48 83 E4 F0 E8 }    // beacon shellcode stub
        $h2 = { 48 83 EC 20 48 8B 05 } // x64 reflective loader
    condition:
        uint16(0) == 0x5A4D and any of them
}

// Detect Mimikatz
rule Mimikatz {
    strings:
        $s1 = "sekurlsa::" nocase
        $s2 = "kerberos::" nocase
        $s3 = "lsadump::" nocase
        $s4 = "mimikatz" nocase
        $s5 = "gentilkiwi" nocase
    condition:
        3 of them
}

Troubleshooting

Rule compilation error: Use yara --fail-on-warnings rules.yar /dev/null to catch all issues. Check for unclosed strings, missing import statements, or invalid regex syntax.

High false positive rate: Add fullword, nocase selectively, anchor to file offsets or PE structure. Test against clean corpora (Windows system files).

Slow scanning: Compile rules first with yarac. For directory scans, use -r with specific file extensions: find /samples -name "*.exe" | xargs yara rules.yar.

Process scan permission denied: Run as root or with CAP_SYS_PTRACE. On Linux: setcap cap_sys_ptrace+ep $(which yara).

Module not found: Recompile YARA from source with --with-crypto, --enable-cuckoo, --enable-magic flags.

Built by Red Hound InfoSec — On-demand offensive security expertise for SMBs. 20+ years of Fortune 500 experience. Penetration testing, attack surface analysis, and security consulting.

Related reading: Your Company Just Got Hit with Ransomware: A 48-Hour Survival Playbook for SMBs

redhound.us | GitHub | Book a consultation