Analyzing Linux Elf Malware

Analyzing Linux ELF Malware

When to Use

• A Linux server or container has been compromised and suspicious ELF binaries are found
• Analyzing Linux botnets (Mirai, Gafgyt, XorDDoS), cryptominers, or ransomware
• Investigating malware targeting cloud infrastructure, Docker containers, or Kubernetes pods
• Reverse engineering Linux rootkits and kernel modules
• Analyzing cross-platform malware compiled for Linux x86_64, ARM, or MIPS architectures

Do not use for Windows PE binary analysis; use PEStudio, Ghidra, or IDA for Windows malware.

Prerequisites

• Ghidra or IDA with Linux ELF support for disassembly and decompilation
• Linux analysis VM (Ubuntu 22.04 recommended) with development tools installed
• strace, ltrace, and GDB for dynamic analysis and debugging
• readelf, objdump, and nm from GNU binutils for static inspection
• Radare2 for quick binary triage and scripted analysis
• Docker for isolated container-based malware execution

Workflow

Step 1: Identify ELF Binary Properties

Examine the ELF header and basic properties:

# File type identification
file suspect_binary

# Detailed ELF header analysis
readelf -h suspect_binary

# Section headers
readelf -S suspect_binary

# Program headers (segments)
readelf -l suspect_binary

# Symbol table (if not stripped)
readelf -s suspect_binary
nm suspect_binary 2>/dev/null

# Dynamic linking information
readelf -d suspect_binary
ldd suspect_binary 2>/dev/null  # Only on matching architecture!

# Compute hashes
md5sum suspect_binary
sha256sum suspect_binary

# Check for packing/UPX
upx -t suspect_binary

# Python-based ELF analysis
from elftools.elf.elffile import ELFFile
import hashlib

with open("suspect_binary", "rb") as f:
    data = f.read()
    sha256 = hashlib.sha256(data).hexdigest()

with open("suspect_binary", "rb") as f:
    elf = ELFFile(f)

    print(f"SHA-256:      {sha256}")
    print(f"Class:        {elf.elfclass}-bit")
    print(f"Endian:       {elf.little_endian and 'Little' or 'Big'}")
    print(f"Machine:      {elf.header.e_machine}")
    print(f"Type:         {elf.header.e_type}")
    print(f"Entry Point:  0x{elf.header.e_entry:X}")

    # Check if stripped
    symtab = elf.get_section_by_name('.symtab')
    print(f"Stripped:     {'Yes' if symtab is None else 'No'}")

    # Section entropy analysis
    import math
    from collections import Counter
    for section in elf.iter_sections():
        data = section.data()
        if len(data) > 0:
            entropy = -sum((c/len(data)) * math.log2(c/len(data))
                          for c in Counter(data).values() if c > 0)
            if entropy > 7.0:
                print(f"  [!] High entropy section: {section.name} ({entropy:.2f})")

Step 2: Extract Strings and Indicators

Search for embedded IOCs and functionality clues:

# ASCII strings
strings suspect_binary > strings_output.txt

# Search for network indicators
grep -iE "(http|https|ftp)://" strings_output.txt
grep -iE "([0-9]{1,3}\.){3}[0-9]{1,3}" strings_output.txt
grep -iE "[a-zA-Z0-9.-]+\.(com|net|org|io|ru|cn)" strings_output.txt

# Search for shell commands
grep -iE "(bash|sh|wget|curl|chmod|/tmp/|/dev/)" strings_output.txt

# Search for crypto mining indicators
grep -iE "(stratum|xmr|monero|pool\.|mining)" strings_output.txt

# Search for SSH/credential theft
grep -iE "(ssh|authorized_keys|id_rsa|shadow|passwd)" strings_output.txt

# Search for persistence mechanisms
grep -iE "(crontab|systemd|init\.d|rc\.local|ld\.so\.preload)" strings_output.txt

# FLOSS for obfuscated strings (if available)
floss suspect_binary

Step 3: Analyze System Calls and Library Usage

Identify what system calls and libraries the malware uses:

# List imported functions (dynamically linked)
readelf -r suspect_binary | grep -E "socket|connect|exec|fork|open|write|bind|listen"

# Trace system calls during execution (in isolated VM only)
strace -f -e trace=network,process,file -o strace_output.txt ./suspect_binary

# Trace library calls
ltrace -f -o ltrace_output.txt ./suspect_binary

# Key system calls to watch:
# Network: socket, connect, bind, listen, accept, sendto, recvfrom
# Process: fork, execve, clone, kill, ptrace
# File:    open, read, write, unlink, rename, chmod
# Persistence: inotify_add_watch (file monitoring)

Step 4: Dynamic Analysis with GDB

Debug the malware to observe runtime behavior:

# Start GDB with the binary
gdb ./suspect_binary

# Set breakpoints on key functions
(gdb) break main
(gdb) break socket
(gdb) break connect
(gdb) break execve
(gdb) break fork

# Run and analyze
(gdb) run
(gdb) info registers    # View register state
(gdb) x/20s $rdi        # Examine string argument
(gdb) bt                # Backtrace
(gdb) continue

# For stripped binaries, break on entry point
(gdb) break *0x400580   # Entry point from readelf
(gdb) run

# Monitor network connections during execution
# In another terminal:
ss -tlnp  # List listening sockets
ss -tnp   # List established connections

Step 5: Reverse Engineer with Ghidra

Perform deep code analysis on the ELF binary:

Ghidra Analysis for Linux ELF:
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
1. Import: File -> Import -> Select ELF binary
   - Ghidra auto-detects ELF format and architecture
   - Accept default analysis options

2. Key analysis targets:
   - main() function (or entry point if stripped)
   - Socket creation and connection functions
   - Command dispatch logic (switch/case on received data)
   - Encryption/encoding routines
   - Persistence installation code
   - Self-propagation/scanning functions

3. For Mirai-like botnets, look for:
   - Credential list for brute-forcing (telnet/SSH)
   - Attack module selection (UDP flood, SYN flood, ACK flood)
   - Scanner module (port scanning for vulnerable devices)
   - Killer module (killing competing botnets)

4. For cryptominers, look for:
   - Mining pool connection (stratum protocol)
   - Wallet address strings
   - CPU/GPU utilization functions
   - Process hiding techniques

Step 6: Analyze Linux-Specific Persistence

Check for persistence mechanisms:

# Check for LD_PRELOAD rootkit
strings suspect_binary | grep "ld.so.preload"
# Malware writing to /etc/ld.so.preload can hook all dynamic library calls

# Check for crontab persistence
strings suspect_binary | grep -i "cron"

# Check for systemd service creation
strings suspect_binary | grep -iE "systemd|\.service|systemctl"

# Check for init script creation
strings suspect_binary | grep -iE "init\.d|rc\.local|update-rc"

# Check for SSH key injection
strings suspect_binary | grep -i "authorized_keys"

# Check for kernel module (rootkit) loading
strings suspect_binary | grep -iE "insmod|modprobe|init_module"

# Check for process hiding
strings suspect_binary | grep -iE "proc|readdir|getdents"

Key Concepts

Tools & Systems

• Ghidra: NSA reverse engineering tool with full ELF support for x86, x86_64, ARM, MIPS, and other Linux architectures
• Radare2: Open-source reverse engineering framework with command-line interface for quick binary analysis and scripting
• strace: Linux system call tracing tool for observing binary behavior including file, network, and process operations
• GDB: GNU Debugger for setting breakpoints, examining memory, and stepping through Linux binary execution
• pyelftools: Python library for parsing ELF files programmatically for automated analysis pipelines

Common Scenarios

Scenario: Analyzing a Cryptominer Found on a Compromised Linux Server

Context: A cloud server shows 100% CPU usage. Investigation reveals an unknown binary running from /tmp with a suspicious name. The binary needs analysis to confirm it is a cryptominer and identify the attacker's wallet and pool.

Approach:

1. Copy the binary to an analysis VM and compute SHA-256 hash
2. Run file and readelf to identify architecture and linking type
3. Extract strings and search for mining pool addresses (stratum+tcp://) and wallet addresses
4. Run with strace in a sandbox to observe network connections (mining pool connection)
5. Import into Ghidra to identify the mining algorithm and configuration extraction
6. Check for persistence mechanisms (crontab, systemd service, SSH keys)
7. Document all IOCs including pool address, wallet, C2 for updates, and persistence artifacts

Pitfalls:

• Running ldd on malware outside a sandbox (ldd can execute code in the binary)
• Not checking for ARM/MIPS architecture before attempting x86_64 execution
• Missing companion scripts (.sh files) that may handle persistence and cleanup
• Ignoring the initial access vector (how the miner was deployed: SSH brute force, web exploit, container escape)

Output Format

LINUX ELF MALWARE ANALYSIS REPORT
====================================
File:             /tmp/.X11-unix/.rsync
SHA-256:          e3b0c44298fc1c149afbf4c8996fb924...
Type:             ELF 64-bit LSB executable, x86-64
Linking:          Statically linked (all libraries embedded)
Stripped:         Yes
Size:             2,847,232 bytes
Packer:           UPX 3.96 (unpacked for analysis)

CLASSIFICATION
Family:           XMRig Cryptominer (modified)
Variant:          Custom build with C2 update mechanism

FUNCTIONALITY
[*] XMR (Monero) mining via RandomX algorithm
[*] Stratum pool connection for work submission
[*] C2 check-in for configuration updates
[*] Process name masquerading (argv[0] = "[kworker/0:0]")
[*] Competitor process killing (kills other miners)
[*] SSH key injection for re-access

NETWORK INDICATORS
Mining Pool:      stratum+tcp://pool.minexmr[.]com:4444
C2 Server:        hxxp://update.malicious[.]com/config
Wallet:           49jZ5Q3b...Monero_Wallet_Address...

PERSISTENCE
[1] Crontab entry: */5 * * * * /tmp/.X11-unix/.rsync
[2] SSH key added to /root/.ssh/authorized_keys
[3] Systemd service: /etc/systemd/system/rsync-daemon.service
[4] Modified /etc/ld.so.preload for process hiding

PROCESS HIDING
LD_PRELOAD:       /usr/lib/.libsystem.so
Hook:             readdir() to hide /tmp/.X11-unix/.rsync from ls
Hook:             fopen() to hide from /proc/*/maps reading