Deobfuscating Powershell Obfuscated Malware

Deobfuscating PowerShell Obfuscated Malware

Overview

PowerShell is heavily abused by malware authors due to its deep Windows integration and powerful scripting capabilities. Obfuscation techniques include string concatenation, Base64 encoding, character substitution, Invoke-Expression layering, SecureString abuse, environment variable manipulation, and tick-mark insertion. Modern malware uses multiple obfuscation layers requiring iterative deobfuscation. Tools like PSDecode, PowerDecode, and PowerPeeler automate much of this process, while manual AST (Abstract Syntax Tree) analysis handles custom obfuscation. PowerPeeler achieves a 95% deobfuscation correctness rate using instruction-level dynamic analysis of expression-related AST nodes.

When to Use

• When performing authorized security testing that involves deobfuscating powershell obfuscated malware
• When analyzing malware samples or attack artifacts in a controlled environment
• When conducting red team exercises or penetration testing engagements
• When building detection capabilities based on offensive technique understanding

Detection Gaps & Validation

• One -EncodedCommand decode is rarely the end. Analysts decode the base64 UTF-16LE blob and stop - but the result usually contains another layer: FromBase64String + GZipStream/DeflateStream, -bxor loops, or ConvertTo-SecureString with an embedded key. Re-classify the decoded output and keep peeling.
• AMSI/logging bypass hides the real payload. Samples patch amsi.dll/AmsiScanBuffer or null EtwEventWrite so Script Block Logging (4104) never captures the deobfuscated stage. If 4104 looks empty for a known-malicious run, assume a bypass and recover the payload from memory instead.
• IEX-replacement tracing has limits. Swapping IEX->Write-Output works for staged droppers, but misses &/. invocation, .Invoke(), runspace/[scriptblock]::Create(), and Start-Process hand-offs - grep for every execution sink, not just IEX/Invoke-Expression.
• Confirm a hit by detonating in an isolated VM with Script Block Logging + Sysmon and PSDecode/PowerDecode, then diff the captured final stage against your static decode; extract URLs, IPs, and dropped paths and verify they match observed network/file behavior.
• Tick-mark/case/concatenation noise causes both FPs and FNs. Backticks are stripped except real escapes (`n,`t), and $env:-substring tricks split keywords - normalize before regex matching or you'll miss commands and misclassify benign admin scripts.
• Benign lookalikes: legitimate installers, EDR agents, and DSC/automation use -EncodedCommand, compression, and remoting. Judge on the decoded intent (download-exec, AMSI tamper, defender exclusion), not on encoding presence.

Prerequisites

• Python 3.9+ with base64, re, subprocess modules
• PowerShell 5.1+ or PowerShell 7+ (for AST access)
• PSDecode (Install-Module PSDecode)
• PowerDecode (https://github.com/Malandrone/PowerDecode)
• Isolated VM or sandbox for safe script execution
• CyberChef for manual encoding transformations
• Understanding of PowerShell AST and Invoke-Expression patterns

Key Concepts

Common Obfuscation Techniques

PowerShell malware employs layered obfuscation to evade static detection. String concatenation splits commands across variables ($a='In'+'voke'). Base64 encoding wraps entire scripts in -EncodedCommand parameters. Character code arrays use [char] casting ([char[]](73,69,88)|%{$r+=$_}). Environment variable abuse reads substrings from $env: paths. Tick-mark insertion adds backticks between characters that PowerShell ignores (Invoke-Expression`). SecureString conversion encrypts strings using ConvertTo-SecureString with embedded keys.

AST-Based Deobfuscation

PowerShell's Abstract Syntax Tree exposes the parsed structure of scripts regardless of surface-level obfuscation. By walking the AST and evaluating expression nodes, analysts can resolve concatenated strings, decode encoded values, and reconstruct the original commands. PowerPeeler uses this approach at the instruction level, monitoring the execution process to correlate AST nodes with their evaluated results.

Dynamic Execution Tracing

By replacing Invoke-Expression (IEX) with Write-Output, analysts can safely capture the deobfuscated script content that would normally be executed. This technique works across multiple layers by iteratively replacing IEX calls until the final payload is revealed.

Workflow

Step 1: Identify Obfuscation Layers

#!/usr/bin/env python3
"""Identify and classify PowerShell obfuscation techniques."""
import re
import base64
import sys


def analyze_obfuscation(script_content):
    """Identify obfuscation techniques used in PowerShell script."""
    techniques = []

    # Check for Base64 encoded command
    b64_pattern = re.compile(
        r'-[Ee](?:nc(?:odedcommand)?)\s+([A-Za-z0-9+/=]{20,})',
        re.IGNORECASE
    )
    if b64_pattern.search(script_content):
        techniques.append("Base64 EncodedCommand")

    # Check for FromBase64String
    if re.search(r'\[Convert\]::FromBase64String', script_content, re.IGNORECASE):
        techniques.append("Base64 FromBase64String")

    # Check for string concatenation
    concat_count = script_content.count("'+'") + script_content.count('"+"')
    if concat_count > 3:
        techniques.append(f"String Concatenation ({concat_count} joins)")

    # Check for char array construction
    if re.search(r'\[char\]\s*\d+', script_content, re.IGNORECASE):
        techniques.append("Character Code Array")

    # Check for Invoke-Expression variants
    iex_patterns = [
        r'Invoke-Expression',
        r'\bIEX\b',
        r'\.\s*\(\s*\$',
        r'&\s*\(\s*\$',
        r'\|\s*IEX',
        r'\|\s*Invoke-Expression',
    ]
    for pattern in iex_patterns:
        if re.search(pattern, script_content, re.IGNORECASE):
            techniques.append(f"Invoke-Expression variant: {pattern}")

    # Check for tick-mark obfuscation
    tick_count = script_content.count('`')
    if tick_count > 5:
        techniques.append(f"Tick-mark Insertion ({tick_count} backticks)")

    # Check for environment variable abuse
    if re.search(r'\$env:', script_content, re.IGNORECASE):
        env_refs = re.findall(r'\$env:\w+', script_content, re.IGNORECASE)
        if len(env_refs) > 2:
            techniques.append(f"Environment Variable Abuse ({len(env_refs)} refs)")

    # Check for SecureString
    if re.search(r'ConvertTo-SecureString', script_content, re.IGNORECASE):
        techniques.append("SecureString Encryption")

    # Check for compression
    if re.search(r'IO\.Compression|DeflateStream|GZipStream',
                 script_content, re.IGNORECASE):
        techniques.append("Compression (Deflate/GZip)")

    # Check for XOR encoding
    if re.search(r'-bxor\s+\d+', script_content, re.IGNORECASE):
        techniques.append("XOR Encoding")

    # Check for Replace chain
    replace_count = len(re.findall(r'\.Replace\(', script_content))
    if replace_count > 2:
        techniques.append(f"Replace Chain ({replace_count} replacements)")

    return techniques


def decode_base64_command(script_content):
    """Extract and decode Base64 encoded commands."""
    b64_match = re.search(
        r'-[Ee](?:nc(?:odedcommand)?)\s+([A-Za-z0-9+/=]{20,})',
        script_content, re.IGNORECASE
    )
    if b64_match:
        encoded = b64_match.group(1)
        try:
            decoded = base64.b64decode(encoded).decode('utf-16-le')
            return decoded
        except Exception:
            return None
    return None


def remove_tick_marks(script_content):
    """Remove PowerShell tick-mark obfuscation."""
    # Remove backticks that are not escape sequences
    escape_chars = {'`n', '`r', '`t', '`a', '`b', '`f', '`v', '`0', '``'}
    result = []
    i = 0
    while i < len(script_content):
        if script_content[i] == '`' and i + 1 < len(script_content):
            pair = script_content[i:i+2]
            if pair in escape_chars:
                result.append(pair)
                i += 2
            else:
                # Skip the backtick, keep the next char
                result.append(script_content[i+1])
                i += 2
        else:
            result.append(script_content[i])
            i += 1
    return ''.join(result)


def resolve_string_concat(script_content):
    """Resolve simple string concatenation patterns."""
    # Pattern: 'str1' + 'str2'
    pattern = re.compile(r"'([^']*)'\s*\+\s*'([^']*)'")
    while pattern.search(script_content):
        script_content = pattern.sub(lambda m: f"'{m.group(1)}{m.group(2)}'",
                                      script_content)
    # Pattern: "str1" + "str2"
    pattern = re.compile(r'"([^"]*)"\s*\+\s*"([^"]*)"')
    while pattern.search(script_content):
        script_content = pattern.sub(lambda m: f'"{m.group(1)}{m.group(2)}"',
                                      script_content)
    return script_content


if __name__ == "__main__":
    if len(sys.argv) < 2:
        print(f"Usage: {sys.argv[0]} <powershell_script>")
        sys.exit(1)

    with open(sys.argv[1], 'r', errors='replace') as f:
        content = f.read()

    print("[+] Obfuscation Analysis")
    print("=" * 60)
    techniques = analyze_obfuscation(content)
    for t in techniques:
        print(f"  - {t}")

    # Attempt automatic deobfuscation
    print("\n[+] Attempting Deobfuscation")
    print("=" * 60)

    # Layer 1: Remove tick marks
    deobfuscated = remove_tick_marks(content)

    # Layer 2: Resolve string concatenation
    deobfuscated = resolve_string_concat(deobfuscated)

    # Layer 3: Decode Base64
    b64_decoded = decode_base64_command(deobfuscated)
    if b64_decoded:
        print("[+] Base64 decoded content:")
        print(b64_decoded[:2000])
        deobfuscated = b64_decoded

    print(f"\n[+] Deobfuscated script length: {len(deobfuscated)} chars")
    output_file = sys.argv[1] + ".deobfuscated.ps1"
    with open(output_file, 'w') as f:
        f.write(deobfuscated)
    print(f"[+] Saved to {output_file}")

Step 2: Multi-Layer IEX Replacement

import subprocess
import tempfile
import os

def iex_replacement_deobfuscate(script_content, max_layers=10):
    """Iteratively replace IEX with Write-Output to unwrap layers."""
    # IEX replacement patterns
    replacements = [
        (r'\bInvoke-Expression\b', 'Write-Output'),
        (r'\bIEX\b', 'Write-Output'),
        (r'\|\s*IEX\b', '| Write-Output'),
    ]

    current = script_content
    layers = []

    for layer_num in range(max_layers):
        # Apply IEX replacements
        modified = current
        for pattern, replacement in replacements:
            modified = re.sub(pattern, replacement, modified, flags=re.IGNORECASE)

        if modified == current and layer_num > 0:
            print(f"  [+] No more IEX layers found at layer {layer_num}")
            break

        # Write to temp file and execute in constrained PowerShell
        with tempfile.NamedTemporaryFile(mode='w', suffix='.ps1',
                                          delete=False) as tmp:
            tmp.write(modified)
            tmp_path = tmp.name

        try:
            result = subprocess.run(
                ['powershell', '-NoProfile', '-ExecutionPolicy', 'Bypass',
                 '-File', tmp_path],
                capture_output=True, text=True, timeout=30
            )

            output = result.stdout.strip()
            if output and output != current:
                print(f"  [+] Layer {layer_num + 1}: Unwrapped "
                      f"{len(output)} chars")
                layers.append({
                    "layer": layer_num + 1,
                    "technique": "IEX replacement",
                    "content_length": len(output),
                })
                current = output
            else:
                break

        except subprocess.TimeoutExpired:
            print(f"  [!] Layer {layer_num + 1}: Execution timeout")
            break
        finally:
            os.unlink(tmp_path)

    return current, layers

Step 3: Extract IOCs from Deobfuscated Script

def extract_iocs_from_script(deobfuscated_content):
    """Extract indicators of compromise from deobfuscated PowerShell."""
    iocs = {
        "urls": [],
        "ips": [],
        "domains": [],
        "file_paths": [],
        "registry_keys": [],
        "commands": [],
        "base64_blobs": [],
    }

    # URLs
    url_pattern = re.compile(
        r'https?://[^\s'"<>)\]]+', re.IGNORECASE
    )
    iocs["urls"] = list(set(url_pattern.findall(deobfuscated_content)))

    # IP addresses
    ip_pattern = re.compile(
        r'\b(?:\d{1,3}\.){3}\d{1,3}\b'
    )
    iocs["ips"] = list(set(ip_pattern.findall(deobfuscated_content)))

    # File paths
    path_pattern = re.compile(
        r'[A-Za-z]:\\[^\s'"<>|]+|'
        r'\\\\[^\s'"<>|]+|'
        r'%(?:APPDATA|TEMP|USERPROFILE|PROGRAMFILES)%[^\s'"<>|]*',
        re.IGNORECASE
    )
    iocs["file_paths"] = list(set(path_pattern.findall(deobfuscated_content)))

    # Registry keys
    reg_pattern = re.compile(
        r'(?:HKLM|HKCU|HKCR|HKU|HKCC)(?:\\[^\s'"<>|]+)+',
        re.IGNORECASE
    )
    iocs["registry_keys"] = list(set(reg_pattern.findall(deobfuscated_content)))

    # Suspicious commands
    suspicious_cmds = [
        'New-Object Net.WebClient',
        'DownloadString', 'DownloadFile', 'DownloadData',
        'Start-Process', 'Invoke-WebRequest',
        'New-Object IO.MemoryStream',
        'Reflection.Assembly',
        'Add-MpPreference -ExclusionPath',
        'Set-MpPreference -DisableRealtimeMonitoring',
        'New-ScheduledTask', 'Register-ScheduledTask',
    ]
    for cmd in suspicious_cmds:
        if cmd.lower() in deobfuscated_content.lower():
            iocs["commands"].append(cmd)

    return iocs

Validation Criteria

• All obfuscation layers identified and classified correctly
• Base64 encoded commands decoded to readable PowerShell
• Tick-mark and string concatenation obfuscation resolved
• IEX replacement reveals next-stage payloads
• URLs, IPs, and file paths extracted from final deobfuscated stage
• Deobfuscated script matches observed malware behavior in sandbox

References

• PSDecode - PowerShell Deobfuscation
• PowerDecode - Multi-layer Deobfuscation
• PowerPeeler - Instruction-level Deobfuscation
• SentinelOne - Deconstructing PowerShell Obfuscation
• MITRE ATT&CK T1059.001 - PowerShell