Reverse Engineering .NET Malware with dnSpy

When to Use

• A malware sample is identified as a .NET assembly (C#, VB.NET, F#) requiring decompilation
• Analyzing .NET-based malware families (AgentTesla, AsyncRAT, RedLine Stealer, Quasar RAT)
• Deobfuscating .NET code protected by ConfuserEx, SmartAssembly, or custom obfuscators
• Extracting hardcoded C2 configurations, encryption keys, and credentials from managed assemblies
• Debugging .NET malware at runtime to observe decryption routines and dynamic behavior

Do not use for native (unmanaged) PE binaries; use Ghidra or IDA for native code analysis.

Prerequisites

• dnSpy or dnSpyEx installed (https://github.com/dnSpyEx/dnSpy - community maintained fork)
• de4dot for automated .NET deobfuscation (https://github.com/de4dot/de4dot)
• ILSpy as an alternative decompiler for cross-validation
• .NET SDK installed for recompiling modified assemblies during analysis
• Isolated Windows VM for running dnSpy debugger on live malware
• Detect It Easy (DIE) for identifying the .NET obfuscator used

Workflow

Step 1: Identify .NET Assembly and Obfuscator

Verify the sample is a .NET binary and detect protection:

bash
# Check if file is .NET assembly
file suspect.exe
# Output should contain "PE32 executable" with .NET metadata

# Detect obfuscator with Detect It Easy
diec suspect.exe

# Python-based .NET detection
python3 << 'PYEOF'
import pefile

pe = pefile.PE("suspect.exe")

# Check for .NET COM descriptor
if hasattr(pe, 'DIRECTORY_ENTRY_COM_DESCRIPTOR'):
    print("[*] .NET assembly detected")
    print(f"    Runtime version: {pe.DIRECTORY_ENTRY_COM_DESCRIPTOR}")
else:
    # Check for mscoree.dll import (alternative detection)
    for entry in pe.DIRECTORY_ENTRY_IMPORT:
        if entry.dll.decode().lower() == "mscoree.dll":
            print("[*] .NET assembly detected (mscoree.dll import)")
            break
    else:
        print("[!] Not a .NET assembly")

# Check section names for .NET indicators
for section in pe.sections:
    name = section.Name.decode().rstrip('\x00')
    if name in ['.text', '.rsrc', '.reloc']:
        print(f"    Section: {name} (typical .NET)")
PYEOF

Step 2: Deobfuscate with de4dot

Remove common .NET obfuscation before manual analysis:

bash
# Run de4dot to identify and remove obfuscation
de4dot suspect.exe -o suspect_cleaned.exe

# Force specific deobfuscator
de4dot suspect.exe -p cf  # ConfuserEx
de4dot suspect.exe -p sa  # SmartAssembly
de4dot suspect.exe -p dr  # Dotfuscator
de4dot suspect.exe -p rv  # Reactor
de4dot suspect.exe -p bl  # Babel.NET

# Verbose output for debugging
de4dot -v suspect.exe -o suspect_cleaned.exe

# Handle multi-file assemblies
de4dot suspect.exe suspect_helper.dll -o cleaned/

Common .NET Obfuscators:
━━━━━━━━━━━━━━━━━━━━━━━
ConfuserEx:      String encryption, control flow, anti-debug, anti-tamper
SmartAssembly:   String encoding, flow obfuscation, pruning
Dotfuscator:     Renaming, string encryption, control flow
.NET Reactor:    Native code generation, necrobit, anti-debug
Babel.NET:       String encryption, resource encryption, code virtualization
Crypto Obfuscator: String encryption, anti-debug, watermarking
Custom:          Malware-specific obfuscation (manual de4dot configuration needed)

Step 3: Open in dnSpy and Analyze Code

Load the deobfuscated assembly in dnSpy for source-level analysis:

dnSpy Analysis Workflow:
━━━━━━━━━━━━━━━━━━━━━━━
1. File -> Open -> Select cleaned assembly
2. Navigate to the entry point:
   - Assembly Explorer -> <namespace> -> Program class -> Main method
   - Or: Right-click assembly -> Go to Entry Point

3. Key areas to examine:
   - Entry point (Main) for initialization and execution flow
   - Form classes for UI-based malware (RATs, stealers)
   - Network/HTTP classes for C2 communication
   - Crypto/encryption classes for data protection
   - Resource access for embedded payloads
   - Timer/Thread classes for persistence and scheduling

4. Navigation shortcuts:
   Ctrl+G       - Go to token/address
   Ctrl+Shift+K - Search assemblies
   F12          - Go to definition
   Ctrl+R       - Analyze (find usages)
   F5           - Start debugging
   F9           - Toggle breakpoint

Step 4: Extract Configuration and C2 Data

Locate hardcoded configuration in the decompiled source:

csharp
// Common .NET malware configuration patterns:

// Pattern 1: Static class with hardcoded values
public static class Config {
    public static string Host = "185.220.101.42";
    public static int Port = 4782;
    public static string Key = "GhOsT_RaT_2025";
    public static string Mutex = "AsyncMutex_6SI8OkPnk";
    public static bool Install = true;
    public static string InstallFolder = "%AppData%";
}

// Pattern 2: Encrypted strings decrypted at runtime
public static string Decrypt(string input) {
    byte[] data = Convert.FromBase64String(input);
    byte[] key = Encoding.UTF8.GetBytes("SecretKey123");
    for (int i = 0; i < data.Length; i++) {
        data[i] ^= key[i % key.Length];
    }
    return Encoding.UTF8.GetString(data);
}

// Pattern 3: Resource-embedded configuration
byte[] configData = Properties.Resources.config;
string config = AES.Decrypt(configData, derivedKey);

python
# Python script to extract .NET resource strings
import subprocess
import re
import base64

# Use monodis (Mono) or ildasm (.NET SDK) to dump IL
result = subprocess.run(
    ["monodis", "--output=il_dump.il", "suspect_cleaned.exe"],
    capture_output=True, text=True
)

# Search for string literals in IL dump
with open("il_dump.il", errors="ignore") as f:
    il_code = f.read()

# Find ldstr (load string) instructions
strings = re.findall(r'ldstr\s+"([^"]+)"', il_code)
for s in strings:
    # Check for Base64 encoded strings
    try:
        decoded = base64.b64decode(s).decode('utf-8', errors='ignore')
        if len(decoded) > 3 and decoded.isprintable():
            print(f"  Base64: {s[:40]}... -> {decoded[:100]}")
    except:
        pass
    # Check for URLs/IPs
    if re.match(r'https?://', s) or re.match(r'\d+\.\d+\.\d+\.\d+', s):
        print(f"  Network: {s}")

Step 5: Debug with dnSpy

Set breakpoints and debug the malware to observe runtime behavior:

dnSpy Debugging Workflow:
━━━━━━━━━━━━━━━━━━━━━━━
1. Set breakpoints on key methods:
   - String decryption functions (to capture decrypted values)
   - Network connection methods (to capture C2 URLs)
   - File write operations (to see what is dropped)
   - Registry modification methods (to see persistence)

2. Debug -> Start Debugging (F5)
   - Select the assembly to debug
   - Set command-line arguments if needed
   - Configure exception handling (break on all CLR exceptions)

3. At each breakpoint:
   - Inspect local variables (Locals window)
   - Evaluate expressions (Immediate window)
   - View call stack to understand execution context
   - Step over (F10) / Step into (F11) / Step out (Shift+F11)

4. Capture decrypted strings:
   - Set breakpoint after decryption function returns
   - Read the return value from the Locals window
   - Document all decrypted configuration values

Step 6: Document Findings

Compile analysis results into a structured report:

Analysis documentation should include:
- .NET assembly metadata (CLR version, target framework, compilation info)
- Obfuscator identified and deobfuscation method used
- Complete C2 configuration (hosts, ports, encryption keys, mutex names)
- Malware capabilities (keylogging, screen capture, file theft, etc.)
- Persistence mechanisms (registry, scheduled tasks, startup folder)
- Anti-analysis techniques (VM detection, debugger detection, sandbox evasion)
- Extracted IOCs (C2 IPs/domains, file hashes, mutex names, registry keys)
- YARA rule based on unique code patterns or strings

Key Concepts

| Term | Definition | |------|------------| | CIL/MSIL | Common Intermediate Language; the bytecode format .NET assemblies compile to, which can be decompiled back to high-level C#/VB.NET | | Metadata Token | Unique identifier for .NET types, methods, and fields within the assembly metadata tables; used for navigation in dnSpy | | de4dot | Open-source .NET deobfuscator that identifies and removes protection from many commercial and malware-specific obfuscators | | ConfuserEx | Popular open-source .NET obfuscator frequently used by malware authors for string encryption and control flow obfuscation | | String Encryption | Obfuscation technique replacing string literals with encrypted data and runtime decryption calls to hide IOCs from static analysis | | Resource Embedding | Storing configuration, payloads, or additional assemblies in .NET embedded resources, often encrypted with a key derived from assembly metadata | | Assembly.Load | .NET method loading assemblies from byte arrays in memory, enabling fileless execution of embedded payloads |

Tools & Systems

• dnSpy/dnSpyEx: Open-source .NET assembly editor, decompiler, and debugger supporting C# and VB.NET decompilation
• de4dot: Automated .NET deobfuscator supporting ConfuserEx, SmartAssembly, Dotfuscator, Reactor, and many other protectors
• ILSpy: Open-source .NET decompiler providing C#, VB.NET, and IL views of assembly code
• dotPeek: JetBrains' free .NET decompiler with symbol server and cross-reference navigation
• Detect It Easy (DIE): Multi-format file analyzer identifying .NET framework version, obfuscator, and compiler information

Common Scenarios

Scenario: Analyzing an AgentTesla Information Stealer

Context: A phishing email delivers a .NET executable identified as AgentTesla. The sample needs analysis to determine what credentials it steals, how it exfiltrates data, and its C2 configuration.

Approach:

1. Run Detect It Easy to identify the obfuscator (commonly ConfuserEx or custom)
2. Deobfuscate with de4dot to restore readable class/method names and decrypt strings
3. Open in dnSpy and navigate to the entry point to understand initialization
4. Locate the credential harvesting modules (browser, email, FTP, VPN password theft classes)
5. Find the exfiltration method (SMTP email, FTP upload, HTTP POST, Telegram bot API)
6. Extract C2 configuration (SMTP server, credentials, recipient email, or HTTP URL)
7. Set debugger breakpoints on the decryption function to capture all decrypted strings at once

Pitfalls:

• Analyzing without de4dot first (ConfuserEx makes manual analysis extremely difficult)
• Not checking for multi-stage loading (initial .NET executable may load additional assemblies from resources)
• Missing configuration stored in .NET resources rather than hardcoded strings
• Running the debugger without network isolation (AgentTesla will attempt to exfiltrate immediately)

Output Format

.NET MALWARE ANALYSIS REPORT
================================
Sample:           invoice_scanner.exe
SHA-256:          e3b0c44298fc1c149afbf4c8996fb924...
Type:             .NET Assembly (C#)
Framework:        .NET Framework 4.8
Obfuscator:       ConfuserEx v1.6
Deobfuscated:     Yes (de4dot -p cf)

CLASSIFICATION
Family:           AgentTesla v3
Type:             Information Stealer / Keylogger
Compile Date:     2025-09-10

C2 CONFIGURATION
Exfil Method:     SMTP (Email)
SMTP Server:      smtp.yandex[.]com:587
SMTP User:        exfil.account@yandex[.]com
SMTP Pass:        Str0ngP@ssw0rd2025
Recipient:        operator@protonmail[.]com
Interval:         30 minutes
Encryption:       AES-256 with key "AgentTesla_2025_key"

CAPABILITIES
[*] Browser credential theft (Chrome, Firefox, Edge, Opera)
[*] Email client passwords (Outlook, Thunderbird)
[*] FTP client credentials (FileZilla, WinSCP)
[*] VPN credentials (NordVPN, OpenVPN)
[*] Keylogging (SetWindowsHookEx)
[*] Screenshot capture (every 30 seconds)
[*] Clipboard monitoring

PERSISTENCE
Method:           Registry Run key + Scheduled Task
Registry:         HKCU\Software\Microsoft\Windows\CurrentVersion\Run\WindowsUpdate
Task:             \Microsoft\Windows\WindowsUpdate\Updater

EXTRACTED IOCs
SMTP Server:      smtp.yandex[.]com
Exfil Email:      exfil.account@yandex[.]com
Recipient:        operator@protonmail[.]com
Mutex:            AgentTesla_2025_Q3_MUTEX
Install Path:     %AppData%\Microsoft\Windows\svchost.exe