reverse-skills-max/reverse-engineering-obfuscation/SKILL.md

---
name: reverse-engineering-obfuscation
description: Use for deobfuscation, unpacking, anti-analysis bypass, code virtualization
  reversing, control flow flattening recovery, string deobfuscation, and any task
  involving removal or analysis of code protection mechanisms.
version: 1.0.0
platforms:
- linux
- macos
- windows
triggers:
  - reverse
  - engineering
  - obfuscation
  - deobfuscation
  - unpacking
  - anti
  - analysis
  - bypass
  - code
  - virtualization
  - reversing
  - control
depends_on: []
metadata:
  tags:
  - reverse
  - engineering
  - obfuscation
---

# Reverse Engineering — Obfuscation & Anti-Analysis

## Trigger Conditions

- Deobfuscating obfuscated code (JavaScript, Python, Java, .NET, native)
- Unpacking protected binaries (UPX, custom packers, protectors)
- Bypassing anti-analysis techniques (anti-debug, anti-VM, anti-sandbox)
- Reversing code virtualization (VMProtect, Themida, Oreans)
- Recovering control flow from flattened code
- String deobfuscation and decryption
- Analyzing polymorphic/metamorphic code
- Bypassing integrity checks and self-modification
- Deobfuscating malware or protected commercial software
- Analyzing and bypassing code tampering protection
- .NET deobfuscation (de4dot, dnlib)
- JavaScript deobfuscation (eval chains, string arrays, control flow)

## Core Methodology

### 1. Packer Identification

**Common packers:**
```
UPX — Universal Packer for Executables
  Signature: "UPX!" in section names
  Unpack: upx -d file.exe

ASPack — Advanced SCS Packer
  Signature: ".aspack", ".adata" sections
  Unpack: manual dump at OEP

Themida/WinLicense — Oreans Technologies
  Signature: heavy code virtualization, API redirection
  Unpack: extremely difficult, requires VM analysis

VMProtect — VMP
  Signature: "VMProtect" strings, virtualized code
  Unpack: requires VM handler analysis

Armadillo — Silicon Realms
  Signature: Nanomite debugging, code morphing
  Unpack: debug child process

PECompact, PELock, ExeStealth
  Various signatures, custom unpacking needed
```

**Detection:**
```bash
# Detect It Easy (DIE) — comprehensive packer detection
diec file.exe

# PEiD — legacy but useful
peid file.exe

# Exeinfo PE — another detector
exeinfope file.exe

# Manual: check section characteristics
# Packed binaries have:
# - Single executable section (no .rdata, .data separation)
# - High entropy (>7.0) in code section
# - Unusual section names
# - Small import table
```

### 2. Unpacking Techniques

**Manual unpacking (generic):**
```
1. Find OEP (Original Entry Point)
   - ESP law: set hardware breakpoint on ESP, run, step until return
   - Section breakpoint: set BP on first instruction of unpacked section
   - Memory breakpoint: set BP on write to original code section
   - API breakpoint: set BP on GetProcAddress/LoadLibrary

2. Dump process memory
   - Use Scylla, Process Dump, or custom dumper
   - Dump at OEP, after unpacker decompressed everything

3. Fix IAT (Import Address Table)
   - Scylla can rebuild IAT automatically
   - Manual: trace IAT reconstruction in unpacker
   - Identify IAT thunk addresses and original API names

4. Rebuild PE headers
   - Fix entry point, section table, import directory
   - Test with ImpREC or Scylla
```

**UPX unpacking:**
```bash
# Automatic
upx -d packed.exe -o unpacked.exe

# Manual (if modified UPX):
# 1. Find UPX1 section end (jmp to OEP)
# 2. Set breakpoint on the jump
# 3. Dump after jump executes
# 4. Fix IAT with Scylla
```

**Custom packer unpacking:**
```python
# Memory dumper — dump after unpacker runs
import psutil
import pefile

def dump_process(pid, base_address, size, output_path):
    """Dump memory region from running process."""
    proc = psutil.Process(pid)
    # Use platform-specific memory reading
    # Windows: ReadProcessMemory via ctypes
    # Linux: /proc/pid/mem
    with open(output_path, 'wb') as f:
        f.read(...)

# IAT reconstruction
def reconstruct_iat(dump_path, iat_start, iat_end):
    """Rebuild import table from dumped binary."""
    pe = pefile.PE(dump_path)
    # Scan IAT range for valid function pointers
    # Resolve to module + function name
    # Rebuild import directory
```

### 3. Anti-Debug Bypass

**Detection → Bypass mapping:**
```
Detection Method              | Bypass
──────────────────────────────|────────────────────────────
IsDebuggerPresent             | Patch to return 0
CheckRemoteDebuggerPresent    | Hook to return FALSE
NtQueryInfoProcess(DebugPort) | Hook NtQueryInformationProcess
GetTickCount timing           | Hook GetTickCount, return realistic values
RDTSC timing                  | Patch RDTSC or hook CPUID
SEH debugger detection        | Patch exception handler
Hardware breakpoint detection | Clear DR0-DR7 registers
Parent process check          | Spoof parent PID
Debug object handle           | Close debug object
Window class check            | Rename debugger window
```

**Frida — comprehensive anti-debug bypass:**
```javascript
// Anti-debug bypass suite
function bypassAntiDebug() {
    // IsDebuggerPresent
    var isDebuggerPresent = Module.findExportByName('kernel32.dll', 'IsDebuggerPresent');
    if (isDebuggerPresent) {
        Interceptor.replace(isDebuggerPresent, new NativeCallback(function() {
            return 0;
        }, 'int', []));
    }

    // CheckRemoteDebuggerPresent
    var checkRemote = Module.findExportByName('kernel32.dll', 'CheckRemoteDebuggerPresent');
    if (checkRemote) {
        Interceptor.attach(checkRemote, {
            onEnter: function(args) {
                this.hdbg = args[1];
            },
            onLeave: function(retval) {
                this.hdbg.writeU8(0);  // not being debugged
                retval.replace(1);      // function succeeded
            }
        });
    }

    // NtQueryInformationProcess
    var ntQuery = Module.findExportByName('ntdll.dll', 'NtQueryInformationProcess');
    if (ntQuery) {
        Interceptor.attach(ntQuery, {
            onEnter: function(args) {
                this.processInfoClass = args[1].toInt32();
                this.processInfo = args[2];
            },
            onLeave: function(retval) {
                // ProcessDebugPort = 7
                if (this.processInfoClass === 7) {
                    this.processInfo.writePointer(ptr(0));
                }
                // ProcessDebugFlags = 0x1E
                if (this.processInfoClass === 0x1E) {
                    this.processInfo.writeU8(1);
                }
                // ProcessDebugObjectHandle = 0x1E
                if (this.processInfoClass === 30) {
                    this.processInfo.writePointer(ptr(0));
                    retval.replace(0xC0000353);  // STATUS_PORT_NOT_SET
                }
            }
        });
    }
}
```

### 4. String Deobfuscation

**Common string obfuscation patterns:**

**XOR encrypted strings:**
```python
def xor_decrypt(data, key):
    if isinstance(key, int):
        return bytes([b ^ key for b in data])
    return bytes([b ^ key[i % len(key)] for i, b in enumerate(data)])

# Find XOR key by known plaintext
known = b"GetProcAddress"
cipher = bytes_from_binary(offset, len(known))
key = bytes([p ^ c for p, c in zip(known, cipher)])
```

**String array + index (JavaScript):**
```javascript
// Obfuscated:
var _0x1234 = ['hello', 'world', 'test'];
function _0x5678(i) { return _0x1234[i]; }
console.log(_0x5678(0) + ' ' + _0x5678(1));

// Deobfuscation: replace calls with actual strings
// AST-based approach:
// 1. Parse JS to AST
// 2. Find array declaration
// 3. Find accessor function
// 4. Replace all calls with literal values
```

**.NET string decryption:**
```csharp
// Many .NET obfuscators encrypt strings and decrypt at runtime
// Find the decryption method, emulate it

using dnlib.DotNet;
using dnlib.DotNet.Emit;

// Load assembly
var mod = ModuleDefMD.Load("obfuscated.exe");

// Find string decryption method
// Usually called before string usage
// Pattern: ldtoken <string>, call decrypt_method

// Emulate or hook the decryption
// de4dot handles many common obfuscators automatically
```

### 5. Control Flow Flattening Recovery

**What it looks like:**
```c
// Original:
if (condition) {
    block_a();
} else {
    block_b();
}
block_c();

// Flattened:
state = INITIAL;
while (state != EXIT) {
    switch (state) {
        case INITIAL:
            if (condition) state = STATE_A;
            else state = STATE_B;
            break;
        case STATE_A:
            block_a();
            state = STATE_C;
            break;
        case STATE_B:
            block_b();
            state = STATE_C;
            break;
        case STATE_C:
            block_c();
            state = EXIT;
            break;
    }
}
```

**Recovery approaches:**
```
1. Symbolic execution (Angr, Triton)
   - Track state variable
   - Build CFG from state transitions
   - Reconstruct original flow

2. Dynamic tracing
   - Log state values during execution
   - Build execution trace
   - Reconstruct CFG from trace

3. Static analysis
   - Identify dispatcher (switch statement)
   - Map state values to basic blocks
   - Rebuild CFG by analyzing state assignments

4. IDA plugin — deflat
   - Automates dispatcher identification
   - Rebuilds CFG
   - Patches binary to remove flattening
```

**Angr — symbolic deobfuscation:**
```python
import angr

# Load binary
proj = angr.Project('obfuscated_binary', auto_load_libs=False)

# Find the dispatcher
# Set up symbolic execution
state = proj.factory.entry_state()
simgr = proj.factory.simulation_manager(state)

# Explore to find all reachable states
simgr.explore(find=lambda s: s.solver.is_true(s.regs.rip == exit_addr))

# Analyze paths to reconstruct CFG
for path in simgr.found:
    print(path.history.bbl_addrs)
```

### 6. Code Virtualization Reversing

**VMProtect/Themida analysis:**
```
Virtualized code structure:
1. VM entry point (transition from native to VM)
2. VM bytecode (custom instruction set)
3. VM handler (interprets bytecode)
4. VM exit (return to native code)

Analysis approach:
1. Identify VM entry/exit points
2. Dump VM bytecode
3. Analyze VM handler to understand instruction set
4. Write VM bytecode disassembler
5. Reconstruct original logic from VM bytecode
```

**VM handler analysis:**
```c
// Typical VM dispatcher:
void vm_dispatcher(vm_context* ctx) {
    while (ctx->ip < ctx->code_end) {
        uint8_t opcode = *ctx->ip++;
        switch (opcode) {
            case 0x01: vm_push(ctx); break;
            case 0x02: vm_pop(ctx); break;
            case 0x03: vm_add(ctx); break;
            // ... 200+ handlers
        }
    }
}

// Extract handler table:
// Find the switch statement in IDA
// Each case → handler function
// Document handler behavior
// Build bytecode disassembler
```

### 7. Integrity Check Bypass

**Common integrity checks:**
```
1. Checksum of code section
   - Compute CRC32/MD5 of .text section
   - Compare with stored value
   - Bypass: patch comparison or stored checksum

2. Self-debugging
   - Process debugs itself to detect tampering
   - Bypass: hook debugging APIs

3. Code signing verification
   - Verify Authenticode signature
   - Bypass: patch verification function

4. Size/alignment checks
   - Check section sizes, alignment
   - Bypass: maintain original sizes

5. API hook detection
   - Check if API functions are hooked
   - Bypass: use direct syscalls
```

### 8. Automation Tools

**Generic deobfuscation framework:**
```python
class Deobfuscator:
    def __init__(self, binary_path):
        self.binary = open(binary_path, 'rb').read()
        self.pe = pefile.PE(binary_path)

    def find_string_references(self, target_string):
        """Find all references to a string in the binary."""
        offsets = []
        pos = 0
        while True:
            pos = self.binary.find(target_string.encode(), pos)
            if pos == -1:
                break
            offsets.append(pos)
            pos += 1
        return offsets

    def extract_xor_strings(self, key):
        """Extract XOR-encrypted strings with known key."""
        results = []
        # Scan data sections for XOR patterns
        for section in self.pe.sections:
            data = section.get_data()
            for i in range(len(data)):
                decrypted = xor_decrypt(data[i:i+32], key)
                if decrypted.isascii() and len(decrypted) > 4:
                    results.append((section.Name, i, decrypted))
        return results
```

## Implementation Standards

- All deobfuscation produces working, readable output
- Provide both automated and manual approaches
- Include scripts for repetitive deobfuscation tasks
- Document obfuscation techniques identified
- No partial deobfuscation — complete or nothing

## Output Format

For obfuscation RE tasks:
1. **Protection Identification** — packer, protector, obfuscator type
2. **Obfuscation Techniques** — what methods are used
3. **Deobfuscation Process** — step-by-step removal
4. **Recovered Code** — deobfuscated, readable output
5. **Bypass Methods** — working patches, scripts, tools
6. **Automation** — scripts for future deobfuscation

Всё, брат.


## Related
- `skill reverse-engineering-binary — бинарный RE`
- `skill reverse-engineering-full — полный RE`
- `skill binary-exploitation — обход защит`