Hardware Breakpoint SSN Extraction

Overview

Hardware Breakpoint SSN extraction is the most advanced resolution technique in SysWhispers4, using CPU debug registers (DR0-DR3) and a Vectored Exception Handler (VEH) to capture syscall numbers without reading potentially-hooked function bytes. This method extracts SSNs at the exact moment they’re loaded into EAX — after mov eax, <SSN> executes but before the syscall instruction.

Complexity: This technique requires understanding of:

x86/x64 debug registers (DR0-DR7)
Vectored Exception Handling (VEH)
CONTEXT manipulation
Single-step exceptions

Use RecycledGate or FreshyCalls unless you specifically need runtime SSN capture.

The Technique: Breakpoint-on-Syscall

Core Concept

Instead of reading the mov eax, <SSN> opcode, we execute it under a hardware breakpoint and capture the result from the CPU register:

; NT stub (e.g., NtAllocateVirtualMemory):
NtAllocateVirtualMemory:
    4C 8B D1              mov r10, rcx          ; Save arg1
    B8 18 00 00 00        mov eax, 0x18         ; ← We want to capture this value
    0F 05                 syscall               ; ← Set breakpoint HERE
    C3                    ret

Strategy:

Set hardware breakpoint (DR0) at the syscall instruction address
Call into the NT stub
When execution hits the breakpoint, our VEH handler fires
Read EAX from the exception context — it contains the SSN
Clear the breakpoint and skip the syscall to avoid actual kernel entry

CPU Debug Registers

Register Layout (x64)

Register	Purpose	Size	Usage
DR0-DR3	Breakpoint addresses	64-bit	Hold linear addresses to break on
DR4-DR5	Reserved	—	Aliased to DR6/DR7 on old CPUs
DR6	Debug status	64-bit	Flags indicating which breakpoint fired
DR7	Debug control	64-bit	Enable/disable DR0-DR3, set conditions

DR7 Control Register

DR7 bit layout (simplified):
Bit  0:     L0 — DR0 local enable (current task)
Bit  1:     G0 — DR0 global enable (all tasks)
Bit  2-3:   L1/G1 (DR1)
Bit  4-5:   L2/G2 (DR2)
Bit  6-7:   L3/G3 (DR3)
Bit 16-17:  R/W0 — DR0 condition (00=exec, 01=write, 11=read/write)
Bit 18-19:  LEN0 — DR0 length (00=1 byte, 01=2 bytes, 11=4/8 bytes)
Bit 20-31:  Similar for DR1-DR3

Setting an Execution Breakpoint

// Set DR0 to break on execution at 'address'
CONTEXT ctx;
ctx.ContextFlags = CONTEXT_DEBUG_REGISTERS;
GetThreadContext(GetCurrentThread(), &ctx);

ctx.Dr0 = (DWORD64)address;        // Breakpoint address
ctx.Dr6 = 0;                       // Clear status
ctx.Dr7 = (1 << 0);                // Enable DR0 local
ctx.Dr7 |= (0 << 16);              // R/W0 = 00 (execute)
ctx.Dr7 |= (0 << 18);              // LEN0 = 00 (1 byte)

SetThreadContext(GetCurrentThread(), &ctx);

Vectored Exception Handler (VEH)

What is VEH?

VEH allows user-mode applications to register handlers for hardware exceptions before structured exception handling (SEH) runs. Critical for catching debug register breakpoints.

Registration

PVOID vehHandle = AddVectoredExceptionHandler(
    1,  // Call this handler first
    SW4_DebugExceptionHandler
);

Handler Structure

LONG WINAPI SW4_DebugExceptionHandler(PEXCEPTION_POINTERS ExceptionInfo) {
    if (ExceptionInfo->ExceptionRecord->ExceptionCode == EXCEPTION_SINGLE_STEP) {
        // Debug register breakpoint fired
        PCONTEXT ctx = ExceptionInfo->ContextRecord;
        
        // Check which breakpoint hit
        if (ctx->Dr6 & (1 << 0)) {  // DR0 fired
            // DR0 was set at the syscall instruction
            // EAX contains the SSN!
            DWORD ssn = (DWORD)ctx->Rax;  // x64: RAX lower 32 bits
            
            // Store SSN in global table
            g_CurrentSsn = ssn;
            
            // Skip the syscall instruction (avoid kernel entry)
            ctx->Rip += 2;  // syscall is 2 bytes: 0F 05
            
            // Clear DR0 and status
            ctx->Dr0 = 0;
            ctx->Dr6 = 0;
            ctx->Dr7 = 0;
            
            return EXCEPTION_CONTINUE_EXECUTION;
        }
    }
    return EXCEPTION_CONTINUE_SEARCH;
}

Full Implementation

Initialization

typedef struct _SW4_HW_BP_CONTEXT {
    PVOID   SyscallAddress;  // Address of syscall instruction
    DWORD   FunctionHash;    // DJB2 hash of function name
    DWORD   CapturedSsn;     // SSN captured from EAX
    BOOL    Complete;        // Capture successful
} SW4_HW_BP_CONTEXT;

static SW4_HW_BP_CONTEXT g_BpContext;

LONG WINAPI SW4_VehHandler(PEXCEPTION_POINTERS ExceptionInfo) {
    if (ExceptionInfo->ExceptionRecord->ExceptionCode == EXCEPTION_SINGLE_STEP) {
        PCONTEXT ctx = ExceptionInfo->ContextRecord;
        
        if (ctx->Dr6 & (1 << 0)) {  // DR0 fired
            // Capture SSN from RAX
            g_BpContext.CapturedSsn = (DWORD)(ctx->Rax & 0xFFFFFFFF);
            g_BpContext.Complete = TRUE;
            
            // Skip syscall (2 bytes: 0F 05)
            ctx->Rip += 2;
            
            // Disable breakpoint
            ctx->Dr0 = 0;
            ctx->Dr6 = 0;
            ctx->Dr7 = 0;
            
            return EXCEPTION_CONTINUE_EXECUTION;
        }
    }
    return EXCEPTION_CONTINUE_SEARCH;
}

BOOL SW4_HardwareBreakpoint(PVOID pNtdll) {
    PVOID vehHandle = AddVectoredExceptionHandler(1, SW4_VehHandler);
    if (!vehHandle) return FALSE;

    // Parse ntdll exports
    PIMAGE_DOS_HEADER dos = (PIMAGE_DOS_HEADER)pNtdll;
    PIMAGE_NT_HEADERS nt = (PIMAGE_NT_HEADERS)((PBYTE)pNtdll + dos->e_lfanew);
    PIMAGE_EXPORT_DIRECTORY exports = 
        (PIMAGE_EXPORT_DIRECTORY)((PBYTE)pNtdll + 
        nt->OptionalHeader.DataDirectory[IMAGE_DIRECTORY_ENTRY_EXPORT].VirtualAddress);

    PDWORD nameRvas = (PDWORD)((PBYTE)pNtdll + exports->AddressOfNames);
    PDWORD funcRvas = (PDWORD)((PBYTE)pNtdll + exports->AddressOfFunctions);
    PWORD ordinals = (PWORD)((PBYTE)pNtdll + exports->AddressOfNameOrdinals);

    // For each NT function we want
    for (DWORD i = 0; i < exports->NumberOfNames; i++) {
        PCHAR name = (PCHAR)((PBYTE)pNtdll + nameRvas[i]);
        if (name[0] != 'N' || name[1] != 't') continue;

        DWORD hash = djb2_hash(name);
        
        // Check if this is one of our target functions
        DWORD funcIdx = 0xFFFFFFFF;
        for (DWORD j = 0; j < SW4_FUNC_COUNT; j++) {
            if (SW4_FunctionHashes[j] == hash) {
                funcIdx = j;
                break;
            }
        }
        if (funcIdx == 0xFFFFFFFF) continue;

        // Get function address
        PVOID funcAddr = (PBYTE)pNtdll + funcRvas[ordinals[i]];
        PBYTE code = (PBYTE)funcAddr;

        // Locate syscall instruction (scan for 0F 05)
        PVOID syscallAddr = NULL;
        for (int offset = 0; offset < 32; offset++) {
            if (code[offset] == 0x0F && code[offset + 1] == 0x05) {
                syscallAddr = &code[offset];
                break;
            }
        }
        if (!syscallAddr) continue;

        // Set up breakpoint context
        g_BpContext.SyscallAddress = syscallAddr;
        g_BpContext.FunctionHash = hash;
        g_BpContext.Complete = FALSE;

        // Set DR0 to syscall address
        CONTEXT ctx;
        ctx.ContextFlags = CONTEXT_DEBUG_REGISTERS;
        GetThreadContext(GetCurrentThread(), &ctx);
        ctx.Dr0 = (DWORD64)syscallAddr;
        ctx.Dr6 = 0;
        ctx.Dr7 = (1 << 0);  // Enable DR0 local, execute breakpoint
        SetThreadContext(GetCurrentThread(), &ctx);

        // Call the NT function with dummy arguments
        // The VEH handler will catch it and extract SSN
        typedef NTSTATUS(NTAPI* NT_FUNC)(...);
        NT_FUNC pFunc = (NT_FUNC)funcAddr;
        
        __try {
            pFunc(NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL);
        } __except (EXCEPTION_EXECUTE_HANDLER) {
            // Ignore — we're just triggering the breakpoint
        }

        // Check if capture succeeded
        if (g_BpContext.Complete) {
            SW4_SsnTable[funcIdx] = g_BpContext.CapturedSsn;
        }

        // Clear context for next function
        memset(&g_BpContext, 0, sizeof(g_BpContext));
    }

    RemoveVectoredExceptionHandler(vehHandle);
    return TRUE;
}

Advantages

No Opcode Reading

Never inspects potentially-hooked function bytes — SSN comes from CPU register

Runtime Capture

Extracts actual SSN during execution — guaranteed to match kernel expectations

Hook Proof

Works even if hooks redirect execution — breakpoint fires after SSN is loaded

Educational Value

Demonstrates advanced Windows internals (debug registers, VEH, context manipulation)

Limitations

1. Performance Overhead

Cost: ~20-30ms initialization (vs. ~2ms for FreshyCalls)

VEH registration/removal
Setting debug registers per function (~64 times)
Exception handling overhead
Context switching

2. Anti-Debug Detection

Using DR0-DR3 may trigger anti-debug checks by EDRs:

// EDR may periodically check:
CONTEXT ctx;
ctx.ContextFlags = CONTEXT_DEBUG_REGISTERS;
GetThreadContext(hThread, &ctx);
if (ctx.Dr0 || ctx.Dr1 || ctx.Dr2 || ctx.Dr3) {
    // Debug registers in use — possible debugger
}

Mitigation: Clear DR7 immediately after each capture.

3. Instrumentation Callbacks

Some EDRs use NtSetInformationProcess(ProcessInstrumentationCallback) to detect debug register manipulation:

// Kernel callback fires when DR registers change
typedef NTSTATUS (*INSTRUMENTATION_CALLBACK)(
    PCONTEXT Context, PVOID Reserved
);

Detection: EDR receives notification when SetThreadContext modifies DR0-DR7.

4. Complexity

Highest complexity of all SSN resolution methods:

VEH management
Debug register programming
Exception handling
Edge case handling (hooked VEH APIs, thread state issues)

When to Use Hardware Breakpoints

Use When

Research/PoC demonstrating advanced techniques
Maximum paranoia + willingness to accept performance cost
Exotic EDR that defeats all other methods (extremely rare)
Educational purposes — learning Windows internals

Avoid When

Performance matters — use FreshyCalls or RecycledGate
Production operations — complexity increases failure risk
Anti-debug present — DR register usage is a detection vector
Simpler methods work — don’t over-engineer

Comparison with Other Methods

Feature	FreshyCalls	RecycledGate	SyscallsFromDisk	HW Breakpoint
Hook resistance	Very High	Maximum	Maximum	Maximum
Speed	Fast (~2ms)	Fast (~5ms)	Slow (~15ms)	Slowest (~25ms)
Complexity	Low	Medium	Medium	Very High
Anti-debug risk	❌	❌	❌	⚠️ High
Opcode dependency	❌	Partial	✅	❌
Runtime capture	❌	❌	❌	✅

Usage in SysWhispers4

Generate with Hardware Breakpoints

# Basic usage
python syswhispers.py --preset injection --resolve hw_breakpoint

# Recommended: limit function count for performance
python syswhispers.py \
    --functions NtAllocateVirtualMemory,NtWriteVirtualMemory,NtCreateThreadEx \
    --resolve hw_breakpoint \
    --method indirect

Integration Example

#include "SW4Syscalls.h"

int main(void) {
    printf("[*] Initializing SSN resolution via hardware breakpoints...\n");

    // Initialize — this will:
    // 1. Register VEH handler
    // 2. Set DR0 breakpoints on each NT function's syscall instruction
    // 3. Trigger execution to capture SSNs from EAX
    // 4. Clean up debug registers and VEH
    if (!SW4_Initialize()) {
        fprintf(stderr, "[!] Hardware breakpoint SSN extraction failed\n");
        return 1;
    }

    printf("[+] SSNs captured via debug registers\n");

    // Optional: verify no debugger attached (DR usage may trigger EDR alerts)
    if (!SW4_AntiDebugCheck()) {
        fprintf(stderr, "[!] Debugger detected\n");
        return 0;
    }

    // Use syscalls normally
    PVOID base = NULL;
    SIZE_T size = 0x1000;
    NTSTATUS st = SW4_NtAllocateVirtualMemory(
        GetCurrentProcess(), &base, 0, &size,
        MEM_COMMIT | MEM_RESERVE, PAGE_READWRITE
    );

    printf("[+] Allocated memory at 0x%p\n", base);
    return NT_SUCCESS(st) ? 0 : 1;
}

Detection & Evasion

Observable Behaviors

Action	EDR Visibility	Kernel Visibility
`AddVectoredExceptionHandler`	✅ (user32.dll hook)	❌
`SetThreadContext` (DR writes)	✅ (ntdll hook)	✅ (instrumentation)
Debug register usage	⚠️ (polling)	✅ (via callbacks)
VEH handler execution	❌ (in-process)	❌

Mitigation Strategies

Clear debug registers immediately

Don’t leave DR0-DR3 set after capture:

ctx.Dr0 = ctx.Dr1 = ctx.Dr2 = ctx.Dr3 = 0;
ctx.Dr6 = ctx.Dr7 = 0;
SetThreadContext(GetCurrentThread(), &ctx);

Use indirect syscall invocation

Keep RIP inside ntdll:

python syswhispers.py --resolve hw_breakpoint --method indirect

Combine with anti-instrumentation checks

Detect if ProcessInstrumentationCallback is set:

ULONG_PTR callback = 0;
NtQueryInformationProcess(
    GetCurrentProcess(),
    ProcessInstrumentationCallback,  // 40
    &callback,
    sizeof(callback),
    NULL
);
if (callback != 0) {
    // EDR instrumentation active
}

Technical Deep Dive: Why It Works

Execution Flow

1. User code calls NtAllocateVirtualMemory(...)
   ↓
2. CPU jumps to ntdll stub:
   4C 8B D1              mov r10, rcx        ; arg1 → r10
   B8 18 00 00 00        mov eax, 0x18       ; SSN → EAX
   0F 05 ← DR0 set here  syscall
   ↓
3. CPU executes mov eax, 0x18 (SSN now in EAX)
   ↓
4. CPU reaches 0F 05, DR0 fires → EXCEPTION_SINGLE_STEP
   ↓
5. Windows delivers exception to VEH handlers
   ↓
6. Our SW4_VehHandler runs:
   - Reads EAX from CONTEXT (0x18)
   - Stores SSN in table
   - ctx.Rip += 2 (skip syscall)
   - Clears DR0
   - Returns EXCEPTION_CONTINUE_EXECUTION
   ↓
7. Execution resumes after the syscall instruction
   ↓
8. Stub returns to user code (no actual syscall occurred)

Why Hooks Don’t Matter

Even if an EDR hooks the first bytes with a JMP:

; Hooked stub:
E9 XX XX XX XX        jmp <EDR_Handler>
...

The EDR handler must eventually execute the real syscall, which means:

Loading the SSN into EAX
Executing the syscall instruction

Our hardware breakpoint catches step 2, after step 1 completes — we still capture the SSN.

LayeredSyscall Research

White Knight Labs on VEH abuse

Intel SDM: Debug Registers

Official documentation (Vol. 3, Chapter 17)

RecycledGate

Simpler alternative with excellent hook resistance

FreshyCalls

Fast default method for most use cases

Techniques Deep Dive

Detection & Evasion

Development

Documentation Index

​Overview

​The Technique: Breakpoint-on-Syscall

​Core Concept

​CPU Debug Registers

​Register Layout (x64)

​DR7 Control Register

​Setting an Execution Breakpoint

​Vectored Exception Handler (VEH)

​What is VEH?

​Registration

​Handler Structure

​Full Implementation

​Initialization

​Advantages

No Opcode Reading

Runtime Capture

Hook Proof

Educational Value

​Limitations

​1. Performance Overhead

​2. Anti-Debug Detection

​3. Instrumentation Callbacks

​4. Complexity

​When to Use Hardware Breakpoints

​Comparison with Other Methods

​Usage in SysWhispers4

​Generate with Hardware Breakpoints

​Integration Example

​Detection & Evasion

​Observable Behaviors

​Mitigation Strategies

​Technical Deep Dive: Why It Works

​Execution Flow

​Why Hooks Don’t Matter

​Further Reading

LayeredSyscall Research

Intel SDM: Debug Registers

RecycledGate

FreshyCalls

Build docs developers (and LLMs) love

Overview

The Technique: Breakpoint-on-Syscall

Core Concept

CPU Debug Registers

Register Layout (x64)

DR7 Control Register

Setting an Execution Breakpoint

Vectored Exception Handler (VEH)

What is VEH?

Registration

Handler Structure

Full Implementation

Initialization

Advantages

Limitations

1. Performance Overhead

2. Anti-Debug Detection

3. Instrumentation Callbacks

4. Complexity

When to Use Hardware Breakpoints

Comparison with Other Methods

Usage in SysWhispers4

Generate with Hardware Breakpoints

Integration Example

Detection & Evasion

Observable Behaviors

Mitigation Strategies

Technical Deep Dive: Why It Works

Execution Flow

Why Hooks Don’t Matter

Further Reading