llvm-security（LLVM 安全）

LLVM Security Skill

This skill covers LLVM-based security features, sanitizers, hardening mechanisms, and secure software development practices.

Sanitizers

AddressSanitizer (ASan)

Detects memory errors: buffer overflow, use-after-free, use-after-scope.

# Compile with ASan
clang -fsanitize=address -g program.c -o program

# Key features
# - Stack buffer overflow detection
# - Heap buffer overflow detection
# - Use-after-free detection
# - Memory leak detection

MemorySanitizer (MSan)

Detects uninitialized memory reads.

clang -fsanitize=memory -g program.c -o program

ThreadSanitizer (TSan)

Detects data races in multithreaded programs.

clang -fsanitize=thread -g program.c -o program

UndefinedBehaviorSanitizer (UBSan)

Detects undefined behavior at runtime.

clang -fsanitize=undefined -g program.c -o program

# Specific checks
clang -fsanitize=signed-integer-overflow,null program.c

Custom Sanitizer Development

// Implementing custom memory tracking
extern "C" void __asan_poison_memory_region(void const volatile *addr, size_t size);
extern "C" void __asan_unpoison_memory_region(void const volatile *addr, size_t size);

class SecureAllocator {
public:
    void* allocate(size_t size) {
        // Add red zones around allocation
        void* ptr = malloc(size + 2 * REDZONE_SIZE);
        __asan_poison_memory_region(ptr, REDZONE_SIZE);
        __asan_poison_memory_region((char*)ptr + REDZONE_SIZE + size, REDZONE_SIZE);
        return (char*)ptr + REDZONE_SIZE;
    }
};

Hardening Techniques

Stack Protection

# Stack canaries
clang -fstack-protector-strong program.c

# Stack clash protection
clang -fstack-clash-protection program.c

# Safe stack (separate stacks for safe/unsafe data)
clang -fsanitize=safe-stack program.c

Control Flow Integrity (CFI)

# Forward-edge CFI
clang -fsanitize=cfi -flto program.c

# Specific CFI schemes
clang -fsanitize=cfi-vcall      # Virtual call checks
clang -fsanitize=cfi-nvcall     # Non-virtual member call checks
clang -fsanitize=cfi-icall      # Indirect call checks

Shadow Call Stack

# Backward-edge protection (return address protection)
clang -fsanitize=shadow-call-stack program.c

Position Independent Executables

# Full ASLR support
clang -fPIE -pie program.c

# Position independent code for shared libraries
clang -fPIC -shared library.c -o library.so

Symbolic Execution

Integration with KLEE

// Mark symbolic inputs
#include <klee/klee.h>

int main() {
    int input;
    klee_make_symbolic(&input, sizeof(input), "input");

    if (input > 0) {
        // Path 1
    } else {
        // Path 2
    }
    return 0;
}

SymCC (Symbolic Execution via Compilation)

Compile-time instrumentation for symbolic execution:

• Faster than IR interpretation
• Supports complex real-world programs
• Integrates with fuzzing workflows

Symbolic Analysis Tools

• Caffeine: LLVM-based symbolic executor
• SymSan: Symbolic execution + sanitizers
• Haybale: Rust-based LLVM symbolic executor

Security-Focused Analysis

Type Checking at Runtime

// LLVM TypeSanitizer concepts
// Track type information through allocations
struct TypeInfo {
    const char* typeName;
    size_t typeSize;
    uint64_t typeHash;
};

void checkType(void* ptr, TypeInfo expected) {
    TypeInfo* actual = getTypeInfo(ptr);
    if (actual->typeHash != expected.typeHash) {
        reportTypeMismatch(ptr, actual, expected);
    }
}

Memory Leak Detection

// LeakSanitizer integration
extern "C" void __lsan_do_leak_check();
extern "C" void __lsan_disable();
extern "C" void __lsan_enable();

// Custom leak tracking
class PreciseLeakSanitizer {
    std::unordered_map<void*, AllocationInfo> allocations;

public:
    void recordAlloc(void* ptr, size_t size, const char* file, int line) {
        allocations[ptr] = {size, file, line, getStackTrace()};
    }

    void recordFree(void* ptr) {
        allocations.erase(ptr);
    }

    void reportLeaks() {
        for (auto& [ptr, info] : allocations) {
            fprintf(stderr, "Leak: %zu bytes at %s:%d\n",
                    info.size, info.file, info.line);
        }
    }
};

Exploit Mitigation Implementation

Return Address Protection

; Shadow stack concept in LLVM IR
define void @protected_function() {
entry:
    %return_addr = call ptr @llvm.returnaddress(i32 0)
    call void @shadow_stack_push(ptr %return_addr)

    ; Function body...

    %saved_addr = call ptr @shadow_stack_pop()
    %current_addr = call ptr @llvm.returnaddress(i32 0)
    %match = icmp eq ptr %saved_addr, %current_addr
    br i1 %match, label %safe_return, label %attack_detected

safe_return:
    ret void

attack_detected:
    call void @abort()
    unreachable
}

Pointer Authentication (ARM)

// Using pointer authentication on ARM64
__attribute__((target("sign-return-address")))
void signed_function() {
    // Return address is cryptographically signed
}

Secure Compilation Pipeline

Build Flags Checklist

# Comprehensive hardening
CFLAGS="-O2 \
    -fstack-protector-strong \
    -fstack-clash-protection \
    -fcf-protection=full \
    -fPIE \
    -D_FORTIFY_SOURCE=2 \
    -Wformat -Wformat-security \
    -fsanitize=cfi -flto"

LDFLAGS="-pie \
    -Wl,-z,relro \
    -Wl,-z,now \
    -Wl,-z,noexecstack"

Compiler Security Checks

• -Wformat-security: Format string vulnerabilities
• -Warray-bounds: Array bounds violations
• -Wshift-overflow: Shift operation overflows
• -Wnull-dereference: Null pointer dereferences

Fuzzing Integration

libFuzzer

// Fuzz target template
extern "C" int LLVMFuzzerTestOneInput(const uint8_t *Data, size_t Size) {
    // Parse/process Data
    processInput(Data, Size);
    return 0;
}

Sanitizer + Fuzzer Combination

# Comprehensive fuzzing setup
clang -fsanitize=fuzzer,address,undefined \
      -fno-omit-frame-pointer \
      -g fuzz_target.c -o fuzzer

Windows-Specific Security

Control Flow Guard (CFG)

clang-cl /guard:cf program.c

SEH (Structured Exception Handling)

• LLVM supports Windows SEH
• Use for secure exception handling
• Integrate with security monitoring

Resources

See Security Features, Sanitizer, and Symbolic Execution sections in README.md for comprehensive tool listings.

Getting Detailed Information

When you need detailed and up-to-date resource links, tool lists, or project references, fetch the latest data from:

https://raw.githubusercontent.com/gmh5225/awesome-llvm-security/refs/heads/main/README.md

This README contains comprehensive curated lists of:

• Security features and hardening (Security Features section)
• Sanitizers and memory safety tools (Sanitizer section)
• Symbolic execution frameworks (Symbolic Execution section)
• Memory leak detectors and runtime checkers