实现一个压缩壳，并给它加点“料”

语言: CN / TW / HK
时间 2022-08-25 18:59:26
主题:

本文为看雪论坛优秀文章

看雪论坛作者ID：橘喵Cat

一

前言

实现压缩壳，必须对PE格式十分熟悉，其次，解压缩代码需要编写shellcode，也是十分麻烦的环节。有了两者的结合，我们才能写好一个真正的压缩壳。

二

设计思路

首先上一张图，让大家直观地感受到一个壳程序是如何运行起来的。

左边是壳PE，壳程序有一个PE头，节表1是空节，用来存放解压缩后的原程序PE，节表2此时存储的是压缩后的原PE。节表3则是壳代码节，壳PE运行起来后，首先就是进入入口点，运行节表3的代码，解压缩节表2，然后将结果覆盖PE头＋节表1的位置，修复完导入表、重定位表，jmp到原程序的入口点处即可。

 原理不变，我这里加了点“料”，新增了节4和节5，存储了相关的信息，让压缩壳的脱壳过程变难，往后看就知道了。

三

壳代码实现

1.为了生成一个新的壳PE，我们一步步来，首先是PE头，俗话说靠山吃山，靠水吃水，这个PE头我就直接拿原程序的PE头来代替了，只不过需要改一些数据：

NumberOfSections --节表数量，要改为5
AddressOfEntryPoint --入口点，要改为节表3里的代码入口
SizeOfImage --PE在内存中的大小，要改为新的PE的内存大小
pSecHdr --节表头，要拓展为5个节表
 
void CPacker::GetNewPeHdr()
{
    //拷贝原PE的PE头
    m_dwNewPeHdrSize = m_pNtHdr->OptionalHeader.SizeOfHeaders;
    m_pNewPeHdr = new BYTE[m_dwNewPeHdrSize];
    CopyMemory(m_pNewPeHdr, m_pDosHdr, m_dwNewPeHdrSize);
 
    //修改
    auto pDosHdr = (PIMAGE_DOS_HEADER)m_pNewPeHdr;
    auto pNtHdr = (PIMAGE_NT_HEADERS)(m_pNewPeHdr + pDosHdr->e_lfanew);
    auto pSecHdr = (PIMAGE_SECTION_HEADER)
        ((LPBYTE)&pNtHdr->OptionalHeader + pNtHdr->FileHeader.SizeOfOptionalHeader);
 
    pNtHdr->FileHeader.NumberOfSections = 5;
    pNtHdr->OptionalHeader.AddressOfEntryPoint = m_newSecHdr[2].VirtualAddress;
    pNtHdr->OptionalHeader.SizeOfImage = m_newSecHdr[4].VirtualAddress + m_newSecHdr[4].Misc.VirtualSize;
    //清空DataDirectory目录
    ZeroMemory(pNtHdr->OptionalHeader.DataDirectory, sizeof(pNtHdr->OptionalHeader.DataDirectory));
 
    //修改新的节表头
    CopyMemory(pSecHdr, m_newSecHdr, sizeof(m_newSecHdr));
 
}

2.新PE头里一些需要修改的数据，比如SizeOfImage，我们目前还没有，需要等我们构造出节表1、2、3、4、5之后，才知道。接下来，先构造节表1的表头，节表1是个空节，它的大小只要够存放原程序的节表即可，多给一点也没关系，我这里直接给了SizeOfImage。（由于是空节，所以这里并不需要考虑节表1的数据内容）

//空节
    strcpy((char*)m_newSecHdr[0].Name, ".cr42");
    m_newSecHdr[0].Misc.VirtualSize = m_pNtHdr->OptionalHeader.SizeOfImage;
    m_newSecHdr[0].VirtualAddress = m_pSecHdr[0].VirtualAddress;
    m_newSecHdr[0].SizeOfRawData = 0;
    m_newSecHdr[0].PointerToRawData = 0;
    m_newSecHdr[0].Characteristics =
        IMAGE_SCN_MEM_EXECUTE | IMAGE_SCN_MEM_WRITE | IMAGE_SCN_MEM_READ;

3.接着是节表2，这个节要存放原PE的压缩数据，先设计表头（使用了PointerToRelocations和PointerToLinenumbers这两个没啥用的字段，存放压缩大小信息，留着给后面shellcode用）。

//压缩数据节
    strcpy((char*)m_newSecHdr[1].Name, ".data");
    m_newSecHdr[1].Misc.VirtualSize = GetAlign(m_dwComSecSize, m_pNtHdr->OptionalHeader.SectionAlignment);
    m_newSecHdr[1].VirtualAddress = m_newSecHdr[0].VirtualAddress + m_newSecHdr[0].Misc.VirtualSize;
    m_newSecHdr[1].SizeOfRawData = m_dwComSecSize;
    m_newSecHdr[1].PointerToRawData = m_pNtHdr->OptionalHeader.SizeOfHeaders;
    m_newSecHdr[1].Characteristics = IMAGE_SCN_MEM_READ;
 
    m_newSecHdr[1].PointerToRelocations = m_dwComSize; //压缩后大小
    m_newSecHdr[1].PointerToLinenumbers = m_dwSrcPeSize;//压缩前大小

然后将PE压缩，压缩前我把节表、导入表、重定位表保存并清空，到时候由shellcode进行还原。

bool CPacker::GetCompressData()
{
    COMPRESSOR_HANDLE hCompressor = NULL;
    BOOL Success = CreateCompressor(
        COMPRESS_ALGORITHM_XPRESS_HUFF,
        NULL,
        &hCompressor
    );
 
    m_pComData = new BYTE[m_dwSrcPeSize + 0x28];
    LPBYTE m_pSrcPeTmp = new BYTE[m_dwSrcPeSize];
    CopyMemory(m_pSrcPeTmp, m_pSrcPe, m_dwSrcPeSize);
 
    PIMAGE_DOS_HEADER m_pDosHdrTmp = (PIMAGE_DOS_HEADER)m_pSrcPeTmp;
    PIMAGE_NT_HEADERS m_pNtHdrTmp = (PIMAGE_NT_HEADERS)(m_pSrcPeTmp + m_pDosHdrTmp->e_lfanew);
    PIMAGE_SECTION_HEADER m_pSecHdrTmp = (PIMAGE_SECTION_HEADER)
        ((LPBYTE)&m_pNtHdrTmp->OptionalHeader + m_pNtHdrTmp->FileHeader.SizeOfOptionalHeader);
 
    //1.在压缩前，把节表保存并清空
    nSecNum = m_pNtHdrTmp->FileHeader.NumberOfSections;
    CopyMemory(m_pSaveSecHdr, m_pSecHdrTmp, nSecNum * 40);
    ZeroMemory(m_pSecHdrTmp, nSecNum * 40);
 
    //2.把导入表保存并清空
    m_pImportAddr = m_pNtHdrTmp->OptionalHeader.DataDirectory[1].VirtualAddress;
    nImpSize = m_pNtHdrTmp->OptionalHeader.DataDirectory[1].Size;
    ZeroMemory(&(m_pNtHdrTmp->OptionalHeader.DataDirectory[1]), 8);
 
    //3.把重定位表保存并清空
    m_pRelocAddr = m_pNtHdrTmp->OptionalHeader.DataDirectory[5].VirtualAddress;
    nRelocSize = m_pNtHdrTmp->OptionalHeader.DataDirectory[5].Size;
    ZeroMemory(&(m_pNtHdrTmp->OptionalHeader.DataDirectory[5]), 8);
 
 
 
    Success = Compress(
        hCompressor,
        m_pSrcPeTmp,
        m_dwSrcPeSize,
        m_pComData,
        m_dwSrcPeSize + 0x28,
        &m_dwComSize
    );
 
 
    return true;
}

4.接着是节表3，该节表存放的是解压缩PE、还原导入表、重定位表，运行原程序的至关重要的shellcode，先设计表头。

//代码节
    strcpy((char*)m_newSecHdr[2].Name, ".text");
    m_newSecHdr[2].Misc.VirtualSize = GetAlign(m_dwCodeSecSize, m_pNtHdr->OptionalHeader.SectionAlignment);
    m_newSecHdr[2].VirtualAddress = m_newSecHdr[1].VirtualAddress + m_newSecHdr[1].Misc.VirtualSize;
    m_newSecHdr[2].SizeOfRawData = m_dwCodeSecSize;
    m_newSecHdr[2].PointerToRawData = m_newSecHdr[1].PointerToRawData + m_newSecHdr[1].SizeOfRawData;
    m_newSecHdr[2].Characteristics = IMAGE_SCN_MEM_EXECUTE | IMAGE_SCN_MEM_WRITE | IMAGE_SCN_MEM_READ;

至于节表3的数据内容，也就是shellcode，等我把节表、PE设计完，我再说。


5.接着是节表4，节表4是我额外增加的一个，用来存储步骤3中保存的原节表表头、原导入表、原重定位表，首先设计节表4表头。 

//存放原节表、导入表、重定位表的节
    strcpy((char*)m_newSecHdr[3].Name, ".info");
    m_newSecHdr[3].Misc.VirtualSize = GetAlign(m_dwTableSecSize, m_pNtHdr->OptionalHeader.SectionAlignment);
    m_newSecHdr[3].VirtualAddress = m_newSecHdr[2].VirtualAddress + m_newSecHdr[2].Misc.VirtualSize;
    m_newSecHdr[3].SizeOfRawData = m_dwTableSecSize;
    m_newSecHdr[3].PointerToRawData = m_newSecHdr[2].PointerToRawData + m_newSecHdr[2].SizeOfRawData;
    m_newSecHdr[3].Characteristics = IMAGE_SCN_MEM_EXECUTE | IMAGE_SCN_MEM_WRITE | IMAGE_SCN_MEM_READ;

然后把刚才保存的原节表表头、原导入表、原重定位表信息，按顺序写入到缓冲区里。

bool CPacker::GetTable()
{
    m_dwTableSize = nSecNum * 40 + 4 + 8 + 8;
    m_pTable = new BYTE[m_dwTableSize];
    RtlCopyMemory(m_pTable, &nSecNum, 4);
    RtlCopyMemory(m_pTable + 4, m_pSaveSecHdr, nSecNum * 40);
    RtlCopyMemory(m_pTable + 4 + nSecNum * 40, &m_pImportAddr, 4);
    RtlCopyMemory(m_pTable + 4 + nSecNum * 40 + 4, &nImpSize, 4);
    RtlCopyMemory(m_pTable + 4 + nSecNum * 40 + 4 + 4, &m_pRelocAddr, 4);
    RtlCopyMemory(m_pTable + 4 + nSecNum * 40 + 4 + 4 + 4, &nRelocSize, 4);
 
    return true;
}

6.还剩最后一个节表5，是一个空节，壳代码还原原PE时，要还原导入表，于是这个节的作用就体现出来了，shellcode在这里玩了一波偷梁换柱，直接毙掉了x64dbg的脱壳后导入表自动修复功能，等会介绍shellcode的时候你们就知道了。设计节表5的表头：

//绕过x64搜索导入表的节(空节)
    strcpy((char*)m_newSecHdr[4].Name, ".imp");
    m_newSecHdr[4].Misc.VirtualSize = GetAlign(0x10000, m_pNtHdr->OptionalHeader.SectionAlignment);
    m_newSecHdr[4].VirtualAddress = m_newSecHdr[3].VirtualAddress + m_newSecHdr[3].Misc.VirtualSize;
    m_newSecHdr[4].SizeOfRawData = 0;
    m_newSecHdr[4].PointerToRawData = m_newSecHdr[3].PointerToRawData + m_newSecHdr[2].SizeOfRawData;
    m_newSecHdr[4].Characteristics = IMAGE_SCN_MEM_EXECUTE | IMAGE_SCN_MEM_WRITE | IMAGE_SCN_MEM_READ;

至此，新的壳PE结构我们就设计好了，然后将新PE头、节表234写入到新文件（节表1、5是空节，不用写入），就大功告成，加壳完毕！

bool CPacker::WriteNewPe(CString strNewPe)
{
    //创建文件
    HANDLE hFile = CreateFile(strNewPe,
        GENERIC_WRITE,
        0,
        NULL,
        CREATE_ALWAYS,
        FILE_ATTRIBUTE_NORMAL,
        NULL);
 
    //写入PE头
    DWORD dwBytesWrited = 0;
    WriteFile(hFile, m_pNewPeHdr, m_dwNewPeHdrSize, &dwBytesWrited, NULL);
 
    //写入数据节
    WriteFile(hFile, m_pComSec, m_dwComSecSize, &dwBytesWrited, NULL);
 
    //写入代码节
    WriteFile(hFile, m_pCodeSec, m_dwCodeSecSize, &dwBytesWrited, NULL);
 
    //写入存放表数据节
    WriteFile(hFile, m_pTableSec, m_dwTableSecSize, &dwBytesWrited, NULL);
 
    CloseHandle(hFile);
 
    return true;
}

四

shellcode代码实现

由于壳PE可能是随机基址，所以执行shellcode时，一定要确保它的代码跟地址无关。我的shellcode是在VS上编译的，为了保证VS不会生成多余的代码，要修改以下几个设置：

1.使用Release版（Debug会加地址有关代码）
2.不使用main函数，自己定义一个，放在链接器->高级->入口点（main不是程序真正入口点）
3.关掉代码生成->安全检查
4.关掉增强指令集
5.关掉全程序优化

然后就是最麻烦的shellcode代码编写了。首先要清楚shellcode代码的功能：

1.解压缩节表2里面的压缩数据
2.将解压缩数据覆盖到PE头处，连带着空节表1也被覆盖
3.还原节表、导入表、重定位表

1.要解压缩，那么必定要调用库的API，如果直接调用的话，call的就是死地址，违背了地址无关原则。那么使用LoadLibrary然后GetProcAddress？显然也不行，LoadLibrary和GetProcAddress也是死地址，所以我们要自己实现LoadLibrary和GetProcAddress的功能。


首先通过_PEB来拿到kernel32的模块基址。 

HMODULE GetKernel32()
{
    HMODULE hKer;
    __asm {
 
        mov eax, dword ptr fs:[0x30]
        mov eax, dword ptr[eax + 0x0C]
        mov eax, dword ptr[eax + 0x0C]
        mov eax, dword ptr[eax]
        mov eax, dword ptr[eax]
        mov eax, dword ptr[eax + 0x18]
        mov hKer, eax
    }
 
    return hKer;
}

然后通过kernel32的导出函数表，拿到GetProcAddress函数地址，代码如下：

FARPROC MyGetProcAddress(HMODULE hMod, LPCSTR lpProcName) {
 
    IMAGE_DOS_HEADER* pDosHdr;
    IMAGE_NT_HEADERS* pNTHdr;
    IMAGE_EXPORT_DIRECTORY* pExpDir;
    DWORD pAddrTbl;
    DWORD pNameTbl;
    DWORD pOrdTbl;
 
    //解析dos头
    pDosHdr = (IMAGE_DOS_HEADER*)hMod;
 
    //nt头
    pNTHdr = (IMAGE_NT_HEADERS*)(pDosHdr->e_lfanew + (DWORD)hMod);
 
    //获取导出表
    pExpDir = (IMAGE_EXPORT_DIRECTORY*)(pNTHdr->OptionalHeader.DataDirectory[0].VirtualAddress + (DWORD)hMod);
 
    //导出函数地址表
    pAddrTbl = (DWORD)(pExpDir->AddressOfFunctions + (DWORD)hMod);
 
    //导出函数名称表
    pNameTbl = (DWORD)(pExpDir->AddressOfNames + (DWORD)hMod);
 
    //导出序号表
    pOrdTbl = (DWORD)(pExpDir->AddressOfNameOrdinals + (DWORD)hMod);
 
    //判断是序号还是名称
    if ((int)lpProcName & 0xffff0000) {
 
        //名称
        int i = 0;
        while (i < pExpDir->NumberOfNames) {
 
            //获取名称地址
            int nNameOff = (int)(*(DWORD*)(pNameTbl + i * 4) + (DWORD)hMod);
 
            //字符串比较
            if (((char*)nNameOff)[0] == 'G'&& ((char*)nNameOff)[1] == 'e'&& ((char*)nNameOff)[2] == 't'&&
                ((char*)nNameOff)[3] == 'P'&& ((char*)nNameOff)[4] == 'r'&& ((char*)nNameOff)[5] == 'o'&&
                ((char*)nNameOff)[6] == 'c'&& ((char*)nNameOff)[7] == 'A'&& ((char*)nNameOff)[8] == 'd'&&
                ((char*)nNameOff)[9] == 'd'&& ((char*)nNameOff)[10] == 'r'&& ((char*)nNameOff)[11] == 'e'&&
                ((char*)nNameOff)[12] == 's'&& ((char*)nNameOff)[13] == 's') {
 
                //找到了, 从导出序号表取出函数地址下标
                int nOrdinal = *(WORD*)(pOrdTbl + i * 2);
 
                //从导出地址表，下标寻址，获取导出函数地址
                int nFuncAddr = *(DWORD*)(pAddrTbl + nOrdinal * 4);
 
 
 
                //不是转发
                nFuncAddr += (int)hMod;
 
 
 
 
                //返回地址
                if (nFuncAddr != NULL) {
 
                    return (FARPROC)nFuncAddr;
                }
            }
 
            i++;
 
        }
    }
 
    else {
 
        //序号
        int nOrdinal = (DWORD)lpProcName - pExpDir->Base;
 
        //从导出地址表，下标寻址，获取导出函数地址
        int nFuncAddr = *(DWORD*)(pAddrTbl + nOrdinal * 4);
 
        //返回地址
        if (nFuncAddr != NULL) {
 
            return (FARPROC)(nFuncAddr + (DWORD)hMod);
        }
 
    }
 
 
    return 0;
}

有了GetProcAddress函数地址和kernel32的基址，同理就能拿到LoadLibrary函数地址了。（注意定义字符串变量时，用单个字符一个一个排列，在汇编里面看就是db出来的，否则字符串会有一个常量区地址，影响shellcode的通用性）

char szLoadLibrary[] = { 'L', 'o','a', 'd', 'L', 'i', 'b', 'r', 'a', 'r', 'y', 'A', '\0' };
    pEnv->pfnLoadLibraryA = (PFN_LoadLibraryA)pEnv->pfnGetProcAddress(hKer, szLoadLibrary);

2.有了LoadLibrary和GetProcAddress，就可以使用任何库函数了，解压缩便是小菜一碟。

//有了LoadLibrary和GetProcAddress，就可以使用任意函数了
    //获取解压缩相关函数
 
    char szCab[] = { 'C','a','b','i','n','e','t', '\0' };
    HMODULE hCab = pEnv->pfnLoadLibraryA(szCab);
 
    char szCreateDecompressor[] = { 'C','r','e','a','t','e','D','e','c','o','m','p','r','e','s','s','o','r', '\0' };
    pEnv->pfnCreateDecompressor = (PFN_CreateDecompressor)pEnv->pfnGetProcAddress(hCab, szCreateDecompressor);
 
    char szDecompress[] = { 'D','e','c','o','m','p','r','e','s','s', '\0' };
    pEnv->pfnDecompress = (PFN_Decompress)pEnv->pfnGetProcAddress(hCab, szDecompress);
 
    char szVirtualAlloc[] = { 'V','i','r','t','u','a','l','A','l','l','o','c', '\0' };
    pEnv->pfnVirtualAlloc = (PFN_VirtualAlloc)pEnv->pfnGetProcAddress(hKer, szVirtualAlloc);
 
    char szVirtualProtect[] = { 'V','i','r','t','u','a','l','P','r','o','t','e','c','t','\0' };
    pEnv->pfnVirtualProtect = (PFN_VirtualProtect)pEnv->pfnGetProcAddress(hKer, szVirtualProtect);
 
 
//解压缩
 
    LPBYTE pPEBuff = (LPBYTE)env.pfnVirtualAlloc(NULL, dwDeComSize, MEM_COMMIT, PAGE_READWRITE);
 
    DECOMPRESSOR_HANDLE hDecompressor;
    BOOL bSuccess = env.pfnCreateDecompressor(
        COMPRESS_ALGORITHM_XPRESS_HUFF,
        NULL,
        &hDecompressor
    );
 
    DWORD dwDecompressedBufferSize = 0;
    bSuccess = env.pfnDecompress(
        hDecompressor,
        pComData,
        dwComSize,
        pPEBuff,
        dwDeComSize,
        &dwDecompressedBufferSize
    );

3.解压缩完毕，得到了原PE，接下来就是将原PE覆盖到现在的PE头+节表1的地方，同时还原节表头、导入表和重定位表，相当于LoadPE的功能了。

#define VirtualProtect pEnv->pfnVirtualProtect
DWORD MyLoadLibrary(LPBYTE pPEBuff, Environment* pEnv, LPBYTE pTableBuf, LPBYTE pMyImpBuf) {
 
    DWORD dwImageBase;
    HANDLE hFile;
    HANDLE hFileMap;
    LPVOID pPEBuf;
    IMAGE_DOS_HEADER* pDosHdr;
    IMAGE_NT_HEADERS* pNTHdr;
    IMAGE_SECTION_HEADER* pSecHdr;
    DWORD dwNumOfSecs;
    IMAGE_IMPORT_DESCRIPTOR* pImpHdr;
    DWORD dwSizeOfHeaders;
    IMAGE_IMPORT_DESCRIPTOR hdrZeroImp;
    HMODULE hDll;
    DWORD dwOep;
    DWORD dwOldProc;
    IMAGE_BASE_RELOCATION* pReloc;
    DWORD dwOfReloc;
    DWORD dwOff;
 
    //RtlZeroMemory(&hdrZeroImp, sizeof(IMAGE_IMPORT_DESCRIPTOR));
 
    pPEBuf = pPEBuff;
 
    //解析
    //dos 头
    pDosHdr = (IMAGE_DOS_HEADER*)pPEBuf;
 
    //nt头
    pNTHdr = (IMAGE_NT_HEADERS*)(pDosHdr->e_lfanew + (DWORD)pPEBuf);
 
    //还原节表
    DWORD nSecNum = *(DWORD*)pTableBuf;
    mymemcpy((void*)((DWORD)&pNTHdr->OptionalHeader + pNTHdr->FileHeader.SizeOfOptionalHeader),
        pTableBuf + 4, nSecNum * 40);
 
    //还原导入表
    mymemcpy(&(pNTHdr->OptionalHeader.DataDirectory[1].VirtualAddress),
        pTableBuf + 4 + nSecNum * 40, 4);
    mymemcpy(&(pNTHdr->OptionalHeader.DataDirectory[1].Size),
        pTableBuf + 4 + nSecNum * 40 + 4, 4);
 
    //还原重定位表
    mymemcpy(&(pNTHdr->OptionalHeader.DataDirectory[5].VirtualAddress),
        pTableBuf + 4 + nSecNum * 40 + 4 + 4, 4);
    mymemcpy(&(pNTHdr->OptionalHeader.DataDirectory[5].Size),
        pTableBuf + 4 + nSecNum * 40 + 4 + 4 + 4, 4);
 
    //选项头信息
    dwSizeOfHeaders = pNTHdr->OptionalHeader.SizeOfHeaders;
 
 
    //自己的模块基址
    dwImageBase = (DWORD)GetModuleBase();
 
    dwOff = dwImageBase - pNTHdr->OptionalHeader.ImageBase; //新旧ImageBase的偏移差
 
    dwOep = pNTHdr->OptionalHeader.AddressOfEntryPoint + dwImageBase;
    //节表
    dwNumOfSecs = pNTHdr->FileHeader.NumberOfSections;
    pSecHdr = (IMAGE_SECTION_HEADER*)((DWORD)&pNTHdr->OptionalHeader + pNTHdr->FileHeader.SizeOfOptionalHeader);
 
 
 
 
    //拷贝PE头
    VirtualProtect((LPVOID)dwImageBase, pNTHdr->OptionalHeader.SizeOfHeaders, PAGE_EXECUTE_READWRITE, &dwOldProc);
    mymemcpy((void*)dwImageBase, pPEBuf, dwSizeOfHeaders);
    VirtualProtect((LPVOID)dwImageBase, pNTHdr->OptionalHeader.SizeOfHeaders, dwOldProc, &dwOldProc);
 
 
    //按照节表，拷贝节区数据
    int i = 0;
    IMAGE_SECTION_HEADER* dwSecTmp = pSecHdr;
    while (i < dwNumOfSecs) {
 
        //目标
        DWORD dwDstMem = dwImageBase;
        dwDstMem += dwSecTmp->VirtualAddress;
 
        //源
        DWORD dwSrcFile = (DWORD)pPEBuf + dwSecTmp->PointerToRawData;
 
        //拷贝
        VirtualProtect((LPVOID)dwDstMem, dwSecTmp->SizeOfRawData, PAGE_EXECUTE_READWRITE, &dwOldProc);
        mymemcpy((void*)dwDstMem, (void*)dwSrcFile, dwSecTmp->SizeOfRawData);
        VirtualProtect((LPVOID)dwImageBase, pNTHdr->OptionalHeader.SizeOfHeaders, dwOldProc, &dwOldProc);
 
        i++;
        dwSecTmp = (IMAGE_SECTION_HEADER*)((char*)dwSecTmp + sizeof(IMAGE_SECTION_HEADER));
    }
 
 
    //获取导入表
    if (pNTHdr->OptionalHeader.DataDirectory[IMAGE_DIRECTORY_ENTRY_IMPORT].VirtualAddress != 0) {
 
        pImpHdr = (IMAGE_IMPORT_DESCRIPTOR*)(dwImageBase + pNTHdr->OptionalHeader.DataDirectory[IMAGE_DIRECTORY_ENTRY_IMPORT].VirtualAddress);
 
 
        //处理导入表
        IMAGE_IMPORT_DESCRIPTOR* pImpHdrTmp = pImpHdr;
        DWORD dwNum = 0;
        int nImpNum = 0;
 
        while (true) {
 
            //判断结束，全0项结束
            if (memcmp(pImpHdrTmp, &hdrZeroImp, sizeof(IMAGE_IMPORT_DESCRIPTOR)) == 0) {
 
                break;
            }
 
            //判断字段, 为空则结束
            if (pImpHdrTmp->Name == NULL || pImpHdrTmp->FirstThunk == NULL) {
 
                break;
            }
 
            //加载dll
            hDll = pEnv->pfnLoadLibraryA((LPCSTR)(dwImageBase + pImpHdrTmp->Name));
 
            //获取导入地址表, IAT
            DWORD dwIAT = pImpHdrTmp->FirstThunk + dwImageBase;
            DWORD dwINT = dwIAT;
            //获取导入名称表, INT
            if (pImpHdrTmp->OriginalFirstThunk != NULL) {
 
                dwINT = pImpHdrTmp->OriginalFirstThunk + dwImageBase;
            }
 
            //遍历导入名称表
            while (*(DWORD*)(dwINT) != 0) {
 
                if ((*(DWORD*)pImpHdrTmp) >> 31) {
 
                    //序号导入, 获取序号
                    dwNum = *(DWORD*)pImpHdrTmp;
                    dwNum = (dwNum << 16) >> 16;
 
                }
                else {
 
                    //名称导入
                    dwNum = *(DWORD*)pImpHdrTmp;
                    dwNum += dwImageBase;
                    dwNum += 2;
 
                }
 
                //获取函数地址后，先不要把地址直接写入IAT
                //而是先将.imp节地址写入IAT(每16个字节写一次)
                //在.imp节里写代码指令push 函数地址 retn
                //如果函数名是_acmdln，那么这里就不要混淆导入表
                if (((LPCSTR)((*(DWORD*)(dwINT)) + dwImageBase + 2))[0] == '_' && ((LPCSTR)((*(DWORD*)(dwINT)) + dwImageBase + 2))[1] == 'a' &&
                    ((LPCSTR)((*(DWORD*)(dwINT)) + dwImageBase + 2))[2] == 'c' && ((LPCSTR)((*(DWORD*)(dwINT)) + dwImageBase + 2))[3] == 'm' &&
                    ((LPCSTR)((*(DWORD*)(dwINT)) + dwImageBase + 2))[4] == 'd' && ((LPCSTR)((*(DWORD*)(dwINT)) + dwImageBase + 2))[5] == 'l' &&
                    ((LPCSTR)((*(DWORD*)(dwINT)) + dwImageBase + 2))[6] == 'n') {
 
                    *(DWORD*)dwIAT = (DWORD)pEnv->pfnGetProcAddress(hDll, (LPCSTR)((*(DWORD*)(dwINT)) + dwImageBase + 2));
                }
                else {
 
                    DWORD dwMyAddr = (DWORD)pMyImpBuf + nImpNum * 0x10;
                    DWORD dwFuncAddr = (DWORD)pEnv->pfnGetProcAddress(hDll, (LPCSTR)((*(DWORD*)(dwINT)) + dwImageBase + 2));
                    *(DWORD*)dwIAT = dwMyAddr;
                    *(BYTE*)dwMyAddr = 0x68;    //push
                    *(DWORD*)(dwMyAddr + 1) = dwFuncAddr;  //真实函数地址
                    *(BYTE*)(dwMyAddr + 5) = 0xC3;  //retn
                }
 
 
                dwIAT += 4;
                dwINT += 4;
                nImpNum++;
 
            }
 
 
            pImpHdrTmp = (IMAGE_IMPORT_DESCRIPTOR*)((char*)pImpHdrTmp + sizeof(IMAGE_IMPORT_DESCRIPTOR));
        }
    }
 
    if (pNTHdr->OptionalHeader.DataDirectory[5].VirtualAddress != 0) {
 
        //定位重定位表
        pReloc = (IMAGE_BASE_RELOCATION*)(pNTHdr->OptionalHeader.DataDirectory[5].VirtualAddress + dwImageBase);
        dwOfReloc = pNTHdr->OptionalHeader.DataDirectory[5].Size;
 
 
        int nSize = 0;
 
        while (nSize < dwOfReloc) {
 
            //数组首地址
            int nOff = (DWORD)pReloc + 8;
 
            //数组元素个数
            int nCnt = (pReloc->SizeOfBlock - 8) >> 1;
 
            //遍历数组
            int j = 0;
            while (j < nCnt) {
 
                //取出一项
                int nDataOff = *(WORD*)(nOff + j * 2);
 
                //判断是否是有效重定位项
                if (nDataOff & 0x00003000) {
 
                    //修正
                    nDataOff = nDataOff & 0x0fff;  //页偏移
                    nDataOff = nDataOff + pReloc->VirtualAddress;
                    nDataOff = nDataOff + dwImageBase;
 
                    *(int*)nDataOff = *(int*)nDataOff + dwOff;
                }
 
                j++;
            }
 
            //处理下一个分页
            nSize += pReloc->SizeOfBlock;
            pReloc = (IMAGE_BASE_RELOCATION*)((char*)pReloc + pReloc->SizeOfBlock);
 
        }
    }
 
 
 
    return dwOep;
}

这里重点要介绍的，就是还原导入表的过程，也是整篇文章的核心主题，加点“料”。我在遍历还原导入表时，并没有直接将API的地址填入到IAT里，而是将节表5的地址，从起始位置开始，每隔16个字节，将地址填入到IAT里，然后在对应的节表5地址上填入push 真实函数地址 + retn的汇编指令。这样一来，原PE程序运行调用API时，就会跳到节表5里面，再从节表5里面跳到真实API地址，直接干掉了x64dbg的脱壳导入表自动修复功能。

//获取函数地址后，先不要把地址直接写入IAT
                //而是先将.imp节地址写入IAT(每16个字节写一次)
                //在.imp节里写代码指令push 函数地址 retn
                //如果函数名是_acmdln，那么这里就不要混淆导入表
                if (((LPCSTR)((*(DWORD*)(dwINT)) + dwImageBase + 2))[0] == '_' && ((LPCSTR)((*(DWORD*)(dwINT)) + dwImageBase + 2))[1] == 'a' &&
                    ((LPCSTR)((*(DWORD*)(dwINT)) + dwImageBase + 2))[2] == 'c' && ((LPCSTR)((*(DWORD*)(dwINT)) + dwImageBase + 2))[3] == 'm' &&
                    ((LPCSTR)((*(DWORD*)(dwINT)) + dwImageBase + 2))[4] == 'd' && ((LPCSTR)((*(DWORD*)(dwINT)) + dwImageBase + 2))[5] == 'l' &&
                    ((LPCSTR)((*(DWORD*)(dwINT)) + dwImageBase + 2))[6] == 'n') {
 
                    *(DWORD*)dwIAT = (DWORD)pEnv->pfnGetProcAddress(hDll, (LPCSTR)((*(DWORD*)(dwINT)) + dwImageBase + 2));
                }
                else {
 
                    DWORD dwMyAddr = (DWORD)pMyImpBuf + nImpNum * 0x10;
                    DWORD dwFuncAddr = (DWORD)pEnv->pfnGetProcAddress(hDll, (LPCSTR)((*(DWORD*)(dwINT)) + dwImageBase + 2));
                    *(DWORD*)dwIAT = dwMyAddr;
                    *(BYTE*)dwMyAddr = 0x68;    //push
                    *(DWORD*)(dwMyAddr + 1) = dwFuncAddr;  //真实函数地址
                    *(BYTE*)(dwMyAddr + 5) = 0xC3;  //retn
                }

五

总结

最后来看看效果，对扫雷进行加壳，加壳后的程序可以正常运行。

先用x64找到真实入口点，进行一波dump操作。

dump后无法直接运行。

然后去x64里使用自动搜索修复导入表的功能，可以看到，IAT这里存放的压根就不是真实API的地址，所以x64也无法识别出来。

 搞定！

看雪ID：橘喵Cat

http://bbs.pediy.com/user-home-940656.htm

*本文由看雪论坛 橘喵Cat 原创，转载请注明来自看雪社区

#

往期推荐

1. formbook脱壳记

2. CVE-2021-1732提权漏洞学习笔记

3. 带加密字符串的.NET样本分析的一些技巧

4. 对一个随身WIFI设备的漏洞挖掘尝试

5. [安全运维向]模拟搭建小型企业内网

6. 某设备CoAP协议漏洞挖掘实战

球分享

球点赞

球在看

点击“阅读原文”，了解更多！

「」