Simple Hooking of Functions not Exported by Ntoskrnl.exe

本文介绍了一种Hooking未被NtOsKrnl导出的函数的方法,通过映射ntdll.dll到进程空间并直接从DLL中获取所需函数位置,避免了硬编码位置带来的问题。该方法只需了解PE文件格式及一些未公开API即可实现。
Simple Hooking of Functions not Exported by Ntoskrnl.exe
  By: gareth

 

As many of you will know, hooking functions not exported by ntoskrnl.exe is a real pain, as you need to hard code their position in KeServiceDescriptorTable, and this will change between windows releases.

Hardcoding the positions seems a poor solution, since it means after a new service pack, the rootkit may no longer work and become discovered.

As I have found the code on this site extremely helpful, I think it is only fair that I return the favour ;-)

I have implemented the method described in previous posts, whereby I have mapped a view of ntdll.dll into the process space of whoever loads the driver initially, and then retrieve the required function positions directly from the dll.

This was relatively simple to do, and only requires knowledge of the pe file format, and a few undocumented apis.

Using the function pasted below, when hooking you simply do as follows:


RtlInitUnicodeString(&dllName, L"//Device//HarddiskVolume1//Windows//System32//ntdll.dll");
functionAddress = GetDllFunctionAddress(functionName, &dllName);
position = *((WORD*)(functionAddress+1));
    
g_OriginalZwCreateProcessEx = (ZWCREATEPROCESSEX)(KeServiceDescriptorTable.ServiceTableBase[position]);


and here's the function GetDllFunctionAddress:


DWORD GetDllFunctionAddress(char* lpFunctionName, PUNICODE_STRING pDllName)
{
    HANDLE hThread, hSection, hFile, hMod;
    SECTION_IMAGE_INFORMATION sii;
    IMAGE_DOS_HEADER* dosheader;
    IMAGE_OPTIONAL_HEADER* opthdr;
    IMAGE_EXPORT_DIRECTORY* pExportTable;
    DWORD* arrayOfFunctionAddresses;
    DWORD* arrayOfFunctionNames;
    WORD* arrayOfFunctionOrdinals;
    DWORD functionOrdinal;
    DWORD Base, x, functionAddress;
    char* functionName;
    STRING ntFunctionName, ntFunctionNameSearch;
    PVOID BaseAddress = NULL;
    SIZE_T size=0;

    OBJECT_ATTRIBUTES oa = {sizeof oa, 0, pDllName, OBJ_CASE_INSENSITIVE};

    IO_STATUS_BLOCK iosb;

    //_asm int 3;
    ZwOpenFile(&hFile, FILE_EXECUTE | SYNCHRONIZE, &oa, &iosb, FILE_SHARE_READ, FILE_SYNCHRONOUS_IO_NONALERT);

    oa.ObjectName = 0;

    ZwCreateSection(&hSection, SECTION_ALL_ACCESS, &oa, 0,PAGE_EXECUTE, SEC_IMAGE, hFile);
    
    ZwMapViewOfSection(hSection, NtCurrentProcess(), &BaseAddress, 0, 1000, 0, &size, (SECTION_INHERIT)1, MEM_TOP_DOWN, PAGE_READWRITE);
    
    ZwClose(hFile);
    
    hMod = BaseAddress;
    
    dosheader = (IMAGE_DOS_HEADER *)hMod;
    
    opthdr =(IMAGE_OPTIONAL_HEADER *) ((BYTE*)hMod+dosheader->e_lfanew+24);

    pExportTable =(IMAGE_EXPORT_DIRECTORY*)((BYTE*) hMod + opthdr->DataDirectory[ IMAGE_DIRECTORY_ENTRY_EXPORT]. VirtualAddress);

    // now we can get the exported functions, but note we convert from RVA to address
    arrayOfFunctionAddresses = (DWORD*)( (BYTE*)hMod + pExportTable->AddressOfFunctions);

    arrayOfFunctionNames = (DWORD*)( (BYTE*)hMod + pExportTable->AddressOfNames);

    arrayOfFunctionOrdinals = (WORD*)( (BYTE*)hMod + pExportTable->AddressOfNameOrdinals);

    Base = pExportTable->Base;

    RtlInitString(&ntFunctionNameSearch, lpFunctionName);

    for(x = 0; x < pExportTable->NumberOfFunctions; x++)
    {
        functionName = (char*)( (BYTE*)hMod + arrayOfFunctionNames[x]);

        RtlInitString(&ntFunctionName, functionName);

        functionOrdinal = arrayOfFunctionOrdinals[x] + Base - 1; // always need to add base, -1 as array counts from 0
        // this is the funny bit.  you would expect the function pointer to simply be arrayOfFunctionAddresses[x]...
        // oh no... thats too simple.  it is actually arrayOfFunctionAddresses[functionOrdinal]!!
        functionAddress = (DWORD)( (BYTE*)hMod + arrayOfFunctionAddresses[functionOrdinal]);
        if (RtlCompareString(&ntFunctionName, &ntFunctionNameSearch, TRUE) == 0)
        {
            ZwClose(hSection);
            return functionAddress;
        }
    }

    ZwClose(hSection);
    return 0;
}


Hopefully this will be useful to you.
内容概要:本文提出了一种基于融合鱼鹰算法和柯西变异的改进麻雀优化算法(OCSSA),用于优化变分模态分解(VMD)的参数,进而结合卷积神经网络(CNN)与双向长短期记忆网络(BiLSTM)构建OCSSA-VMD-CNN-BILSTM模型,实现对轴承故障的高【轴承故障诊断】基于融合鱼鹰和柯西变异的麻雀优化算法OCSSA-VMD-CNN-BILSTM轴承诊断研究【西储大学数据】(Matlab代码实现)精度诊断。研究采用西储大学公开的轴承故障数据集进行实验验证,通过优化VMD的模态数和惩罚因子,有效提升了信号分解的准确性与稳定性,随后利用CNN提取故障特征,BiLSTM捕捉时间序列的深层依赖关系,最终实现故障类型的智能识别。该方法在提升故障诊断精度与鲁棒性方面表现出优越性能。; 适合人群:具备一定信号处理、机器学习基础,从事机械故障诊断、智能运维、工业大数据分析等相关领域的研究生、科研人员及工程技术人员。; 使用场景及目标:①解决传统VMD参数依赖人工经验选取的问题,实现参数自适应优化;②提升复杂工况下滚动轴承早期故障的识别准确率;③为智能制造与预测性维护提供可靠的技术支持。; 阅读建议:建议读者结合Matlab代码实现过程,深入理解OCSSA优化机制、VMD信号分解流程以及CNN-BiLSTM网络架构的设计逻辑,重点关注参数优化与故障分类的联动关系,并可通过更换数据集进一步验证模型泛化能力。
评论 1
成就一亿技术人!
拼手气红包6.0元
还能输入1000个字符
 
红包 添加红包
表情包 插入表情
 条评论被折叠 查看
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值