【Hook技术】实现从"任务管理器"中保护进程不被关闭 + 附带源码 + 进程保护知识扩展
公司有个监控程序涉及到进程的保护问题，需要避免用户通过任务管理器结束掉监控进程，这里使用了HOOK技术，通过Hook OperProcess来实现进程的保护.
正常的结束进程的流程是(应用层)
a.OpenProcess 打开进程,获取进程的句柄.
b.将a获取的进程句柄传递给TerminateProcess,最后由TermianteProcess来完成进程的关闭.
ps:TerminateProcess又会调用系统的NtTerminateProcess，然后逐步深入内核层，最终调用内核API完成进程的关闭和进程相关资源的释放.
(在应用层大多数进程的关闭都是走上面流程的，包括任务管理器，所以我们知道在Opernprocess和TerminateProcess间，我们只要Hook OpenProcess,让程序无法打开进程获取到句柄就可以了，那样TerminateProcess调用自然也就失败了)
HANDLE WINAPI OpenProcess(
_In_ DWORD dwDesiredAccess,
_In_ BOOL bInheritHandle,
_In_ DWORD dwProcessId //进程ID，我们所关注的
);
由OpenProcess函数原型我们知道，第三个参数dwProcessId标识要打开进程的ID，而我们可以通过拦截OpenProcess，然后判断dwProcess是否为保护的进程ID.
例如:我们自定义的Hook_OpenProcess
HANDLE WINAPI Hook_OpenProcess(DWORD dwDesiredAccess,BOOL bInheritHandle,DWORD dwProcessId) { //判断打开进程的权限和PID，其中lpData->dwProcessId存储我们要保护的进程ID if(dwDesiredAccess == PROCESS_TERMINATE && dwProcessId == lpData->dwProcessId) { //为保护的进程直接返回NULL给TerminateProcess return NULL; } return OpenProcess(dwDesiredAccess,bInheritHandle,dwProcessId); }
定制好了自己的OpenProcess的函数，接下来我们要做的就是如何将原函数（OpenProcess）替换为我们自己的Hook_OpenProcess，只有程序在调用OpenProcess时候先走我们的Hook_OpenProcess那么保护才能顺序的进行. 而替换的主要方法主要有:
1. 通过Patch IAT表,要求对PE文件格式熟悉
2. 通过修改原函数的入口处几个字节，实现跳转到我们定制的函数,这种技术我们称为inline hook，主要是通过修改入口前5个字节或者7个字节，或者patch函数的中部代码，也可以patch尾部.
我们这里用的是第一种方法，这里HOOK类，使用的《核心编程》中的CAPIHook,例如:
class QHookSrv { public: // Hook a function in all modules QHookSrv(PSTR pszCalleeModName, PSTR pszFuncName, PROC pfnHook, BOOL fExcludeAPIHookMod); // Unhook a function from all modules ~QHookSrv(); // Returns the original address of the hooked function operator PROC() { return(m_pfnOrig); } private: static PVOID sm_pvMaxAppAddr; // Maximum private memory address static QHookSrv* sm_pHead; // Address of first object QHookSrv* m_pNext; // Address of next object PCSTR m_pszCalleeModName; // Module containing the function (ANSI) PCSTR m_pszFuncName; // Function name in callee (ANSI) PROC m_pfnOrig; // Original function address in callee PROC m_pfnHook; // Hook function address BOOL m_fExcludeAPIHookMod; // Hook module w/CAPIHook implementation? private: // Replaces a symbol's address in a module's import section static void WINAPI ReplaceIATEntryInAllMods(PCSTR pszCalleeModName, PROC pfnOrig, PROC pfnHook, BOOL fExcludeAPIHookMod); // Replaces a symbol's address in all module's import sections static void WINAPI ReplaceIATEntryInOneMod(PCSTR pszCalleeModName, PROC pfnOrig, PROC pfnHook, HMODULE hmodCaller); };
QHookSrv实现代码:
View Code
PVOID QHookSrv::sm_pvMaxAppAddr = NULL;
const BYTE cPushOpCode = 0x68; // The PUSH opcode on x86 platforms
QHookSrv* QHookSrv::sm_pHead = NULL;
QHookSrv::QHookSrv(PSTR pszCalleeModName, PSTR pszFuncName, PROC pfnHook, BOOL fExcludeAPIHookMod)
{
if(sm_pvMaxAppAddr == NULL)
{
SYSTEM_INFO si;
GetSystemInfo(&si);
sm_pvMaxAppAddr = si.lpMaximumApplicationAddress;
}
m_pNext = sm_pHead; // The next node was at the head
sm_pHead = this; // This node is now at the head
// Save information about this hooked function
m_pszCalleeModName = pszCalleeModName;
m_pszFuncName = pszFuncName;
m_pfnHook = pfnHook;
m_fExcludeAPIHookMod = fExcludeAPIHookMod;
m_pfnOrig = GetProcAddress(GetModuleHandleA(pszCalleeModName), m_pszFuncName);
if (m_pfnOrig > sm_pvMaxAppAddr) {
// The address is in a shared DLL; the address needs fixing up
PBYTE pb = (PBYTE) m_pfnOrig;
if (pb[0] == cPushOpCode) {
// Skip over the PUSH op code and grab the real address
PVOID pv = * (PVOID*) &pb[1];
m_pfnOrig = (PROC) pv;
}
}
// Hook this function in all currently loaded modules
ReplaceIATEntryInAllMods(m_pszCalleeModName, m_pfnOrig, m_pfnHook,
m_fExcludeAPIHookMod);
}
QHookSrv::~QHookSrv()
{
// Unhook this function from all modules
ReplaceIATEntryInAllMods(m_pszCalleeModName, m_pfnHook, m_pfnOrig, m_fExcludeAPIHookMod);
// Remove this object from the linked list
QHookSrv* p = sm_pHead;
if (p == this) { // Removing the head node
sm_pHead = p->m_pNext;
} else {
BOOL fFound = FALSE;
// Walk list from head and fix pointers
for (; !fFound && (p->m_pNext != NULL); p = p->m_pNext) {
if (p->m_pNext == this) {
// Make the node that points to us point to the our next node
p->m_pNext = p->m_pNext->m_pNext;
break;
}
}
}
}
static HMODULE ModuleFromAddress(PVOID pv)
{
MEMORY_BASIC_INFORMATION mbi;
return((VirtualQuery(pv, &mbi, sizeof(mbi)) != 0)
? (HMODULE) mbi.AllocationBase : NULL);
}
void QHookSrv::ReplaceIATEntryInAllMods(PCSTR pszCalleeModName, PROC pfnCurrent, PROC pfnNew, BOOL fExcludeAPIHookMod)
{
HMODULE hmodThisMod = fExcludeAPIHookMod ? ModuleFromAddress(ReplaceIATEntryInAllMods) : NULL;
HANDLE m_hSnapshot = CreateToolhelp32Snapshot(TH32CS_SNAPMODULE, GetCurrentProcessId());
if(m_hSnapshot != INVALID_HANDLE_VALUE)
{
MODULEENTRY32 me = { sizeof(me) };
for (BOOL fOk = Module32First(m_hSnapshot, &me); fOk; fOk = Module32Next(m_hSnapshot, &me))
{
// NOTE: We don't hook functions in our own module
if (me.hModule != hmodThisMod)
{
// Hook this function in this module
ReplaceIATEntryInOneMod(pszCalleeModName, pfnCurrent, pfnNew, me.hModule);
}
}
CloseHandle(m_hSnapshot);
}
}
void QHookSrv::ReplaceIATEntryInOneMod(PCSTR pszCalleeModName,
PROC pfnCurrent, PROC pfnNew, HMODULE hmodCaller) {
// Get the address of the module's import section
ULONG ulSize;
PIMAGE_IMPORT_DESCRIPTOR pImportDesc = (PIMAGE_IMPORT_DESCRIPTOR)ImageDirectoryEntryToData(hmodCaller, TRUE, IMAGE_DIRECTORY_ENTRY_IMPORT, &ulSize);
if (pImportDesc == NULL)
return; // This module has no import section
// Find the import descriptor containing references to callee's functions
for (; pImportDesc->Name; pImportDesc++) {
PSTR pszModName = (PSTR) ((PBYTE) hmodCaller + pImportDesc->Name);
if (lstrcmpiA(pszModName, pszCalleeModName) == 0)
break; // Found
}
if (pImportDesc->Name == 0)
return; // This module doesn't import any functions from this callee
// Get caller's import address table (IAT) for the callee's functions
PIMAGE_THUNK_DATA pThunk = (PIMAGE_THUNK_DATA)
((PBYTE) hmodCaller + pImportDesc->FirstThunk);
// Replace current function address with new function address
for (; pThunk->u1.Function; pThunk++) {
// Get the address of the function address
PROC* ppfn = (PROC*) &pThunk->u1.Function;
// Is this the function we're looking for?
BOOL fFound = (*ppfn == pfnCurrent);
if (!fFound && (*ppfn > sm_pvMaxAppAddr)) {
// If this is not the function and the address is in a shared DLL,
// then maybe we're running under a debugger on Windows 98. In this
// case, this address points to an instruction that may have the
// correct address.
PBYTE pbInFunc = (PBYTE) *ppfn;
if (pbInFunc[0] == cPushOpCode) {
// We see the PUSH instruction, the real function address follows
ppfn = (PROC*) &pbInFunc[1];
// Is this the function we're looking for?
fFound = (*ppfn == pfnCurrent);
}
}
if (fFound) {
// The addresses match, change the import section address
WriteProcessMemory(GetCurrentProcess(), ppfn, &pfnNew,
sizeof(pfnNew), NULL);
return; // We did it, get out
}
}
}
挂钩OpenProcess ：
//Hook_OpenProcess为定制的拦截函数 QHookSrv g_OpenProcess("kernel32.dll", "OpenProcess",(PROC)Hook_OpenProcess,TRUE);
有了IAT hook类，也有了拦截函数，接下来我们要做的就是如何patch所有的进程中duiOpenProcess的调用了,这里我们通过安装全局的WH_SHELL钩子来实现
应用层消息钩子安装通过API：
HHOOK WINAPI SetWindowsHookEx( _In_ int idHook, //钩子的类型(消息钩子) _In_ HOOKPROC lpfn, //钩子回调，当前响应消息被触发时候调用 _In_ HINSTANCE hMod, //实例模块句柄,一般是DLL句柄 _In_ DWORD dwThreadId //0标识全局钩子，非0标识局部钩子 );
消息钩子的卸载:
//只有一个参数，通过SetWindowsHookEx返回的钩子句柄 BOOL WINAPI UnhookWindowsHookEx( _In_ HHOOK hhk );
: 本Demo中钩子的安装和卸载过程
//安装钩子 ,由Dll导出 //参数pid标识保护进程的ID BOOL InstallHook(DWORD pid) { BOOL bResult=FALSE; if(!glhHook) { glhHook = SetWindowsHookEx(WH_SHELL,ShellHookProc,glhInstance, 0); if(glhHook!=NULL) { //´´½¨ÄÚ´æ¹²Ïí hMapping=CreateFileMapping((HANDLE)0xFFFFFFFF,NULL,PAGE_READWRITE,0,0x100,"PCMONITOR."); if(hMapping != NULL) { lpData=(LPSHWP_STRUCT)MapViewOfFile(hMapping,FILE_MAP_ALL_ACCESS,0,0,0); lpData->dwProcessId = pid; } bResult=TRUE; } } return bResult; } //卸载钩子,UninstallHook由Dll导出 BOOL UninstallHook() { BOOL bResult=FALSE; if(glhHook) { bResult= UnhookWindowsHookEx(glhHook); if(bResult) { glhHook=NULL; } } return bResult; }
消息回调函数,回调不进行任何操作调用CallNextHookEx将消息传递下一个处理程序
static LRESULT WINAPI ShellHookProc(int code, WPARAM wParam, LPARAM lParam) { return ::CallNextHookEx(glhHook, code, wParam, lParam); }
+
NET(C#)中的调用:
// QomoHookSrv.dll [DllImport("QomoHookSrv.dll", CallingConvention=CallingConvention.Cdecl)] private extern static bool InstallHook( int processID); [DllImport("QomoHookSrv.dll", CallingConvention=CallingConvention.Cdecl)] private extern static bool UninstallHook(); private void install_hook_when_app_startup() { //获取进程的ID int pid = System.Diagnostics.Process.GetCurrentProcess().Id; //安装钩子 InstallHook(pid); } private void uninstall_hook_when_app_closed() { UninstallHook(); }
源码仅包含HOOK DLL代码（核心代码)， 测试工程请自行code，文章也已经提及了，相对简单
我是Source
【进程保护知识扩展】
本文中提及是从应用层的角度来保护进程不被关闭，通过Hook OperProcess 和TerminateProcess并不是安全了，应用程进程的关闭有些不是走该流程的，比如说直接调用NtTerminateProcess或者通过NtSetSystemInformation等
进程关闭的流程大致为（系统API间的调用关系);
通过上述流程相应的进程保护方法:
1.应用层Hook OpenProcess/TerminateProcess ，或者调用RtlSetprocessiscritical
2.内核层方式实现进程保护相对方式比较多点:
a. Hook NtTerminaterProcess/ZwTerminateProcess（两个是一样的，在R3是为NtTerminateProcess，内核就有NtTerminateProcess转为ZwTerminateProcess调用),这种拦截可以通过修改内核表SSDT，或者InlineHook实现
b.Hook ObReferenceObjectByHandle ,网上有很多关于该API的HoOK源码，也是进程保护中常用的一种手段，因为后续的很多操作都是需要通过调用该函数返回的进程对象体
c. Hook PspTerminateThreadByPointer/PspTerminateProcess/PspExitThread ,pspXX函数是比较底层的API了，而未导出，获取他们地址需要通过硬编码的方式，所以不同的系统可能存在问题，所以这种方法是相对比较危险的，但是确实很有效的
d. Hook 插APC函数，例如:KeInitializeApc等这种方法也相对比较实用多，网上也有很多这方面的例子，不过大多数都是通过插APC的方式来强制结束进程的.
专注于: Net分布式技术，移动服务端架构及系统安全学习及研究 by Andy