Node add-on for memory reading and writing!
npm install @luchanso/memoryjsmemoryjs is an NPM package for reading and writing process memory! (finally!)
NOTE: version 3 of this library introduces breaking changes that are incompatible with previous versions.
The notable change is that when reading memory, writing memory and pattern scanning you are required to pass the handle
through for the process (that is returned from memoryjs.openProcess). This allows for multi-process support.
- List all open processes
- List all modules associated with a process
- Find a specific module within a process
- Read process memory
- Write process memory
- Read buffers from memory
- Write buffer to memory
- Change memory protection
- Reserve/allocate, commit or change regions of memory
- Fetch a list of memory regions within a process
- Pattern scanning
- Execute a function within a process
- Hardware breakpoints (find out what accesses/writes to this address etc)
Functions that this library directly exposes from the WinAPI:
- ReadProcessMemory
- WriteProcessMemory
- VirtualProtectEx
- VirtualAllocEx
TODO:
- WriteFile support (for driver interactions)
- DLL injections
This is a Node add-on (last tested to be working on v8.11.3) and therefore requires node-gyp to use.
You may also need to follow these steps.
npm install memoryjs
When using memoryjs, the target process should match the platform architecture of the Node version running.
For example if you want to target a 64 bit process, you should try and use a 64 bit version of Node.
You also need to recompile the library and target the platform you want. Head to the memoryjs node module directory, open up a terminal and to run the compile scripts, type:
npm run build32 if you want to target 32 bit processes
npm run build64 if you want to target 64 bit processes
If you are planning to use this module with Node Webkit or Electron, take a look at Liam Mitchell's build notes here.
Initialise:
`` javascript`
const memoryjs = require('memoryjs');
const processName = "csgo.exe";
Open a process (sync):
` javascript`
const processObject = memoryjs.openProcess(processIdentifier);
Open a process (async):
` javascript
memoryjs.openProcess(processIdentifier, (error, processObject) => {
});
`
Get all processes (sync):
` javascript`
const processes = memoryjs.getProcesses();
Get all processes (async):
` javascript
memoryjs.getProcesses((error, processes) => {
});
`
See the Documentation section of this README to see what a process object looks like.
Find a module (sync):
` javascript`
const module = memoryjs.findModule(moduleName, processId);
Find a module (async):
` javascript
memoryjs.findModule(moduleName, processId, (error, module) => {
});
`
Get all modules (sync):
` javascript`
const modules = memoryjs.getModules(processId);
Get all modules (async):
` javascript
memoryjs.getModules(processId, (error, modules) => {
});
`
See the Documentation section of this README to see what a module object looks like.
Read from memory (sync):
` javascript`
const value = memoryjs.readMemory(handle, address, dataType);
Read from memory (async):
` javascript
memoryjs.readMemory(handle, address, dataType, (error, value) => {
});
`
Read buffer from memory (sync):
` javascript`
const buffer = memoryjs.readBuffer(handle, address, size);
Read buffer from memory (async):
` javascript
memoryjs.readBuffer(handle, address, size, (error, buffer) => {
});
`
Write to memory:
` javascript`
memoryjs.writeMemory(handle, address, value, dataType);
Write buffer to memory:
` javascript`
memoryjs.writeBuffer(handle, address, buffer);
Fetch memory regions (sync):
` javascript`
const regions = memoryjs.getRegions(handle);
Fetch memory regions (async):
` javascript
memoryjs.getRegions(handle, (regions) => {
});
`
See the Documentation section of this README to see what values dataType can be.
Set protection of memory:
` javascript`
const oldProtection = memoryjs.virtualProtectEx(handle, address, size, protection);
See the Documentation section of this README to see what values protection can be.
Pattern scanning (sync):
` javascript`
const offset = memoryjs.findPattern(handle, moduleName, signature, signatureType, patternOffset, addressOffset);
Pattern scanning (async):
` javascript
memoryjs.findPattern(handle, moduleName, signature, signatureType, patternOffset, addressOffset, (error, offset) => {
})
`
Function execution (sync):
` javascript`
const result = memoryjs.callFunction(handle, args, returnType, address);
Function execution (async):
` javascript
memoryjs.callFunction(handle, args, returnType, address, (error, result) => {
});
`
Click here to see what a result object looks like.
Clicklick here for details about how to format the arguments and the return type.
Attach a debugger:
` javascript`
const success = memoryjs.attatchDebugger(processId, exitOnDetatch);
Detatch debugger:
` javascript`
const success = memoryjs.detatchDebugger(processId);
Wait for debug devent:
` javascript`
const success = memoryjs.awaitDebugEvent(hardwareRegister, millisTimeout);
Handle debug event:
` javascript`
const success = memoryjs.handleDebugEvent(processId, threadId);
Set a hardware breakpoint:
` javascript`
const success = memoryjs.setHardwareBreakpoint(processId, address, hardwareRegister, trigger, length);
Remove a hardware breakpoint:
` javascript`
const success = memoryjs.removeHardwareBreakpoint(processId, hardwareRegister);
Note: this documentation is currently being updated, refer to the Wiki for more information.
javascript
{ dwSize: 304,
th32ProcessID: 10316,
cntThreads: 47,
th32ParentProcessID: 7804,
pcPriClassBase: 8,
szExeFile: "csgo.exe",
modBaseAddr: 1673789440,
handle: 808 }
`The
handle and modBaseAddr properties are only available when opening a process and not when listing processes.$3
` javascript
{ modBaseAddr: 468123648,
modBaseSize: 80302080,
szExePath: 'c:\\program files (x86)\\steam\\steamapps\\common\\counter-strike global offensive\\csgo\\bin\\client.dll',
szModule: 'client.dll',
th32ProcessID: 10316 }
`$3
` javascript
{ returnValue: 1.23,
exitCode: 2 }
`The
returnValue is the value returned from the function that was called. exitCode is the termination status of the thread.$3
When using the write or read functions, the data type (dataType) parameter can either be a string and be one of the following:
"byte", "int", "int32", "uint32", "int64", "uint64", "dword", "short", "long", "float", "double", "bool", "boolean", "ptr", "pointer", "str", "string", "vec3", "vector3", "vec4", "vector4"or can reference constants from within the library:
memoryjs.BYTE, memoryjs.INT, memoryjs.INT32, memoryjs.UINT32, memoryjs.INT64, memoryjs.UINT64, memoryjs.DWORD, memoryjs.SHORT, memoryjs.LONG, memoryjs.FLOAT, memoryjs.DOUBLE, memoryjs.BOOL, memoryjs.BOOLEAN, memoryjs.PTR, memoryjs.POINTER, memoryjs.STR, memoryjs.STRING, memoryjs.VEC3, memoryjs.VECTOR3, memoryjs.VEC4, memoryjs.VECTOR4This is simply used to denote the type of data being read or written.
Vector3 is a data structure of three floats:
` javascript
const vector3 = { x: 0.0, y: 0.0, z: 0.0 };
memoryjs.writeMemory(address, vector3);
`Vector4 is a data structure of four floats:
` javascript
const vector4 = { w: 0.0, x: 0.0, y: 0.0, z: 0.0 };
memoryjs.writeMemory(address, vector4);
`$3
If you have a structure you want to write to memory, you can use buffers. For an example on how to do this, view the buffers example.
To write a structure to memory, you can use the concentrate library to describe the structure as a buffer
and then write the buffer to memory using the
writeBuffer function.To read a structure from memory, you will need to read a buffer from memory using the
readBuffer function, and then you can use the dissolve library to parse the buffer into a structure.In either case you don't need to use the two libraries mentioned above, they just make it easy to turn your structure into a buffer, and your buffer into a structure.
$3
Protection type is a bit flag DWORD value.
This parameter should reference a constant from the library:
memoryjs.PAGE_NOACCESS, memoryjs.PAGE_READONLY, memoryjs.PAGE_READWRITE, memoryjs.PAGE_WRITECOPY, memoryjs.PAGE_EXECUTE, memoryjs.PAGE_EXECUTE_READ, memoryjs.PAGE_EXECUTE_READWRITE, memoryjs.PAGE_EXECUTE_WRITECOPY, memoryjs.PAGE_GUARD, memoryjs.PAGE_NOCACHE, memoryjs.PAGE_WRITECOMBINE, memoryjs.PAGE_ENCLAVE_THREAD_CONTROL, memoryjs.PAGE_TARGETS_NO_UPDATE, memoryjs.PAGE_TARGETS_INVALID, memoryjs.PAGE_ENCLAVE_UNVALIDATEDRefer to MSDN's Memory Protection Constants for more information.
$3
Memory allocation type is a bit flag DWORD value.
This parameter should reference a constat from the library:
memoryjs.MEM_COMMIT, memoryjs.MEM_RESERVE, memoryjs.MEM_RESET, memoryjs.MEM_RESET_UNDORefer to MSDN's VirtualAllocEx documentation for more information.
$3
You can use this library to read either a "string", or "char*" and to write a string.
In both cases you want to get the address of the char array:
`c++
std::string str1 = "hello";
std::cout << "Address: 0x" << hex << (DWORD) str1.c_str() << dec << std::endl;char* str2 = "hello";
std::cout << "Address: 0x" << hex << (DWORD) str2 << dec << std::endl;
`From here you can simply use this address to write and read memory.
There is one caveat when reading a string in memory however, due to the fact that the library does not know
how long the string is, it will continue reading until it finds the first null-terminator. To prevent an
infinite loop, it will stop reading if it has not found a null-terminator after 1 million characters.
One way to bypass this limitation in the future would be to allow a parameter to let users set the maximum
character count.
$3
When pattern scanning, flags need to be raised for the signature types. The signature type parameter needs to be one of the following:
0x0 or memoryjs.NORMAL which denotes a normal signature.0x1 or memoryjs.READ which will read the memory at the address.0x2 or memoryjs.SUBSTRACT which will subtract the image base from the address.To raise multiple flags, use the bitwise OR operator:
memoryjs.READ | memoryjs.SUBTRACT.$3
Remote function execution works by building an array of arguments and dynamically generating shellcode that is injected into the target process and executed, for this reason crashes may occur.
To call a function in a process, the
callFunction function can be used. The library supports passing arguments to the function and need to be in the following format:`javascript
[{ type: T_INT, value: 4 }]
`The library expects the arguments to be an array of objects where each object has a
type which denotes the data type of the argument, and a value which is the actual value of the argument. The various supported data types can be found below.
` javascript
memoryjs.T_VOID = 0x0,
memoryjs.T_STRING = 0x1,
memoryjs.T_CHAR = 0x2,
memoryjs.T_BOOL = 0x3,
memoryjs.T_INT = 0x4,
memoryjs.T_DOUBLE = 0x5,
memoryjs.T_FLOAT = 0x6,
`When using
callFunction, you also need to supply the return type of the function, which again needs to be one of the above values.For example, given the following C++ function:
` c++
int add(int a, int b) {
return a + b;
}
`You would call this function as so:
`javascript
const args = [{
type: memoryjs.T_INT,
value: 2,
}, {
type: memoryjs.T_INT,
value: 5,
}];
const returnType = T_INT;> memoryjs.callFunction(handle, args, returnType, address);
{ returnValue: 7, exitCode: 7 }
`See the result object documentation for details on what
callFunction returns.Notes: currently passing a
double as an argument is not supported, but returning one is.Much thanks to the various contributors that made this feature possible.
$3
Hardware breakpoints work by attaching a debugger to the process, setting a breakpoint on a certain address and declaring a trigger type (e.g. breakpoint on writing to the address) and then continuously waiting for a debug event to arise (and then consequently handling it).
This library exposes the main functions, but also includes a wrapper class to simplify the process. For a complete code example, checkout our debugging example.
When setting a breakpoint, you are required to pass a trigger type:
-
memoryjs.TRIGGER_ACCESS - breakpoint occurs when the address is accessed
- memoryjs.TRIGGER_WRITE - breakpoint occurs when the address is written toDo note that when monitoring an address containing a string, the
size parameter of the setHardwareBreakpoint function should be the length of the string. When using the Debugger wrapper class, the wrapper will automatically determine the size of the string by attempting to read it.To summarise:
- When using the
Debugger class:
- No need to pass the size parameter to setHardwareBreakpoint
- No need to manually pick a hardware register
- Debug events are picked up via an event listener
- setHardwareBreakpoint returns the register that was used for the breakpoint- When manually using the debugger functions:
- The
size parameter is the size of the variable in memory (e.g. int32 = 4 bytes). For a string, this parameter is the length of the string
- Manually need to pick a hardware register (via memoryjs.DR0 through memoryhs.DR3). Only 4 hardware registers are available (some CPUs may even has less than 4 available). This means only 4 breakpoints can be set at any given time
- Need to manually wait for debug and handle debug events
- setHardwareBreakpoint returns a boolean stating whether the operation as successfulFor more reading about debugging and hardware breakpoints, checkout the following links:
- DebugActiveProcess.aspx) - attatching the debugger
- DebugSetProcessKillOnExit - kill the process when detatching
- DebugActiveProcessStop.aspx) - detatching the debugger
- WaitForDebugEvent.aspx) - waiting for the breakpoint to be triggered
- ContinueDebugEvent.aspx) - handling the event
#### Using the Debugger Wrapper
The Debugger wrapper contains these functions you should use:
` javascript
class Debugger {
attatch(processId, killOnDetatch = false);
detatch(processId);
setHardwareBreakpoint(processId, address, trigger, dataType);
removeHardwareBreakpoint(processId, register);
}
`1. Attach the debugger
` javascript
const hardwareDebugger = memoryjs.Debugger;
hardwareDebugger.attach(processId);
`2. Set a hardware breakpoint
` javascript
const address = 0xDEADBEEF;
const trigger = memoryjs.TRIGGER_ACCESS;
const dataType = memoryjs.INT;
const register = hardwareDebugger.setHardwareBreakpoint(processId, address, trigger, dataType);
`3. Create an event listener for debug events (breakpoints)
` javascript
// debugEvent event emission catches debug events from all registers
hardwareDebugger.on('debugEvent', ({ register, event }) => {
console.log(Hardware Register ${register} breakpoint);
console.log(event);
});// You can listen to debug events from specific hardware registers
// by listening to whatever register was returned from
setHardwareBreakpoint
hardwareDebugger.on(register, (event) => {
console.log(event);
});
`#### When Manually Debugging
1. Attatch the debugger
` javascript
const hardwareDebugger = memoryjs.Debugger;
hardwareDebugger.attach(processId);
`2. Set a hardware breakpoint (determine which register to use and the size of the data type)
` javascript
// available registers: DR0 through DR3
const register = memoryjs.DR0;
// int = 4 bytes
const size = 4;const address = 0xDEADBEEF;
const trigger = memoryjs.TRIGGER_ACCESS;
const dataType = memoryjs.INT;
const success = memoryjs.setHardwareBreakpoint(processId, address, register, trigger, size);
`3. Create the await/handle debug event loop
` javascript
const timeout = 100;setInterval(() => {
//
debugEvent can be null if no event occurred
const debugEvent = memoryjs.awaitDebugEvent(register, timeout); // If a breakpoint occurred, handle it
if (debugEvent) {
memoryjs.handleDebugEvent(debugEvent.processId, debugEvent.threadId);
}
}, timeout);
``Note: a loop is not required, e.g. no loop required if you want to simply wait until the first detection of the address being accessed or written to.