Popular GitHub Repositories for Internals

The nnsight package enables interpreting and manipulating the internals of deep learned models.

Gain insights into MS-RPC implementations that may be vulnerable using an automated approach and make it easy to visualize the data. By following this approach, a security researcher will hopefully identify interesting RPC services in such a time that would take a manual approach significantly more.

Gain insights into COM/DCOM implementations that may be vulnerable using an automated approach and make it easy to visualize the data. By following this approach, a security researcher will hopefully identify interesting (D)COM classes/implementations in such a time that would take a manual approach significantly more.

A performance insights and knowledge assistant agent built on top of Chrome DevTools internals, Mastra, AI SDK and NextJS

An 64-bit operating system I'm developing for the x86 architecture

Original reference implementation of "Analyzing the Internals of Neural Radiance Fields"

capybaradb - a toy Vector DB implementation from scratch in Python. Explore Vector DB internals.

Implementation Of SysWhispers Direct / Indirect System Call Technique In D.

Awesome ELF Internals

NT API Code Injection using NtCreateSection and NtMapViewOfSection to inject shellcode into a remote process. This method allows for efficient memory manipulation and code execution within target processes.

ShellBunny is a malleable shellcode loader written in C and Assembly utilizing direct and indirect syscalls for evading EDR hooks. It can be built as EXE, DLL, or XLL and offers a variety of QoL features that make it more adaptable.

Awesome APK Internals

Awesome PE Internals

This is a tool for obfuscating functions in C/C++ programs on Windows. It hides what a function does by changing its code while the program is running, moves that code to another place in memory, and uses Windows exception handling to jump to the moved code when the function is called.

llama.cpp internals

InMemoryRunPE is a Windows proof-of-concept project demonstrating manual PE (Portable Executable) loading and execution directly from memory. It bypasses the default Windows loader by mapping, fixing relocations, resolving imports, and executing the entry point without writing the executable to disk. Includes support for command-line masquerading.