ggfuchsi-oss/spectrida-reverse_engineering_stack

spectrIDA is a Python tool that makes analyzing compiled binary files significantly faster and less tedious. Compiled binaries (like game DLLs or software executables) can contain hundreds of thousands of unnamed functions with labels like "sub_10001234" that tell you nothing about what they do. Figuring out what those functions are is called reverse engineering, and it is normally slow, painstaking work. The tool does two main things. First, it speeds up the initial analysis phase by splitting the binary into chunks and processing each chunk at the same time in parallel, using multiple workers. The README notes that a file which takes IDA Pro four hours to analyze on its own can be done in 67 seconds using 16 workers. IDA Pro is a professional binary analysis program that spectrIDA depends on. Second, after analysis, spectrIDA uses a local AI model to suggest names for functions. The model was fine-tuned specifically for reverse engineering work and considers the context of which other functions call a given function when making its suggestions. You press a key in the terminal interface and watch it generate a name in real time. The terminal interface has a cyberpunk visual style and supports searching, browsing functions, viewing decompiled pseudocode, and exporting results. Results can be exported as JSON, CSV, or as an IDA script that applies all the AI-generated names back into IDA Pro in one click. There is also a Python API that lets you drive the tool from scripts or notebooks without using the terminal interface at all. The AI model runs locally through a program called Ollama and requires about 8.7 gigabytes of disk space. The whole setup requires IDA Pro 9.x and Python 3.10 or newer. A demo mode lets you try the interface without having IDA Pro or Ollama installed at all. The README describes the target audience as reverse engineers who need orientation on large binaries quickly: game modders, security researchers, and participants in capture-the-flag competitions. It acknowledges the naming accuracy is not perfect (roughly 70 percent on generic functions) but frames the tool as handling the initial survey pass so a human can focus on the parts that matter.