albertan017

6 ms·

1. 1.▲
   by albertan017 3y ago

   We find several work on refining Ghidra decompilation results with GPTs, that could be another interesting directions!
2. 2.▲
   by albertan017 3y ago

   Yes, GPT4 is very impressive, as it's not directly trained on the decompilation. We're working on improving our model, please keep watching updates!
3. 3.▲
   by albertan017 3y ago

   Thanks. We acknowledge that an LLM cannot completely replace human expertise in decompilation, much like GPT-4 has not achieved true human-like intelligence. However, the aim of our llm4decompile project is to do something like GPT-4, and o
4. 4.▲
   by albertan017 3y ago

   Thanks! The concern is how to uniformly uplift binary code from various architectures and configurations to the same IR like RzIL? Is there a method to automate the disassembly process reliably across these different systems?
5. 5.▲
   by albertan017 3y ago

   Ideally, with a substantial dataset of obfuscated JavaScript and corresponding raw code, a language model could potentially make good predictions. The first key difficulty, however, is collecting a large-scale dataset and setting up a syste
6. 6.▲
   by albertan017 3y ago

   Thanks! We're working on Ghidra/IDA pro. The problem we face is the right kind of data to test with and how to evaluate it. It's like there's no "standard" benchmark/metrics that everyone uses for decompil
7. 7.▲
   by albertan017 3y ago

   Thanks! But people want an all-in-one solution for decompilation. Given the vast array of architectures and compilation settings, and the fact that these information are usually not predetermined, finding a way to effectively navigate this
8. 8.▲
   by albertan017 3y ago

   Thanks! The model is trained only for O0-3, not support for obfuscation. There's still a long way for llm to go.
9. 9.▲
   by albertan017 3y ago

   We're interested in the toolchain, could you share the link or reference to it? GPT4 does an amazing work, we're also very surprised that it can work.
10. 10.▲
    by albertan017 3y ago

    Thanks! We're working on Ghidra/IDA pro. The problem we face is the right kind of data to test with and how to evaluate it. It's like there's no "standard" benchmark/metrics that everyone uses for decompil
11. 11.▲
    by albertan017 3y ago

    Yes, it's not easy to train a 33B model. An interesting point is, naive fine-tuning, which means if one followed the standard way to fine-tune the model. Training a larger model is tricky, not only the data amount matters, everything l
12. 12.▲
    by albertan017 3y ago

    Thanks! Our initial experiments indicate that for simple cases, such as short snippets (tens of lines) of code without external dependencies, the LLM can decompile very well. However, for more complicated examples, it tends to offer specula
13. 13.▲
    by albertan017 3y ago

    I found lots of decompilation work are conducted on C. It seems not much python projects are compiled into binaries.
14. 14.▲
    by albertan017 3y ago

    Thanks! Training a language model from scratch is data-intensive; Llama2 was developed using 2 trillion tokens, while our dataset is around 4 billion. The appropriate size of the model is not straightforward to determine. In our experiments
15. 15.▲
    by albertan017 3y ago

    Thanks! Processing raw binary data directly would be inefficient for the language model, as it's not designed to interpret strings of zeros and ones but for understanding higher-level instructions (like code and natural language).