Obfuscation

Overview

In this example, we use CodableLLM to generate a dataset designed for function name recovery — the task of predicting a function's original name based on its source code or decompiled implementation.

This dataset can be used to train models for:

• Reverse engineering tasks
• Deobfuscation tooling
• Enhancing binary analysis by recovering semantic information from stripped binaries

Creating the Datasets

To support this use case, we generate two datasets:

1. A dataset of the original source code and decompiled functions (deobfuscated)
2. A dataset of the transformed versions with obfuscated function names

These can be merged later using metadata to create aligned training pairs for function name recovery tasks.

```python from codablellm import compile_dataset, DecompiledCodeDatasetConfig, ExtractConfig

def obfuscate_name(source: SourceFunction) -> SourceFunction: ... # Replace the function name with an ambiguous identifier (e.g., func_1)

Original (deobfuscated) dataset

deobfuscated_dataset = compile_dataset( 'path/to/demo-c-repo', [ 'path/to/demo-c-repo/main_app', 'path/to/demo-c-repo/tool', ], 'make', dataset_config=DecompiledCodeDatasetConfig( extract_config=ExtractConfig(), generation_mode='path' # No transformation ) )

Transformed (obfuscated) dataset

obfuscated_dataset = compile_dataset( 'path/to/demo-c-repo', [ 'path/to/demo-c-repo/main_app', 'path/to/demo-c-repo/tool', ], 'make', dataset_config=DecompiledCodeDatasetConfig( extract_config=ExtractConfig( transform=obfuscate_name ), generation_mode='temp' # Applies transform in a temp copy ) )

deobfuscated_dataset.save_as('deobfuscated_dataset.csv') obfuscated_dataset.save_as('obfuscated_dataset.csv')