LaMBdA: Learning-based Deobfuscation of MBA

Mixed boolean arithmetic deobfuscation using graph neural networks.

Usage:

python train.py [-h] [--epochs | -e EPOCHS] [-gpu | -g DEVICE_ID] [--lr LR] [--big | -b] [--res | -r]

python test.py [-h] [--gpu | -g DEVICE_ID]

Example Usage:

python train.py --epochs 420 --gpu 2 --lr 0.0001 --big --res

python test.py --gpu -1

Training Arguments

Argument	Type	Description
-h, --help	None	shows argument help message
-g, --gpu	INT	specifies device ID to use. [0, N] for GPU and -1 for CPU (default=-1)
-e, --epochs	INT	number of epochs to train (default=1000000)
-bs, --batch_size	INT	batch size (default=1)
-lr	FLOAT	learning rate (default=0.00005)
-b, --big	BOOL	enables usage of bigger graphs (default=True)
-r, --res	BOOL	enables interim result nodes (default=False)

Testing Arguments

Argument	Type	Description
-h, --help	None	shows argument help message
-g, --gpu	INT	specifies device ID to use. [0, N] for GPU and -1 for CPU (default=-1)