Python reimplementation of BSim's similarity and confidence algorithms.
You need to generate/export function signatures from a Ghidra project before you can use py-bsim, which you can do using Ghidra's built-in bsim command-line tool as shown below.
Below: /path/to/ghidra/project/directory should be the path of a .gpr file, without the ".gpr" suffix. For example, if /home/user/project.gpr exists, the second argument should be ghidra:/home/user/project. Ghidra is very picky about this and has terrible error messages if you get it wrong.
$ ghidra_11.0.1_PUBLIC/support/./bsim generatesigs ghidra:/path/to/ghidra/project/directory /path/to/output/signature/directory --config=medium_64from pathlib import Path
import py_bsim
# Load weights (XML file included with Ghidra)
weights = py_bsim.BSimWeights.load_from_path(
Path('ghidra_11.0.1_PUBLIC/Ghidra/Features/BSim/data/lshweights_64.xml'))
# Load signatures dumped from Ghidra project
signatures = py_bsim.BSimSignatureDump.load_from_path(
Path('/path/to/output/signature/directory/signatures.xml'))
functions = signatures.as_combined_function_dict()
# Calculate self-significance of a function
self_sig = py_bsim.lshcosine_self_significance(functions['example_function_1'], weights)
print(f'example_function_1 self-significance: {self_sig}')
# Calculate similarity between two functions
similarity = py_bsim.lshcosine_similarity(
functions['example_function_1'],
functions['example_function_2'],
weights,
)
print(f'Similarity between example_function_1 and example_function_2: {similarity}')
# Calculate confidence between two functions
confidence = py_bsim.lshcosine_confidence(
functions['example_function_1'],
functions['example_function_2'],
weights,
)
print(f'Confidence between example_function_1 and example_function_2: {similarity}')