fsabiu
/

keras-modelscan-torchmodulewrapper-coverage-gap

Model card Files Files and versions

keras-modelscan-torchmodulewrapper-coverage-gap / detector-summary.md

fsabiu's picture

Add Keras ModelScan coverage validation artifacts

496e58e verified 10 days ago

|

history blame contribute delete

808 Bytes

	# EXP-MFV-20260502-002 Summary

	\| File \| Detector findings \| ModelScan issues \| Gap \|
	\| --- \| ---: \| ---: \| --- \|
	\| `benign_dense.keras` \| `0` \| `0` \| `no` \|
	\| `benign_external_vocab_lookup.keras` \| `0` \| `0` \| `no` \|
	\| `benign_functional.keras` \| `0` \| `0` \| `no` \|
	\| `benign_lambda.keras` \| `2` \| `1` \| `no` \|
	\| `benign_registered_custom.keras` \| `0` \| `0` \| `no` \|
	\| `benign_torch_module.keras` \| `1` \| `0` \| `yes` \|

	## Detector Findings

	### `benign_lambda.keras`

	- `unsafe_class` at `$.config.layers[1]`: Lambda layer may deserialize Python functions/lambdas.
	- `serialized_lambda` at `$.config.layers[1].config.function`: Serialized Python lambda object.

	### `benign_torch_module.keras`

	- `unsafe_class` at `$.config.layers[1]`: TorchModuleWrapper may deserialize torch.nn.Module via torch.load().