Scan Model Files¶

Once the Model Scanner CLI image has been deployed, it can be run as a container using the docker run command as described below.

See Deploy Model Scanner CLI for deployment instructions.
See Command-line Arguments for more information about CLI input and output arguments.

Set Environment Variables¶

Export your HiddenLayer information (product license, API key, and API secret).

Replace %HIDDENLAYER_LICENSE% with your HiddenLayer license.
Replace %HL_API_KEY% and %HL_API_SECRET% with your HiddenLayer API key and secret, generated in the Console.
Replace %HIDDENLAYER_PLATFORM_HOSTNAME% with the URL for your Console. Example: http://server1.test.hiddenlayer.com.
For Windows systems, replace export with set.

export HL_LICENSE=%HIDDENLAYER_LICENSE%
export HL_CLIENT_ID=%HL_API_KEY%
export HL_CLIENT_SECRET=%HL_API_SECRET%
export HL_MODEL_SCANNER_PLATFORM_RESULTS_URL=%HIDDENLAYER_PLATFORM_HOSTNAME%
export HL_MODEL_SCANNER_TOKEN_URL=%HIDDENLAYER_PLATFORM_HOSTNAME%

Scan a Folder¶

Scan all model files within a specified folder.

Select OS

Select your operating system to view file scan instructions.

macOSLinuxWindows

Replace /home/users/models with the path on the local machine where models to be scanned are located.
Replace <model_name_in_console> with either the name of an existing model from the inventory (to add a new scan version to the existing model) or a new, unique name (to create a new model entry in the inventory).
Using --persist adds the scan results to an existing model in the AISec Platform Console.

docker run --rm \
  -e HL_LICENSE \
  -e HL_CLIENT_ID \
  -e HL_CLIENT_SECRET \
  -e HL_MODEL_SCANNER_PLATFORM_RESULTS_URL \
  -e HL_MODEL_SCANNER_TOKEN_URL \
  -v /home/user/models:/files-to-scan \
  images.hiddenlayer.ai/proxy/aisec-platform/ghcr.io/hiddenlayer-engineering/modelscanner-src:airgapped \
  --input /files-to-scan  --persist --model-name=<model_name_in_console>

Replace /home/users/models with the path on the local machine where models to be scanned are located.
Replace <model_name_in_console> with either the name of an existing model from the inventory (to add a new scan version to the existing model) or a new, unique name (to create a new model entry in the inventory).
Using --persist adds the scan results to an existing model in the AISec Platform Console.

docker run --rm \
  -e HL_LICENSE \
  -e HL_CLIENT_ID \
  -e HL_CLIENT_SECRET \
  -e HL_MODEL_SCANNER_PLATFORM_RESULTS_URL \
  -e HL_MODEL_SCANNER_TOKEN_URL \
  -v /home/user/models:/files-to-scan \
  images.hiddenlayer.ai/proxy/aisec-platform/ghcr.io/hiddenlayer-engineering/modelscanner-src:airgapped \
  --input /files-to-scan  --persist --model-name=<model_name_in_console>

Replace c:/work/models with the path on the local machine where models to be scanned are located.
Replace <model_name_in_console> with either the name of an existing model from the inventory (to add a new scan version to the existing model) or a new, unique name (to create a new model entry in the inventory).
Using --persist adds the scan results to an existing model in the AISec Platform Console.

docker run --rm ^
  -e HL_LICENSE ^
  -e HL_CLIENT_ID ^
  -e HL_CLIENT_SECRET ^
  -e HL_MODEL_SCANNER_PLATFORM_RESULTS_URL ^
  -e HL_MODEL_SCANNER_TOKEN_URL ^
  -v c:/work/models:/files-to-scan ^
  images.hiddenlayer.ai/proxy/aisec-platform/ghcr.io/hiddenlayer-engineering/modelscanner-src:airgapped ^
  --input /files-to-scan  --persist --model-name=<model_name_in_console>

Scan a File¶

Scan a single file by including the file name with the --input.

Select OS

Select your operating system to view file scan instructions.

macOSLinuxWindows

Replace /home/users/models with the path on the local machine where models to be scanned are located.
Replace <model_file> with the file name. Example: --input /files-to-scan/test_model.pkl.
Replace <model_name_in_console> with either the name of an existing model from the inventory (to add a new scan version to the existing model) or a new, unique name (to create a new model entry in the inventory).
Using --persist adds the scan results to an existing model in the AISec Platform Console.

docker run --rm \
  -e HL_LICENSE \
  -e HL_CLIENT_ID \
  -e HL_CLIENT_SECRET \
  -e HL_MODEL_SCANNER_PLATFORM_RESULTS_URL \
  -e HL_MODEL_SCANNER_TOKEN_URL \
  -v /home/user/models:/files-to-scan \
  images.hiddenlayer.ai/proxy/aisec-platform/ghcr.io/hiddenlayer-engineering/modelscanner-src:airgapped \
  --input /files-to-scan/<model_file>  --persist --model-name=<model_name_in_console>

Replace /home/users/models with the path on the local machine where models to be scanned are located.
Replace <model_file> with the file name. Example: --input /files-to-scan/test_model.pkl.
Replace <model_name_in_console> with either the name of an existing model from the inventory (to add a new scan version to the existing model) or a new, unique name (to create a new model entry in the inventory).
Using --persist adds the scan results to an existing model in the AISec Platform Console.

docker run --rm \
  -e HL_LICENSE \
  -e HL_CLIENT_ID \
  -e HL_CLIENT_SECRET \
  -e HL_MODEL_SCANNER_PLATFORM_RESULTS_URL \
  -e HL_MODEL_SCANNER_TOKEN_URL \
  -v /home/user/models:/files-to-scan \
  images.hiddenlayer.ai/proxy/aisec-platform/ghcr.io/hiddenlayer-engineering/modelscanner-src:airgapped \
  --input /files-to-scan/<model_file>  --persist --model-name=<model_name_in_console>

Replace c:/work/models with the path on the local machine where models to be scanned are located.
Replace <model_file> with the file name. Example: --input /files-to-scan/test_model.pkl.
Replace <model_name_in_console> with either the name of an existing model from the inventory (to add a new scan version to the existing model) or a new, unique name (to create a new model entry in the inventory).
Using --persist adds the scan results to an existing model in the AISec Platform Console.

docker run --rm ^
  -e HL_LICENSE ^
  -e HL_CLIENT_ID ^
  -e HL_CLIENT_SECRET ^
  -e HL_MODEL_SCANNER_PLATFORM_RESULTS_URL ^
  -e HL_MODEL_SCANNER_TOKEN_URL ^
  -v c:/work/models:/files-to-scan ^
  images.hiddenlayer.ai/proxy/aisec-platform/ghcr.io/hiddenlayer-engineering/modelscanner-src:airgapped ^
  --input /files-to-scan/<model_file>  --persist --model-name=<model_name_in_console>

Inspect Scan Results of Test ML Model¶

At scan completion, a scan result similar to the following will be emitted to stdout.

Scan results will be JSON-minified, but are shown formatted here for readability.

{
    "scan_id": "13c0f8a4-c938-43ee-9d95-2803ccd7de10",
    "start_time": "2025-02-12T17:55:36.511Z",
    "end_time": "2025-02-12T17:55:36.517Z",
    "status": "done",
    "version": "latest",
    "$schema_version": "3.1.0",
    "inventory": {
        "requested_scan_location": "/files-to-scan/ZS_withConfig.nemo",
        "model_id": "00000000-0000-0000-0000-000000000000",
        "model_name": "cli-latest-cdffd55e-802b-4440-8403-698e930e5bb9",
        "model_version": "1739382936",
        "model_version_id": "00000000-0000-0000-0000-000000000000"
    },
    "file_results": [
        {
            "file_instance_id": "01c8c57d-dd77-488b-b459-709f1933f7b2",
            "file_location": "/files-to-scan/ZS_withConfig.nemo",
            "status": "done",
            "start_time": "2025-02-12T17:55:36.511Z",
            "end_time": "2025-02-12T17:55:36.517Z",
            "details": {
                "sha256": "6a15d94d7c9a67d2574ba8226cfa7d678524899d61f18f98cb218da6e30f0570",
                "file_type": "TAR",
                "file_type_details": {},
                "estimated_time": ""
            },
            "seen": "2025-02-12T17:55:36.511Z",
            "detections": [
                {
                    "detection_id": "b3598a21-bdb0-4f4d-9c66-c7f127894b43",
                    "rule_id": "NEMO_0008_202408",
                    "category": "Decompression Vulnerabilities",
                    "description": "Archive Contains Tar Slip attack, which can be used to overwrite files on the system.",
                    "severity": "high",
                    "mitre_atlas": [
                        {
                            "technique": "AML.T0010",
                            "tactic": "AML.TA0004"
                        }
                    ],
                    "owasp": [
                        "ML06",
                        "LLM05"
                    ],
                    "cwe": "",
                    "cwe_href": ""
                }
            ]
        }
    ],
    "detection_count": 1,
    "file_count": 1,
    "files_with_detections_count": 1,
    "summary": {
        "detection_count": 1,
        "file_count": 1,
        "severity": "high",
        "files_with_detections_count": 1,
        "detection_categories": [
            "Decompression Vulnerabilities"
        ]
    }
}