6.3 ‐ How to deploy pre trained Deep Learning models for ESCU

Steps to deploy a pre-trained deep learning model for ESCU

NOTE: The following instructions are specifically for deploying pre-trained deep learning models using Splunk App for Data Science and Deep Learning (DSDL)5.0 in ESCU. The deployed models are then used with "apply" command in ESCU detections.

The steps are outlined as follows -

1. Set up Splunk App for Data Science and Deep Learning (DSDL) 5.0
2. Download the model artifacts
3. Deploy model artifacts into Splunk app DSDL

Set up Splunk App for Data Science and Deep Learning (DSDL)

1. Install the DSDL app (https://splunkbase.splunk.com/app/4607/) 5.0 on Splunk instance and follow the steps in the Overview > User Guide.
2. Additional information and FAQs are available here https://splunkbase.splunk.com/app/4607/#/details.
3. Ensure the containers use Golden Image CPU 5.0.0

Download the model artifacts

1. Download the pre-trained model file .tar.gz from the link provided here.

Detection	Model
Detect DGA domains using pre-trained model in DSDL	pre-trained model
Detect DNS Tunneling using TXT record responses	pre-trained model
Detect suspicious process names	pre-trained model
Detect DNS Data Exfiltration	pre-trained model

2. Download notebook .ipynb for the detection

Detection	Notebook
Detect DGA domains using pre-trained model in DSDL	notebook
Detect DNS Tunneling using TXT record responses	notebook
Detect suspicious process names	notebook
Detect DNS Data Exfiltration	notebook

3. Download model configuration file .json for the detection

Detection	.json file
Detect DGA domains using pre-trained model in DSDL	.json
Detect DNS Tunneling using TXT record responses	.json
Detect suspicious process names	.json
Detect DNS Data Exfiltration	.json

4. Follow the below steps only if you are deploying the model to use outside of ESCU. For ex: in a simple SPL search using |apply command. Place these .mlmodel files into app context /lookup directory. For ex: etc/apps/mltk-container/lookups

Detection	.mlmodel file
Detect DGA domains using pre-trained model in DSDL	.mlmodel
Detect DNS Tunneling using TXT record responses	.mlmodel
Detect suspicious process names	.mlmodel
Detect DNS Data Exfiltration	.mlmodel

Deploy the model artifacts

1. Login into the Splunk instance and launch the Splunk App for Data Science and Deep Learning (DSDL).
2. Select Containers from the drop-down menu and it should list all the containers.
3. Select Container Image as Golden image CPU 5.0.0 and Cluster target as per env setup and start the dev container.
4. Wait for the container to start up and urls to populate for the container.
5. Login into the Jupyter lab of dev container by clicking on the url, ex: http://{container_url}:port_num/lab?
   • Use the password provided in the Overview > User Guide of DSDL app
6. The below steps are performed within the Jupyter Lab of the container.

   • Upload the pre-trained model .tar.gz file into app/model/data path using the upload option in the Jupyter notebook.

   • Open a terminal on Jupyterlab and execute the following commands

      tar -xf app/model/data/<pretrained_model_file_name>.tar.gz -C app/model/data/<pretrained_model_file_name>/
     

     This will extract the artifact <pretrained_model_file_name>.tar.gz into app/model/data/<pretrained_model_file_name>/

   • Upload .ipynb notebook into notebooks folder using the upload option in Jupyter lab.

   • Also, upload .json model configuration into notebooks/data folder.

   • Save the notebook using the save option in Jupyter notebook.

7. Start the container specific to the detection.