Question
Virtusa
US
Last activity: 17 May 2022 12:18 EDT
ADM Model – Encrypted Predictor Value
We have a requirement to migrate on-prem Pega CDH application to the AWS cloud along with sensitive data encryption.
The CDH application has matured Adaptive (ADM) model which is referring customer sensitive data as one of the predictor value, can you please help me with below questions.
- As per my understanding the ADM predictor value is get stored into the “pr_data_adm_factory” (OOTB) table, can we encrypt the intended customer value which is being used as predictor?
- Is there any impact on the ADM learning if we apply encryption on intended predictor value?
- During which event/process/operation the predictor value is getting referred back by ADM from the table “pr_data_adm_factory” ?
- Is it possible to apply encryption and decryption on intended customer value before and after storing the predictor data into the table “pr_data_adm_factory”? so that ADM can refer plane text predictor value for its BAU processing.
Also please suggest in case there any better alternative to achieve the task. Thank you in advance.
***Edited by Moderator Marije to add Capability tags***
Updated: 12 May 2022 6:12 EDT
Tech Mahindra
IN
@Krishna_Das snapshot model/predictor data are two different things. If you want to extract snapshot predictor data, you can use the https://docs-previous.pega.com/decision-management/84/extracting-historical-responses-adaptive-models?check_logged_in=1
Pega ADM models wont store the predictor values. So if you are encrypting them before passing to ADM, it will create bin issues.
And by the way, why do you want to encrypt the ADM predictor data? Can you give one example here? Usually we don't pass PII details to ADM models.
Updated: 16 May 2022 17:04 EDT
Virtusa
US
Thank you @Nizam for your response, please find my comments below.
1. The ADM models are not enabled to record historical data ( "Record historical data check box" configuration is off for the ADM, also we don't have repository configured ).
2. I can see the predictor values in table "pr_data_adm_factory" under column "PYFACTORY", please find attached screenshot for your reference. (the predictor name and value has been replaced with dummy value in the screenshot. ).
We are not intended to pass the encrypted predictor values to the ADM, instead we want the value to be stored in encrypted form in to the database also able to decrypt value before feeding back to the ADM (if ADM needs to refer from DB) to avoid any impact on ADM learning.
3. We are passing some codes associated with the Customer to the ADM as a predictor.
Looking forward for your response. Thank you.
Tech Mahindra
IN
@AbhinayU Thanks for sharing the screenshot. This holds all the predictor values created for the symbolic bins. All the predictor values will be stored in pxADMInputs property under pxDecisionResults data set once the decision is made. After the feedback is received, it will feed the model response back with the outcome. That means, the values which you pass to ADM rule will be stored under pxADMInputs. What is the issue if you pass the actual value to ADM??