Allowing visualize and patient list while low cell masking and noise is enabled.

[glabrie10]

Our group at HDC(University of Colorado) has lowcellmasking enabled. Due to university policy, this cannot be adjusted. With help from nic this is the proposed patch. With hopes that this becomes a feature of leaf. Proposed changes from the leaf team:

1. Comment out / remove this check in the API:
   (

   leaf/src/server/Model/Cohort/DemographicProvider.cs

   Line 138 in 28b216e

   throw new Exception("Demographics cannot be returned if Cohort De-identification Low Cell Size Masking is enabled");

   )

2. Do the admin in API here:
   https://github.com/uwrit/leaf/blob/master/src/server/Model/Cohort/DatasetProvider.cs#L115

3. In the web client, add a check similar to the line, to see if the cohort count is less than your threshold:
   https://github.com/uwrit/leaf/blob/master/src/ui-client/src/containers/Visualize/Visualize.tsx#L108

Add this to your conditional:

'<div className={${c}-error}>

Sorry Visualizations can't be shown for cohorts with 10 patients of less.

'

4. Same here
   https://github.com/uwrit/leaf/blob/master/src/ui-client/src/containers/PatientList/PatientList.tsx#L84

With this I will be adding additional changes for noise. When this is tested I will submit solution here.

Gabe
University of Colorado

[ndobb]

Thanks @glabrie10, this seems reasonable, and really the app should allow for cases where low cell masking is used but cohort obfuscation is not, as the former alone doesn't preclude visualizations unless the cohort count happens to be below the threshold.

[glabrie10]

Final Change proposal for allowing visualize and patient list while low cell masking and noise is enabled @ndobb :

1. Comment out lines 132 through 139 in leaf/src/server/Modal/Cohort/DemographicProvider.cs to allow the visualize feature while low-cell-masking and cohort obfuscation is enabled.

        void ThrowIfSettingsInvalid()
        {
            if (!clientOpts.PatientList.Enabled && !clientOpts.Visualize.Enabled)
            {
                throw new Exception("Both Visualize and Patient List are disabled");
            }
           /* if (deidentOpts.Cohort.Noise.Enabled)
            {
                throw new Exception("Demographics cannot be returned if Cohort De-identification Noise is enabled");
            }
            if (deidentOpts.Cohort.LowCellSizeMasking.Enabled)
            {
                throw new Exception("Demographics cannot be returned if Cohort De-identification Low Cell Size Masking is enabled");
            } */
        }

2. Comment out lines 109-115 in leaf/src/server/Modal/Cohort/DatasetProviders.cs to allow the patientlist feature while low-cell-masking and cohort obfuscation is enabled.

     < void ThrowIfSettingsInvalid()
        {
            if (!clientOpts.PatientList.Enabled && !clientOpts.Visualize.Enabled)
            {
                throw new Exception("Both Visualize and Patient List are disabled");
            }
           /* if (deidentOpts.Cohort.Noise.Enabled)
            {
                throw new Exception("Demographics cannot be returned if Cohort De-identification Noise is enabled");
            }
            if (deidentOpts.Cohort.LowCellSizeMasking.Enabled)
            {
                throw new Exception("Demographics cannot be returned if Cohort De-identification Low Cell Size Masking is enabled");
            } */
        }

3. Create conditional from lines 112 to 122 in leaf/src/ui-client/src/containers/Visualize/Visualize.tsx that blocks the visualize feature if cohort size is less than or equal to 10.

     /**
     * Blocks visualize if the query is 10 or smaller
    */
     if (cohort.count.value < 11 ) {
          return (
               <div className={`${c}-error`}>
                    <p>
                         Sorry, visualize can't be seen for cohorts that are equal to or less than 10
                   </p>
               </div>
          );
     }

4. Create conditional from lines 87 to 98 in leaf/src/ui-client/src/containers/PatientList/Patientlist.tsx that blocks the patientlist feature if cohort size is less than or equal to 10.

     /**
     * Blocks patientlist if the query is 10 or smaller
    */
     if (cohort.count.value < 11 ) {
          return (
               <div className={`${c}-error`}>
                    <p>
                         Sorry, patient list can't be seen for cohorts that are equal to or less than 10
                   </p>
               </div>
          );
     }

[ndobb]

Thanks @glabrie10, I'm trying to carve out some time to do a proper code review. At a glance the main thing I'd want to tweak is the hard-coding of the LowCellCountThreshold at 10.

It looks like the API isn't currently sending the LowCellSizeMaskingOptions, so the client wouldn't know what the threshold value is.

Given that, we'd first need to weave those into the server ConfigDTO object (so they'd be included in config data sent to client) and also add to the corresponding client ConfigDTO TypeScript definition.

Finally the line if (cohort.count.value < 11 ) { would need to be updated to look for that value.

[ndobb]

Actually this brings up a possible security vulnerability as well.

If I'm reading correctly, the current code proposal above puts trust fully in the client to not show Visualize/Demographics data from the API if below a certain threshold (in fact, we'd also need to add a check to not even request demographics here).

A bad actor however could still simply call the API (outside of Leaf) and get this data, as the server would not enforce any checks as to cohort size.

If we needed the server to do a check of cohort size before retrieving demographics, we could add code to do so but it would first necessitate a database call to check, which is not ideal.

So there are quite a few additional considerations.

[glabrie10]

If our data is de-identified could we utilize this: cohort.networkCohorts.size === 1, in the conditional? Like this:

     /**
     * Blocks visualize if the query is 10 or smaller
    */
     if (cohort.networkCohorts.size === 1 && cohort.count.value < 11 ) {
          return (
               <div className={`${c}-error`}>
                    <p>
                         Sorry, visualize can't be seen for cohorts that are equal to or less than 10
                   </p>
               </div>
          );
     }

and

     /**
     * Blocks patientlist if the query is 10 or smaller
    */
     if (cohort.networkCohorts.size === 1 && cohort.count.value < 11 ) {
          return (
               <div className={`${c}-error`}>
                    <p>
                         Sorry, patient list can't be seen for cohorts that are equal to or less than 10
                   </p>
               </div>
          );
     }

[glabrie10]

@ndobb after reading and reviewing your comments, I am wondering if it is even smart to apply this and pen test it. Would the API call in bad faith from the client return de-identified data if it is enabled?

For the change mentioned here;

If I'm reading correctly, the current code proposal above puts trust fully in the client to not show Visualize/Demographics data from the API if below a certain threshold (in fact, we'd also need to add a check to not even request demographics https://github.com/uwrit/leaf/blob/ef63db8e18ee2b4a312bb297437fcb068bf16da8/src/ui-client/src/actions/cohort/count.ts#L197)

Could I add a check to the conditional that checks to see if the cohort size is greater than 10?

As for the coding, it is specific to our use case and not for a general patch to use universally throughout the community.

[ndobb]

Hey @glabrie10, sorry for the delay in my reply. Whether the bogus API call in question returns de-identified data depends on your appsettings.json config.

Regardless, it would return the actual (ie, real number of patients) row-level data for the cohort, unfortunately defeating the purpose of using the "10 patients or less" feature.