Open Bug 1888216 Opened 2 years ago Updated 2 years ago

Data Review Request for Including Sync Device ID in Backend FxA Glean Pings

Categories

(Cloud Services :: Server: Firefox Accounts, task)

task

Tracking

(Not tracked)

People

(Reporter: bchen, Unassigned)

References

Details

Attachments

(1 file)

In https://mozilla-hub.atlassian.net/browse/FXA-9355 we want to include the "device id" of the user's browser that's signed into Mozilla accounts for Sync in a subset of the backend Glean pings. We'd like a data review for the addition.

Attachment #9393567 - Flags: data-review?(tlong)

Comment on attachment 9393567 [details]
Answers to data collection review form

Data Review

  1. Is there or will there be documentation that describes the schema for the ultimate data set in a public, complete, and accurate way?

Yes, through the metrics.yaml file and the Glean Dictionary.

  1. Is there a control mechanism that allows the user to turn the data collection on and off?

Yes, through the data preferences in the Mozilla account settings.

  1. If the request is for permanent data collection, is there someone who will monitor the data over time?

Permanent collection to be monitored over time by Wil Clouser

  1. Using the category system of data types on the Mozilla wiki, what collection type of data do the requested measurements fall under?

Category 1, technical information

  1. Is the data collection request for default-on or default-off?

Default-on

  1. Does the instrumentation include the addition of any new identifiers (whether anonymous or otherwise; e.g., username, random IDs, etc. See the appendix for more details)?

No

  1. Is the data collection covered by the existing Firefox privacy notice?

Yes

  1. Does the data collection use a third-party collection tool?

No

Result

data-review+

Attachment #9393567 - Flags: data-review?(tlong) → data-review+
Component: Firefox Accounts → Server: Firefox Accounts
Product: Firefox → Cloud Services

:markh raised the concern of sending raw device ids in Glean pings on the Github pull request. (Thank you, :markh!) I think that warrants a discussion, and this is probably the appropriate place for that.

I can think of three potential problems.

1. Mapping the devices to a specific user.

The device ids will be stored in a dataset with account identifiers. If an attacker manages to correlate an account identifier back to an actual Mozilla accounts user, then it does not make a difference whether the device ids are hashed, since the account to devices relationship is in the Glean data.

2. Mapping the devices to other datasets.

As far as I can tell, we currently do not use the raw device id in other telemetry. (Probably because we deemed it to be bad practice.) So this is not an issue at the moment, but...

3. Future risks.

If other telemetry datasets do include the raw device ids in the future, then #2 would be an issue. And if those datasets have identifiers that can be mapped back to specific users, then a reverse mapping of #1 will be an issue. To truly prevent these accidental mappings among datasets, a dataset-specific secret would be necessary in order to hash the device ids.

However, with all that being said, given the decision around the "Account ID", I wonder if that rationale can also be applied to the sync device id.

(In reply to Barry Chen from comment #3)

However, with all that being said, given the decision around the "Account ID", I wonder if that rationale can also be applied to the sync device id.

I think it probably can, and all sync pings are careful to not include any other identifiers - so to my mind this is just about consistency and cross-linking. I'd be fine with a decision that the sync pings should move to an unhashed version so long as everyone is fine with a temporary blip as we move - eg, any existing metrics which tries to track things per-device (eg, number of unique devices syncing per day) would be confused as any changes ride the trains and we wait for ESR to catch up.

Thanks for pointing this out that this is a distinct identifier, I was somewhat under the assumption that this was the FxA id when I initially reviewed this.

That being said, I still think this is likely technical information and given that this isn't a device identifier from a third party, but rather one we ourselves generate (correct me if I am wrong here).

In regards to the potential problems Barry mentions:

  1. Mapping devices to users

The identifiers we already make use of (client_id, nimbus_id, fxa_id) all already provide some ability to do this, but this is mitigated by the data itself being protected and only accessible to authorized folks, and through segregation of identifiers as Mark mentioned. An attacker may be able to get their hands on one or many of these identifiers but would also need access to the data that might be associated with it in order to compromise it.

  1. Mapping the devices to other datasets

There is some prior art around collection of hardware identifiers, the one that specifically comes to mind here is collecting the google hardware identifier in Android (hashed and isolated from client_id as mitigations).

  1. Future risks

I agree, there are some risks of linking this to data in such a way as to connect it to other datasets, but again, the access to the data is controlled and this is no different than being careful with identifiers we currently use, so it is the same risk for any unique identifier.

So, to move this forward, would you feel more comfortable if I were to escalate this review to get Trust, Legal, Security, etc. opinions on this?

Flags: needinfo?(markh)
Flags: needinfo?(bchen)

So, to move this forward, would you feel more comfortable if I were to escalate this review to get Trust, Legal, Security, etc. opinions on this?

I don't mind it if you think it's necessary. With the arguments laid out in the Account ID decision brief and input from you and Mark, I'm comfortable with moving forward using the raw sync device ids.

Flags: needinfo?(bchen)

Thanks Barry, I don't want to delay your work here if possible, but after considering this I think I'd feel more comfortable if I raised the question with Trust & Legal on whether or not this should be escalated.

I'll reach out to the right folks there now, and I'll get a more formal opinion on whether or not we need to escalate this or if it falls under the existing decision around the FxA identifier.

Flags: needinfo?(markh)

Comment on attachment 9393567 [details]
Answers to data collection review form

I've spoken with Nneka about this, and just to ensure that this gets proper documentation and awareness the escalation process should be followed. I've withdrawn my data-review+ pending escalation.

Please email data-review@mozilla.com to begin the escalation process, as outlined here under "Step 3: Sensitive Data Collection Review Process": wiki.mozilla.org/Data_Collection

Attachment #9393567 - Flags: data-review+ → data-review-

Please email data-review@mozilla.com to begin the escalation process, as outlined here under "Step 3: Sensitive Data Collection Review Process": wiki.mozilla.org/Data_Collection

Quick question regarding the escalation process. Since Step 3 of the documentation is for "data collection that is classified as category 3 or 4", is the escalation to argue that the data is not category 3 or 4, or, is it to argue for the inclusion of the data even though it is category 3 or 4? Thank you!

Flags: needinfo?(tlong)

The purpose of the escalation here, as communicated to me by Nneka, was to document this identifier as being able to be treated as "technical" data and inform Privacy and Trust&Legal so that they are aware it exists.

It is typical for new identifiers to undergo some additional scrutiny like this, and is part of the process that the data-stewards follow when reviewing data-collection requests for these to be called out.

Flags: needinfo?(tlong)
See Also: → 1893727
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: