This week brought a timely revisit to remind myself just how common the three security extended attributes (xattr) have become, and to see whether we can make use of any of them for our own purposes.

How common?

Checking through one of my ~/Documents folders containing a modest 57,884 files, nearly 60% of them have at least one xattr. By far the most common is com.apple.quarantine on 48%, followed by com.apple.provenance at 14%. Some way behind those, but still one of the most frequent, is com.apple.macl on 2.8%.

Having explained what I think macOS does with all those xattrs, the next step is to ask whether we can use them for our direct benefit. Of the three, quarantine seems least useful to anything beyond Gatekeeper. MACL is like a boil on the bum, and the only time you’ll notice it is when it gets in your way. I can’t make sense from its contents either, but as it’s protected by SIP, there’s little a utility could do to alleviate our suffering, so we just have to learn to live with it. I’m surprised how uncommon it is in comparison with the nuisance it can cause.

Promising provenance

It’s the provenance xattr that looks most promising, and Koh M. Nakagawa has followed up his recent research into its function with an open-source command tool ShowProvenanceInfo that can look up provenance IDs found on files, in the ExecPolicy database’s Provenance Tracking table, although that requires root privileges for access.

Apps and executable code signed by third-parties rather than Apple are added to that table when they’ve successfully completed their first run. Each is given a unique provenance ID number that is attached to them in a com.apple.provenance xattr. When they then perform any of 11 types of file operation, such as creating a file or opening one in write mode, that app’s provenance ID is attached to the file in its own provenance xattr.

As apps and other executables that have been entered into that table have their own provenance xattrs, it shouldn’t be too much of a burden to build an independent database from those, together with other information about the executable with that provenance ID. That can then be used to examine provenance xattrs on arbitrary files, to identify which app last worked with that file, the primary task of a new GUI utility I’ve already dubbed Providable.

In addition to telling you which app with a provenance ID last changed a document, there are other functions that Providable could perform. Those property lists forming the basis for Background Items listed in Login Items & Extensions settings are normally created and changed by their owning app. When a third-party app with a provenance ID does that, the property list gains a provenance xattr that can be used to identify the app responsible. That can in turn provide information that’s often sadly lacking from the list of extensions, including the location of the app rather than that of the property list.

One obvious hole in this plan is the fact that apps that are signed by Apple, including those bundled in macOS and everything installed from the App Store, don’t get assigned provenance IDs. They therefore can’t be traced and identified from the files they create or change, as they operate outside the provenance tracking system.

Providable

My outline design for Providable is therefore to inspect provenance IDs saved as xattrs to the apps in the Data volume, and to display their details in a list you can refer to. Alongside that, a window lets you drop files on it for checking. Each will be examined for a provenance xattr, and those that have one will then be associated with the app in the list, providing its path and other details.

Provenance IDs can also be assigned to command tools and other executables, and a later version will allow you to check those in popular locations, and add them to the database so files can be matched to them as well. I don’t currently know how useful that might be, but we should get a better idea once the first version of Providable is in use.

I invite your ideas and comments, please, before I start coding.