Sloppy ICE Redactions

U.S. Immigration and Customs Enforcement, or ICE, is, allegedly, a law enforcement agency. While its operational sloppiness is widely documented, I haven’t seen much about its sloppy paperwork.

ICE partners with local law enforcement agencies using 287(g) memoranda of agreement.These are not the only agreements between ICE and local agencies. Details of these contracts are not important to this story, other than the fact that many of these agreements are scanned and made available to the public via links in a spreadsheet available from this page. Based on a handful of spot checks, most of the available MOAs are scanned PDFs. I typically think of a few things with scanned PDFs: poor legibility, giant file sizes, poor OCR, etc. More to the point, I do not think of fillable text forms, so I was rather surprised to spot several blank text fields scattered across one of the MOAs I was reading (original, archived).

This screenshot is the full document, with the fillable text fields colored red.
Screenshot of several pages of a PDF with red rectangles illustrating where fillable text fields are.

The first page contains a fillable field in the lower left that looks like it could be intended to hold a page number. That would be a strange way to insert page numbers, especially considering the original document included printed page numbers, but I’ve seen stranger hacks in the past, so I didn’t think much of it. The second page included two fields near the bottom of the page, one of which was significantly larger. Maybe that was supposed to contain a document name? File path? Some other sort of metadata? Page seven finally piqued my interest enough to go digging: it has several fields in and around the signature blocks, arranged in a way that could plausibly be hiding information.

The signature blocks as presented in the original PDF.
Screenshot of a scanned signature block, with some portions covered by opaque, fillable text fields.

Mozilla’s PDF.js library, used in Firefox, makes it remarkably easy to adjust how PDFs are displayed, since it renders PDFs just like any other page. After fiddling with the developer tools a little bit to identify how those fillable fields were displayed, I discovered some of why the seemingly random fields were present: hiding a single mistake in the signature block. The sheriff’s office in question is in Michigan, but the “Agency” line in the signature block mistakenly included “OH” (Ohio) instead. As far as I can see, that is the only piece of information being hidden (poorly) by the text fields. Perhaps the others mean something else, or are intended to be red herrings, or are simply the result of even more sloppy execution.

The signature blocks with the fillable text fields removed.
Screenshot of the same scanned signature block, with the fillable text fields missing, uncovering the incorrect 'OH'.

For the life of me, I cannot figure out why hiding the mistake with fillable fields seemed like a sensible solution. It’s a seemingly inconsequential mistake, and could have easily been struck-through and corrected by hand, or by an annotation on the PDF after it was scanned. I’ve heard stories of failed redactions leaking information in the past, but I never expected to come across such an issue in the wild. Sadly, it seems I stumbled across the most mundane occurrence possible. Ah, well.