On the Definition of Dragnet “Identifier”

Last month, I noted that ODNI failed to redact a reference to Verizon in one of the phone dragnet primary orders, which helped to confirm that Verizon was the provider ordered to provide only its domestic or one-end domestic call records to NSA under this order.

I’d like to look at another redaction fail (also, IIRC, pointed out to me Michael) from that document dump.

In the February 25, 2010 order, part of the footnote describing what identifiers NSA can use to contact chain was left unredacted.

Screen Shot 2014-02-15 at 12.42.04 PM

The footnote starts on the previous page; this is the end of the description (the big redaction below it modifies one of the terms in the list of terror groups associations).

Given all the discussion about whether NSA does or does not collect cell phone data, I think it of particular interest that IMSI and IMEI — two ways to identify cell phone users — appear in this footnote. It’s actually not clear whether their inclusions mean they can or cannot be used as identifiers.

But there’s reason to believe the footnote says they can be used as identifiers.

The footnote first appeared in the March 5, 2009 order — the first written after Judge Reggie Walton started trying to clean up the dragnet mess. Screen Shot 2014-02-15 at 1.01.28 PM

By that point, NSA had informed Walton that an additional querying tool had regularly accessed the 215 dragnet to perform analysis of certain identifiers.

If an analyst conducted research supported by [redacted] the analyst would receive a generic notification that NSA’s signals intelligence (“SIGINT”) databases contained one or more references to the telephone identifier in which the analyst was interested; a count of how many times the identifier was present in SIGINT databases; the dates of the first and last call events associated with the identifier; a count of how many other unique telephone identifiers had direct contact with the identifier that was the subject of the analyst’s research; the total number of calls made to or from the telephone identifier that was the subject of the analyst’s research; the ratio of the count of total calls to the count of unique contacts; and the amount of time it took to process the analyst’s query.

But this was before NSA explained it treated all correlated identifiers for a particular RAS-approved person as RAS-approved,

The end-to-end review revealed the fact that NSA’s practice of using correlated selectors to query the BR FISA metadata had not been fully described to the Court. A communications address or selector, is considered correlated with other communications addresses when each additional address is shown to identify the same communicant(s) as the original address.

Though it had provided some kind of description of this practice in an August 18, 2008 filing that almost certainly served as back-up for the August 19, 2008 order that first started specifically ordering IMSI and IMEI data.

A description of how [redacted] is used to correlate [redacted] was included in the government’s 18 August 2008 filing to the FISA Court, While NSA previously described to the FISC the ractice of using correlated selectors as seeds, the FISC never addressed whether [redacted] correlated selectors met the RAS standard when any one of the correlated selectors met the RAS standard. A notice was filed with the FISC can this issue on 15 June 2009.


All of which is to say that several of the items discussed during the 2009 review pertained to how NSA tracked identities over time, particularly phone-based identities that spanned multiple cell phones.

Which would explain why it would want to track both phone numbers themselves, but especially the handset and SIM identifiers (though in the case of burner phone “correlation,” those details wouldn’t help to make a match).

None of this should be surprising. As I said, it would be shocking if the nation’s counterterrorism professionals accepted a dragnet with less functionality than the one available to DEA under AT&T’s Hemisphere program, and a key part of that program involves matching cell phone identities (though remember, Hemisphere at least used to permit tracking of geolocation, too).

But assuming that footnote defining “identifier” affirmatively includes IMSI and IMEI as potential identifiers, which would seem logical, it’s yet one more data point showing how central the use of cell phones is to the dragnet.

That still doesn’t mean the NSA collected cell phone data, or collected it from providers besides AT&T and Sprint. But it sure seems to indicate an priority on such data.

Is Hemisphere Creating Problems for the Phone Dragnet?

Screen Shot 2014-02-12 at 4.39.40 PMYou are all probably bored with my repeated posts about why the claim that NSA only collects 30% of US data is probably only narrowly true.

So I won’t discuss how absurd it would be to argue that the terrorist dragnet drawing on the records of at least 3 phone companies was less comprehensive than Hemisphere, the similar AT&T-specific database it makes available to hunt drug crime.

I just want to raise a methodological issue.

In her declaration submitted in support of the suits challenging the Section 215 dragnet, Theresa Shea emphasized something implicit in the Business Records order: the telecoms are only turning over records they already have.

[P]ursuant to the FISC’s orders, telecommunications service providers turn over to the NSA business records that the companies already generate and maintain for their own pre-existing business purposes (such as billing and fraud prevention).

Presumably, AT&T provides precisely this same data to the NSA for its master phone dragnet. That is, to the extent that AT&T compiles this data in particular form, that may well be the form it hands onto NSA.

And that’s interesting for several reasons.

Hemisphere includes not just AT&T call records. It includes records from “CDRs for any telephone carrier that uses an AT&T switch to process a telephone call.” It gets 4 billion call records a day, including international ones and cell ones. As Scott Shane explained,

AT&T operates what are called switches, through which telephone calls travel all around the country. And what AT&T does in this program is it collects all the—what are called the CDRs, the call data records, the so-called metadata from the calls that we’ve heard about in the NSA context. This is the phone number—phone numbers involved in a call, its time, its duration, and in this case it’s also the location. Some are cellphone calls; some are land line calls. Anything that travels through an AT&T switch, even if it’s not made by an AT&T customer—for example, if you’re using your T-Mobile cellphone but your call travels through an AT&T switch somewhere in the country, it will be picked up by this project and dumped into this database.

Which supports the report from last summer that the government can get T-Mobile calls off AT&T’s records. These are the pre-existing records that NSA can come get and they include T-Mobile calls.

There’s another interesting part of that. As I noted the first two phone dragnet orders provided for compensation to the providers, even though the statute doesn’t envision that. That would bring you to November 2006; Hemisphere started in 2007, with funding from ONCDP, the White House Drug Czar. Remember, too, that FBI had the equivalent of Hemisphere onsite until late 2007-2008. That is, one thing Hemisphere does is pay for one provider to store what serves as a good baseline dragnet that can then be handed over to the NSA. That’s significant especially given Geoffrey Stone’s claims that the dragnet is not comprehensive because the cost involved: there should be no cost, but somehow it’s driving decisions.

In any case, as luck would have it, Hemisphere got exposed at the same time as the dragnet.

Hemisphere operates with different legal problems than the NSA phone dragnet. At least with the phone dragnet, after all, AT&T has been compelled to turn over records; with Hemisphere they’re effectively retaining them voluntarily to turn surveillance into a profit center (though they do get compelled on an order-by-order basis). Moreover, AT&T’s far more exposed by the publication on Hemisphere than it is on the NSA dragnet (or perhaps, than even Verizon is under the phone dragnet). The exposure of Hemisphere might make AT&T more hesitant to “voluntarily” retain this data.

Finally, there’as the amicus challenge EFF and ACLU submitted in a criminal case in Northern California notes, Hemisphere includes precisely the data the NSA is struggling with: cell location data.

Hemisphere goes even further than the NSA’s mass call-tracking program, as the CDRs stored in the Hemisphere database contain location information about callers (see Hemisphere Slide Deck at 3, 13), thus implicating the specific concerns raised by five Justices in Jones. See 132 S. Ct. at 955 (Sotomayor, J., concurring) (“wealth of detail about [a person’s] familial, political, professional, religious, and sexual associations” revealed through “trips to the psychiatrist, the plastic surgeon, the abortion clinic,” etc.) (internal quotation marks, citation omitted); id. at 964 (Alito, J., concurring).

The FISC has created all sorts of problems for NSA to store cell location data, most explicitly with Claire Eagan’s order in July specifically prohibiting it.

But here AT&T is, creating the opportunity for the perfect challenge to use Jones to challenge location in a dragnet specifically.

Which is all a way of saying that the tensions with the phone dragnet may not be entirely unrelated from the fact that Hemisphere also got challenged.