John Bates

1 2 3 10

Another Probable Reason to Shut Down the Internet Dragnet: Dissemination Restrictions

Screen Shot 2015-11-27 at 10.27.12 PMI noted the other day that an NSA IG document liberated by Charlie Savage shows the agency had 4 reasons to shut down the domestic Internet (PRTT) dragnet, only one of which is the publicly admitted reason — that NSA could accomplish what it needed to using SPCMA and FAA collection.

I’m fairly sure another of the reasons NSA shut down the dragnet is because of dissemination restrictions that probably got newly reinvigorated in mid-2011.

I laid out a timeline of events leading up to the shutdown of the Internet dragnet here. I’ve added one date: that of the draft training program, several modules of which are dated October 17, 2011, released under FOIA (given other dates in the storyboard, the program had clearly been in development as early as November 2010). How odd is that? The NSA was just finalizing a training program on the Internet (and phone) dragnet as late as 6 weeks before NSA hastily shut it down starting in late November 2011. The training program — which clearly had significant Office of General Counsel involvement — provides a sense of what compliance issues OGC was emphasizing just as NSA decided to shut down the Internet dragnet.

The training program was done in the wake of two things: a series of audits mandated by the FISA Court (see PDF 36) that lasted from May 2010 until early 2011, and the resumption of the PRTT Internet dragnet between July and October 2010.

The series of audits revealed several things. First, as I have long argued was likely, the technical personnel who monitor the data for integrity may also use their access to make inappropriate queries, as happened in an incident during this period (see PDF 95 and following); I plan to return to this issue. In addition, at the beginning of the period — before a new selector tracking tool got introduced in June 2010 — NSA couldn’t track whether some US person selectors had gotten First Amendment review. And, throughout the audit period, the IG simply didn’t review whether less formalized disseminations of dragnet results followed the rules, because it was too hard to audit. The final report summarizing the series of audits from May 2011 (as well as the counterpart one covering the Internet dragnet) identified this as one of the weaknesses of the program, but NSA wanted to manage it by just asking FISC to eliminate the tracking requirements for foreign selectors (see PDF 209).

Screen Shot 2015-11-29 at 9.36.44 AM

I found this blasé attitude about dissemination remarkable given that in June 2009, Reggie Walton had gotten furious with NSA for not following dissemination restrictions, after which NSA did it again in September 2009, and didn’t tell Walton about it, which made him furious all over again. Dissemination restrictions were something Walton had made clear he cared about, and NSA IG’s response was simply to say auditing for precisely the kind of thing he was worried about — informal dissemination — was too hard, so they weren’t going to do it, not even for the audits FISC (probably Walton himself) ordered NSA to do to make sure they had cleaned up all the violations discovered in 2009.

Meanwhile, when NSA got John Bates to authorize the resumption of the dragnet (he signed the order in July 2010, but it appears it didn’t resume in earnest until October 2010), they got him to approve the dissemination of PRTT data broadly within NSA. This was a response to a Keith Alexander claim, made the year before, that all product lines within NSA might have a role in protecting against terrorism (see PDF 89).

Screen Shot 2015-11-29 at 10.00.59 AM

In other words, even as NSA’s IG was deciding it couldn’t audit for informal dissemination because it was too hard to do (even while acknowledging that was one of the control weaknesses of the program), NSA asked for and got FISC to expand dissemination, at least for the Internet dragnet, to basically everyone. (The two dragnets appear to have been synched again in October 2010, as they had been for much of 2009, and when that happened the NSA asked for all the expansions approved for the Internet dragnet to be applied to the phone dragnet.)

Which brings us to the training program.

There are elements of the training program that reflect the violations of the previous years, from an emphasis on reviewing for access restrictions to a warning that tech personnel should only use their sysadmin access to raw data for technical purposes, and not analytical ones.

But the overwhelming emphasis in the training was on dissemination — which is a big part of the reason the NSA used the program to train analysts to rerun PATRIOT-authorized queries under EO 12333 so as to bypass dissemination restrictions. As noted in the screen capture above, the training program gave a detailed list of the things that amounted to dissemination, including oral confirmation that two identifiers — even by name (which of course confirms that these phone numbers are identifiable to analysts) — were in contact.

In addition, any summary of that information would also be a BR or PR/TT query result. So, if you knew that identifier A belonged to Joe and identifier B belonged to Sam, and the fact of that contact was derived from BR or PR/TT metadata, if you communicate orally or in writing that Joe talked to Sam, even if you don’t include the actual e-mail account or telephone numbers that were used to communicate, this is still a BR or PR/TT query result.

The program reminded that NSA has to report every dissemination, no matter how informal.

This refers to information disseminated in a formal report as well as information disseminated informally such as written or oral collaboration with the FBI. We need to count every instance in which we take a piece of information derived from either of these two authorities and disseminate it outside of NSA.

Normally an NSA product report is the record of a formal dissemination. In the context of the BR and PR/TT Programs, an official RFI response or Analyst Collaboration Record will also be viewed as dissemination. Because this FISC requirement goes beyond the more standard NSA procedures, additional diligence must be given to this requirement. NSA is required to report disseminations formal or informal to the FISC every 30 days.

I’m most interested in two other aspects of the training. First, it notes that not all queries obtained via the dragnet will be terrorism related.

It might seem as though the information would most certainly be counterterrorism-related since, due to the RAS approval process, you wouldn’t have this U.S. person information from a query of BR or PR/TT if it weren’t related to counterterrorism. In the majority of cases, it will be counterterrorism-related; however, the nature of the counterterrorism target is that it often overlaps with several other areas that include counternarcotics, counterintelligence, money laundering, document forging, people and weapons trafficking, and other topics that are not CT-centric. Thus, due to the fact that these authorities provide NSA access to a high volume of U.S. person information for counterterrorism purposes, the Court Order requires an explicit finding that the information is in fact related to counterterrorism prior to dissemination. Therefore, one of the approved decision makers must document the finding using the proper terminology. It must state that the information is related to counterterrorism and that it is necessary to understand the counterterrorism information.

Remember, this training was drafted in the wake of NSA’s insistence that all these functional areas needed to be able to receive Internet dragnet data, which, of course, was just inviting the dissemination of information for reasons other than terrorism, especially given FISC’s permission to use the dragnet to track Iranian “terrorism.” Indeed, I still think think it overwhelmingly likely Shantia Hassanshahi got busted for proliferation charges using the phone dragnet (during a period when FISC was again not monitoring NSA very closely). And one of the things NSA felt the need to emphasize a year or so after NSA started being able to share this “counterterrorism” information outside of its counterterrorism unit was that they couldn’t share information about money laundering or drug dealing or … counterproliferation unless there was a counterterrorism aspect to it. Almost as if it had proven to be a problem.

The training program warns that results may not be put into queriable tools that untrained analysts have access to.

Screen Shot 2015-11-29 at 1.54.44 PM\

Note the absolutely hysterical review comment that said there’s no list of which tools analysts couldn’t use with 215 and PRTT dragnet results. Elsewhere, the training module instructs analysts to ask their manager, which from a process standpoint is a virtual guarantee there will be process violations.

This is interesting for two reasons. First, it suggests NSA was still getting in trouble running tools they hadn’t cleared with FISC (the 215 IG Reports also make it clear they were querying the full database using more than just the contact-chaining they claim to have been limited to). Remember there were things like a correlations tool they had to shut down in 2009.

But it’s also interesting given the approval, a year after this point, of an automatic alert system for use with the phone dragnet (which presumably was meant to replace the illegal alert system identified in 2009).

In 2012, the FISA court approved a new and automated method of performing queries, one that is associated with a new infrastructure implemented by the NSA to process its calling records.68 The essence of this new process is that, instead of waiting for individual analysts to perform manual queries of particular selection terms that have been RAS approved, the NSA’s database periodically performs queries on all RAS-approved seed terms, up to three hops away from the approved seeds. The database places the results of these queries together in a repository called the “corporate store.”

The ultimate result of the automated query process is a repository, the corporate store, containing the records of all telephone calls that are within three “hops” of every currently approved selection term.69 Authorized analysts looking to conduct intelligence analysis may then use the records in the corporate store, instead of searching the full repository of records.70

That is, in 2011, NSA was moving towards such an automated system, which would constitute a kind of dissemination by itself. But it wasn’t there yet for the PATRIOT authorized collection. Presumably it was for EO 12333 collection.

As it happened, NSA never did fulfill whatever requirements FISC imposed for using that automatic system with phone dragnet information, and they gave up trying in February 2014 when Obama decided to outsource the dragnet to the telecoms. But it would seem limits on the permission to use other fancy tools because they would amount to dissemination would likely limit the efficacy of these dragnets.

Clearly, in the weeks before NSA decided to shut down the PRTT dragnet, its lawyers were working hard to keep the agency in compliance with rules on dissemination. Then, they stopped trying and shut it down.

Both the replacement of PRTT with SPCMA and 702, and the replacement of the 215 dragnet with USAF, permit the government to disseminate metadata with far looser restrictions (and almost none, in the case of 702 and USAF metadata). It’s highly likely this was one reason the NSA was willing to shut them down.

2011 Internet Dragnet Audit Didn’t Find Significant Violation Reported to IOB

This will be the second of three posts on the NSA IG’s failures to correct problems with the Internet (PRTT) dragnet. In the first, I showed how quickly NSA nuked the PRTT (or at least claimed to) after John Bates ruled, a second time, that NSA could not illegally wiretap the content of Americans’ communications. Here, I’ll examine another IG Report, completed earlier in 2011 and also liberated by Charlie Savage, that appears to show the PRTT dragnet was hunky dory just weeks before it became clear again that it was not.

The report (see PDF 4-23) must date to between March 15 and May 25, 2011. It was related to a series of reports on the phone dragnet (these reports appear to have been solicited by or encouraged by Reggie Walton in the wake of the 2009 dragnet problems) that Savage liberated earlier this year. It lists all those reports on pages A-2 to A-3. But it lists the final, summary report in that series, (ST-10-0004L), as a draft, dated March 15, 2011. The copy provided to Savage is the final, dated May 25, 2011 (see PDF 203).

The reason for doing this, the PRTT report, is curious. The report notes “we began this review in [redacted, would be some time in summer 2009] but suspended it when NSA allowed the PR/TT Order to expire.” That is, this was the report that got started, but then halted, when someone discovered that every single record the NSA had collected under the program included categories of information violating the rules set by FISC in 2004.

But then NSA started a review of the phone dragnet covering all the activity in 2010 (reflected in monthly reports in Savage’s earlier release). So the NSA decided to do a review of PRTT at the same time. But remember: the Internet dragnet was shut down until at least July 2010, when John Bates authorized its resumption, and it took some time to turn the dragnet back on. That means NSA conducted a review of a dragnet that was largely on hiatus or just resuming. During the review period, both the phone and Internet dragnet reflect few finalized reports based on either dragnet. Indeed, it appears likely that there were no phone dragnet disseminations in August 2010 (see 155). There are probably two explanations for that. It suggests that after Reggie Walton told NSA they had to start following the rules, the amount of intelligence they got from the dragnet appears to have gone down from both the phone and Internet dragnet. But there may be a reason for that: we know that in 2011 NSA was training analysts to re-run queries that came up in both FISA and EO 12333 searches using EO 12333, so the results could be disseminated more broadly. So it’s likely that a lot of what had been reports reporting FISA authorized data before 2009 (which didn’t always follow FISC’s rules) started getting disseminated as EO 12333 authorized reports afterward. Still, in the case of the Internet dragnet reviewed for this report, “the dissemination did not contain PR/TT-derived USP information” so they “did not formally test dissemination objectives” (see footnote 1). None of the reports on the US Internet dragnet reviewed in some period in 2010 included US person data.

So much for collecting all of Americans’ email records to catch Americans, I guess.

All that said, both the Internet and phone dragnet found that the dragnets had adequate controls to fulfill the requirements of the FISC orders, but did say (this is laid out in unredacted form more explicitly in the phone dragnet report) that the manual monitoring of dissemination would become unworkable if analysts started using the dragnet more. The phone dragnet reports also suggest they weren’t good at monitoring less formal disseminations (via email or conversation), and by the time of these summary reports, NSA was preparing ask FISC to change the rules on reporting of non-US person dissemination. Overall in spring 2011, NSA’s IG found, the process worked according to the rules, but in part only because it was so little used.

That’s the assessment of the PRTT dragnet as of sometime between March and May 2011, less than 9 months before they’d nuke the dragnet really quickly, based mostly off a review of what NSA was doing during a period when the dragnet was largely inactive.

Which is all very interesting, because sometime before June 30, 2011 there was a PRTT violation that got reported — in a far more extensive description than the actual shut down of the dragnet in 2009 — to Intelligence Oversight Board. (see PDF 10)

Screen shot 2015-11-21 at 12.55.36 PM

There’s no mention of reporting to Congress on this, which is interesting because PATRIOT Act was being reauthorized again during precisely this period, based off notice, dated February 2, 2011, that the compliance problems were largely solved.

So here’s what happened: After having had its IG investigation shut down in fall 2009 because NSA had never been in compliance with limits on the PRTT dragnet, NSA’s IG tried again during a period when the NSA wasn’t using it all that much. It gave NSA a clean bill of health no earlier than March 15, 2011. But by June 30, 2011, something significant enough to get reported in two full paragraphs to IOB happened.

It turns out things weren’t quote so hunky dory after all.

The NSA (Said It) Ate Its Illegal Domestic Content Homework before Having to Turn It in to John Bates

The question of whether NSA can keep its Section 215 dragnet data past November 28 has been fully briefed for at least 10 days, but Judge Michael Mosman has not yet decided whether the NSA can keep it — at least not publicly. But given what the NSA IG Report on NSA’s destruction of the Internet dragnet says (liberated by Charlie Savage and available starting on PDF 60), we should assume the NSA may be hanging onto that data anyway.

This IG Report documents NSA’s very hasty decision to shut down the Internet dragnet and destroy all the data associated with it at the end of 2011, in the wake of John Bates’ October 3, 2011 opinion finding, for the second time, that if NSA knew it had collected US person content, it would be guilty of illegal wiretapping. And even with the redactions, it’s clear the IG isn’t entirely certain NSA really destroyed all those records.

The report adds yet more evidence to support the theory that the NSA shut down the PRTT program because it recognized it amounted to illegal wiretapping. The evidence to support that claim is laid out in the timeline and working notes below.

The report tells how, in early 2011, NSA started assessing whether the Internet dragnet was worth keeping under the form John Bates had approved in July 2010, which was more comprehensive and permissive than what got shut down around October 30, 2009. NSA would have had SPCMA running in big analytical departments by then, plus FAA, so they would have been obtaining these benefits over the PRTT dragnet already. Then, on a date that remains redacted, the Signals Intelligence Division asked to end the dragnet and destroy all the data. That date has to post-date September 10, 2011 (that’s roughly when the last dragnet order was approved), because SID was advising to not renew the order, meaning it happened entirely during the last authorization period. Given the redaction length it’s likely to be October (it appears too short to be September), but could be anytime before November 10. [Update: As late as October 17, SID was still working on a training program that covered PRTT, in addition to BRFISA, so it presumably post-dates that date.] That means that decision happened at virtually the same time or after, but not long after, John Bates raised the problem of wiretapping violations under FISA Section 1809(a)(2) again on October 3, 2011, just 15 months after having warned NSA about Section 1809(a)(2) violations with the PRTT dragnet.

The report explains why SID wanted to end the dragnet, though three of four explanations are redacted. If we assume bullets would be prioritized, the reason we’ve been given — that NSA could do what it needed to do with SPCMA and FAA — is only the third most important reason. The IG puts what seems like a non sequitur in the middle of that paragraph. “In addition, notwithstanding restrictions stemming from the FISC’s recent concerns regarding upstream collection, FAA §702 has emerged as another critical source for collection of Internet communications of foreign terrorists” (which seems to further support that the decision post-dated that ruling). Indeed, this is not only a non sequitur, it’s crazy. Everyone already knew FAA was useful. Which suggests it may not be a non sequitur at all, but instead something that follows off of the redacted discussions.

Given the length of the redacted date (it is one character longer than “9 December 2011”), we can say with some confidence that Keith Alexander approved the end and destruction of the dragnet between November 10 and 30 — during the same period the government was considering appealing Bates’ ruling, close to the day — November 22 — NSA submitted a motion arguing that Section 1809(a)(2)’s wiretapping rules don’t apply to it, and the day, a week later, it told John Bates it could not segregate the pre-October 31 dragnet data from post October 31 dragnet data.

Think how busy a time this already was for the legal and tech people, given the scramble to keep upstream 702 approved! And yet, at precisely the same time, they decided they should nuke the dragnet, and nuke it immediately, before the existing dragnet order expired, creating another headache for the legal and tech people. My apologies to the people who missed Thanksgiving dinner in 2011 dealing with both these headaches at once.

Not only did NSA nuke the dragnet, but they did it quickly. As I said, it appears Alexander approved nuking it November 10 or later. By December 9, it was gone.

At least, it was gone as far as the IG can tell. As far as the 5 parts of the dragnet (which appear to be the analyst facing side) that the technical repository people handled, that process started on December 2, with the IG reviewing the “before” state, and ended mostly on December 7, with final confirmation happening on December 9, the day NSA would otherwise have had to have new approval of the dragnet. As to the the intake side, those folks started destroying the dragnet before the IG could come by and check their before status:

However, S3 had completed its purge before we had the opportunity to observe. As a result we were able to review the [data acquisition database] purge procedures only for reasonableness; we were not able to do the before and after comparisons that we did for the TD systems and databases disclosed to us.

Poof! All gone, before the IG can even come over and take a look at what they actually had.

Importantly, the IG stresses that his team doesn’t have a way of proving the dragnet isn’t hidden somewhere in NSA’s servers.

It is important to note that we lack the necessary system accesses and technical resources to search NSA’s networks to independently verify that only the disclosed repositories stored PR/TT metadata.

That’s probably why the IG repeatedly says he is confirming purging of the data from all the “disclosed” databases (@nailbomb3 observed this point last night). Perhaps he’s just being lawyerly by including that caveat. Perhaps he remembers how he discovered in 2009 that every single record the NSA had received over the five year life of the dragnet had violated Colleen Kollar-Kotelly’s orders, even in spite of 25 spot checks. Perhaps the redacted explanations for eliminating the dragnet explain the urgency, and therefore raise some concerns. Perhaps he just rightly believes that when people don’t let you check their work — as NSA did not by refusing him access to NSA’s systems generally — there’s more likelihood of hanky panky.

But when NSA tells — say — the EFF, which was already several years into a lawsuit against the NSA for illegal collection of US person content from telecom switches, and which already had a 4- year old protection order covering the data relevant to that suit, that this data got purged in 2011?

Even NSA’s IG says he thinks it did but he can’t be sure.

But what we can be sure of is, after John Bates gave NSA a second warning that he would hold them responsible for wiretapping if they kept illegally collecting US person content, the entire Internet dragnet got nuked within 70 days — gone!!! — all before anyone would have to check in with John Bates again in connection with the December 9 reauthorization and tell him what was going on with the Internet dragnet.

Update: Added clarification language.

Update: The Q2 2011 IOB report (reporting on the period through June 30, 2011) shows a 2-paragraph long, entirely redacted violation (PDF 10), which represents a probably more substantive discussion than the systematic overcollection that shut down the system in 2009.

Continue reading

The Reasons to Shut Down the (Domestic) Internet Dragnet: Purpose and Dissemination Limits, Correlations, and Functionality

Charlie Savage has a story that confirms (he linked some of my earlier reporting) something I’ve long argued: NSA was willing to shut down the Internet dragnet in 2011 because it could do what it wanted using other authorities. In it, Savage points to an NSA IG Report on its purge of the PRTT data that he obtained via FOIA. The document includes four reasons the government shut the program down, just one of which was declassified (I’ll explain what is probably one of the still-classified reasons probably in a later post). It states that SPCMA and Section 702 can fulfill the requirements that the Internet dragnet was designed to meet. The government had made (and I had noted) a similar statement in a different FOIA for PRTT materials in 2014, though this passage makes it even more clear that SPCMA — DOD’s self-authorization to conduct analysis including US persons on data collected overseas — is what made the switch possible.

It’s actually clear there are several reasons why the current plan is better for the government than the previous dragnet, in ways that are instructive for the phone dragnet, both retrospectively for the USA F-ReDux debate and prospectively as hawks like Tom Cotton and Jeb Bush and Richard Burr try to resuscitate an expanded phone dragnet. Those are:

  • Purpose and dissemination limits
  • Correlations
  • Functionality

Purpose and dissemination limits

Both the domestic Internet and phone dragnet limited their use to counterterrorism. While I believe the Internet dragnet limits were not as stringent as the phone ones (at least in pre 2009 shutdown incarnation), they both required that the information only be disseminated for a counterterrorism purpose. The phone dragnet, at least, required someone sign off that’s why information from the dragnet was being disseminated.

Admittedly, when the FISC approved the use of the phone dragnet to target Iran, it was effectively authorizing its use for a counterproliferation purpose. But the government’s stated admissions — which are almost certainly not true — in the Shantia Hassanshahi case suggest the government would still pretend it was not using the phone dragnet for counterproliferation purposes. The government now claims it busted Iranian-American Hassanshahi for proliferating with Iran using a DEA database rather than the NSA one that technically would have permitted the search but not the dissemination, and yesterday Judge Rudolph Contreras ruled that was all kosher.

But as I noted in this SPCMA piece, the only requirement for accessing EO 12333 data to track Americans is a foreign intelligence purpose.

Additionally, in what would have been true from the start but was made clear in the roll-out, NSA could use this contact chaining for any foreign intelligence purpose. Unlike the PATRIOT-authorized dragnets, it wasn’t limited to al Qaeda and Iranian targets. NSA required only a valid foreign intelligence justification for using this data for analysis.

The primary new responsibility is the requirement:

  • to enter a foreign intelligence (FI) justification for making a query or starting a chain,[emphasis original]

Now, I don’t know whether or not NSA rolled out this program because of problems with the phone and Internet dragnets. But one source of the phone dragnet problems, at least, is that NSA integrated the PATRIOT-collected data with the EO 12333 collected data and applied the protections for the latter authorities to both (particularly with regards to dissemination). NSA basically just dumped the PATRIOT-authorized data in with EO 12333 data and treated it as such. Rolling out SPCMA would allow NSA to use US person data in a dragnet that met the less-restrictive minimization procedures.

That means the government can do chaining under SPCMA for terrorism, counterproliferation, Chinese spying, cyber, or counter-narcotic purposes, among others. I would bet quite a lot of money that when the government “shut down” the DEA dragnet in 2013, they made access rules to SPCMA chaining still more liberal, which is great for the DEA because SPCMA did far more than the DEA dragnet anyway.

So one thing that happened with the Internet dragnet is that it had initial limits on purpose and who could access it. Along the way, NSA cheated those open, by arguing that people in different function areas (like drug trafficking and hacking) might need to help out on counterterrorism. By the end, though, NSA surely realized it loved this dragnet approach and wanted to apply it to all NSA’s functional areas. A key part of the FISC’s decision that such dragnets were appropriate is the special need posed by counterterrorism; while I think they might well buy off on drug trafficking and counterproliferation and hacking and Chinese spying as other special needs, they had not done so before.

The other thing that happened is that, starting in 2008, the government started putting FBI in a more central role in this process, meaning FBI’s promiscuous sharing rules would apply to anything FBI touched first. That came with two benefits. First, the FBI can do back door searches on 702 data (NSA’s ability to do so is much more limited), and it does so even at the assessment level. This basically puts data collected under the guise of foreign intelligence at the fingertips of FBI Agents even when they’re just searching for informants or doing other pre-investigative things.

In addition, the minimization procedures permit the FBI (and CIA) to copy entire metadata databases.

FBI can “transfer some or all such metadata to other FBI electronic and data storage systems,” which seems to broaden access to it still further.

Users authorized to access FBI electronic and data storage systems that contain “metadata” may query such systems to find, extract, and analyze “metadata” pertaining to communications. The FBI may also use such metadata to analyze communications and may upload or transfer some or all such metadata to other FBI electronic and data storage systems for authorized foreign intelligence or law enforcement purposes.

In this same passage, the definition of metadata is curious.

For purposes of these procedures, “metadata” is dialing, routing, addressing, or signaling information associated with a communication, but does not include information concerning the substance, purport, or meaning of the communication.

I assume this uses the very broad definition John Bates rubber stamped in 2010, which included some kinds of content. Furthermore, the SMPs elsewhere tell us they’re pulling photographs (and, presumably, videos and the like). All those will also have metadata which, so long as it is not the meaning of a communication, presumably could be tracked as well (and I’m very curious whether FBI treats location data as metadata as well).

Whereas under the old Internet dragnet the data had to stay at NSA, this basically lets FBI copy entire swaths of metadata and integrate it into their existing databases. And, as noted, the definition of metadata may well be broader than even the broadened categories approved by John Bates in 2010 when he restarted the dragnet.

So one big improvement between the old domestic Internet dragnet and SPCMA (and 702 to a lesser degree, and I of course, improvement from a dragnet-loving perspective) is that the government can use it for any foreign intelligence purpose.

At several times during the USA F-ReDux debate, surveillance hawks tried to use the “reform” to expand the acceptable uses of the dragnet. I believe controls on the new system will be looser (especially with regards to emergency searches), but it is, ostensibly at least, limited to counterterrorism.

One way USA F-ReDux will be far more liberal, however, is in dissemination. It’s quite clear that the data returned from queries will go (at least) to FBI, as well as NSA, which means FBI will serve as a means to disseminate it promiscuously from there.


Another thing replacing the Internet dragnet with 702 access does it provide another way to correlate multiple identities, which is critically important when you’re trying to map networks and track all the communication happening within one. Under 702, the government can obtain not just Internet “call records” and the content of that Internet communication from providers, but also the kinds of thing they would obtain with a subpoena (and probably far more). As I’ve shown, here are the kinds of things you’d almost certainly get from Google (because that’s what you get with a few subpoenas) under 702 that you’d have to correlate using algorithms under the old Internet dragnet.

  • a primary gmail account
  • two secondary gmail accounts
  • a second name tied to one of those gmail accounts
  • a backup email (Yahoo) address
  • a backup phone (unknown provider) account
  • Google phone number
  • Google SMS number
  • a primary login IP
  • 4 other IP logins they were tracking
  • 3 credit card accounts
  • Respectively 40, 5, and 11 Google services tied to the primary and two secondary Google accounts, much of which would be treated as separate, correlated identifiers

Every single one of these data points provides a potentially new identity that the government can track on, whereas the old dragnet might only provide an email and IP address associated with one communication. The NSA has a great deal of ability to correlate those individual identifiers, but — as I suspect the Paris attack probably shows — that process can be thwarted somewhat by very good operational security (and by using providers, like Telegram, that won’t be as accessible to NSA collection).

This is an area where the new phone dragnet will be significantly better than the existing phone dragnet, which returns IMSI, IMEI, phone number, and a few other identifiers. But under the new system, providers will be asked to identify “connected” identities, which has some limits, but will nonetheless pull some of the same kind of data that would come back in a subpoena.


While replacing the domestic Internet dragnet with SPCMA provides additional data with which to do correlations, much of that might fall under the category of additional functionality. There are two obvious things that distinguish the old Internet dragnet from what NSA can do under SPCMA, though really the possibilities are endless.

The first of those is content scraping. As the Intercept recently described in a piece on the breathtaking extent of metadata collection, the NSA (and GCHQ) will scrape content for metadata, in addition to collecting metadata directly in transit. This will get you to different kinds of connection data. And particularly in the wake of John Bates’ October 3, 2011 opinion on upstream collection, doing so as part of a domestic dragnet would be prohibitive.

In addition, it’s clear that at least some of the experimental implementations on geolocation incorporated SPCMA data.

I’m particularly interested that one of NSA’s pilot co-traveler programs, CHALKFUN, works with SPCMA.

Chalkfun’s Co-Travel analytic computes the date, time, and network location of a mobile phone over a given time period, and then looks for other mobile phones that were seen in the same network locations around a one hour time window. When a selector was seen at the same location (e.g., VLR) during the time window, the algorithm will reduce processing time by choosing a few events to match over the time period. Chalkfun is SPCMA enabled1.

1 (S//SI//REL) SPCMA enables the analytic to chain “from,” “through,” or “to” communications metadata fields without regard to the nationality or location of the communicants, and users may view those same communications metadata fields in an unmasked form. [my emphasis]

Now, aside from what this says about the dragnet database generally (because this makes it clear there is location data in the EO 12333 data available under SPCMA, though that was already clear), it makes it clear there is a way to geolocate US persons — because the entire point of SPCMA is to be able to analyze data including US persons, without even any limits on their location (meaning they could be in the US).

That means, in addition to tracking who emails and talks with whom, SPCMA has permitted (and probably still does) permit NSA to track who is traveling with whom using location data.

Finally, one thing we know SPCMA allows is tracking on cookies. I’m of mixed opinion on whether the domestic Internet ever permitted this, but tracking cookies is not only nice for understanding someone’s browsing history, it’s probably critical for tracking who is hanging out in Internet forums, which is obviously key (or at least used to be) to tracking aspiring terrorists.

Most of these things shouldn’t be available via the new phone dragnet — indeed, the House explicitly prohibited not just the return of location data, but the use of it by providers to do analysis to find new identifiers (though that is something AT&T does now under Hemisphere). But I would suspect NSA either already plans or will decide to use things like Supercookies in the years ahead, and that’s clearly something Verizon, at least, does keep in the course of doing business.

All of which is to say it’s not just that the domestic Internet dragnet wasn’t all that useful in its current form (which is also true of the phone dragnet in its current form now), it’s also that the alternatives provided far more than the domestic Internet did.

Jim Comey recently said he expects to get more information under the new dragnet — and the apparent addition of another provider already suggests that the government will get more kinds of data (including all cell calls) from more kinds of providers (including VOIP). But there are also probably some functionalities that will work far better under the new system. When the hawks say they want a return of the dragnet, they actually want both things: mandates on providers to obtain richer data, but also the inclusion of all Americans.

The FISA Court’s Uncelebrated Good Points

I’m working on a post responding to this post from Chelsea Manning calling to abolish the FISA Court. Spoiler alert: I largely agree with her, but I think the question is not that simple.

As background to that post, I wanted to shift the focus from a common perception of the FISC — that it is a rubber stamp that approves all requests — to a better measure of the FISC — the multiple ways it has tried to rein in the Executive. I think the FISC has, at times, been better at doing so than often given credit for. But as I’ll show in my larger post, those efforts have had limited success.

Minimization procedures

The primary tool the FISC uses is in policing the Executive is minimization procedures approved by the court. Royce Lamberth unsuccessfully tried to use minimization procedures to limit the use of FISA-collected data in prosecutions (and also, tools for investigation, such as informants). Reggie Walton was far more successful at using and expanding very detailed limits on the phone — and later, the Internet — dragnet to force the government to stop treating domestically collected dragnet data under its own EO 12333 rules and start treating it under the more stringent FISC-imposed rules. He even shut down the Internet dragnet in fall (probably October 30) 2009 because it did not abide by limits imposed 5 years earlier by Colleen Kollar-Kotelly.

There was also a long-running discussion (that involved several briefs in 2006 and 2009, and a change in FISC procedure in 2010) about what to do with Post Cut Through Dialed Digits (those things you type in after a call or Internet session has been connected) collected under pen registers. It appears that FISC permitted (and probably still permits) the collection of that data under FISA (that was not permitted under Title III pen registers), but required the data get minimized afterwards, and for a period over collected data got sequestered.

Perhaps the most important use of minimization procedures, however, came when Internet companies stopped complying with NSLs requiring data in 2009, forcing the government to use Section 215 orders to obtain the data. By all appearances, the FISC imposed and reviewed compliance of minimization procedures until FBI, more than 7 years after being required to, finally adopted minimization procedures for Section 215. This surely resulted in a lot less innocent person data being collected and retained than under NSL collection. Note that this probably imposed a higher standard of review on this bulky collection of data than what existed at magistrate courts, though some magistrates started trying to impose what are probably similar requirements in 2014.

Such oversight provides one place where USA Freedom Act is a clear regression from what is (today, anyway) in place. Under current rules, when the government submits an application retroactively for an emergency search of the dragnet, the court can require the government to destroy any data that should not have been collected. Under USAF, the Attorney General will police such things under a scheme that does not envision destroying improperly collected data at all, and even invites the parallel construction of it.

First Amendment review

The FISC has also had some amount — perhaps significant — success in making the Executive use a more restrictive First Amendment review than it otherwise would have. Kollar-Kotelly independently imposed a First Amendment review on the Internet dragnet in 2004. First Amendment reviews were implicated in the phone dragnet changes Walton pushed in 2009. And it appears that in the government’s first uses of the emergency provision for the phone dragnet, it may have bypassed First Amendment review — at least, that’s the most logical explanation for why FISC explicitly added a First Amendment review to the emergency provision last year. While I can’t prove this with available data, I strongly suspect more stringent First Amendment reviews explain the drop in dragnet searches every time the FISC increased its scrutiny of selectors.

In most FISA surveillance, there is supposed to be a prohibition on targeting someone for their First Amendment protected activities. Yet given the number of times FISC has had to police that, it seems that the Executive uses a much weaker standard of First Amendment review than the FISC. Which should be a particularly big concern for National Security Letters, as they ordinarily get no court review (one of the NSL challenges that has been dismissed seemed to raise First Amendment concerns).

Notice of magistrate decisions

On at least two occasions, the FISC has taken notice of and required briefing after magistrate judges found a practice also used under FISA to require a higher standard of evidence. One was the 2009 PCTDD discussion mentioned above. The other was the use of combined orders to get phone records and location data. And while the latter probably resulted in other ways the Executive could use FISA to obtain location data, it suggests the FISC has paid close attention to issues being debated in magistrate courts (though that may have more to do with the integrity of then National Security Assistant Attorney General David Kris than the FISC itself; I don’t have high confidence it is still happening). To the extent this occurs, it is more likely that FISA practices will all adjust to new standards of technology than traditional courts, given that other magistrates will continue to approve questionable orders and warrants long after a few individually object, and given that an individual objection isn’t always made public.

Dissemination limits

Finally, the FISC has limited Executive action by limiting the use and dissemination of certain kinds of information. During Stellar Wind, Lamberth and Kollar-Kotelly attempted to limit or at least know which data came from Stellar Wind, thereby limiting its use for further FISA warrants (though it’s not clear how successful that was). The known details of dragnet minimization procedures included limits on dissemination (which were routinely violated until the FISC expanded them).

More recently John Bates twice pointed to FISA Section 1809(a)(2) to limit the government’s use of data collected outside of legal guidelines. He did so first in 2010 when he limited the government’s use of illegally collected Internet metadata. He used it again in 2011 when he used it to limit the government’s access to illegally collected upstream content. However, I think it likely that after both instances, the NSA took its toys and went elsewhere for part of the relevant collection, in the first case to SPCMA analysis on EO 12333 collected Internet metadata, and in the second to CISA (though just for cyber applications). So long as the FISC unquestioningly accepts EO 12333 evidence to support individual warrants and programmatic certificates, the government can always move collection away from FISC review.

Moreover, with USAF, Congress partly eliminated this tool as a retroactive control on upstream collection; it authorized the use of data collected improperly if the FISC subsequently approved retention of it under new minimization procedures.

These tools have been of varying degrees of usefulness. But FISC has tried to wield them, often in places where all but a few Title III courts were not making similar efforts. Indeed, there are a few collection practices where the FISC probably imposed a higher standard than TIII courts, and probably many more where FISC review reined in collection that didn’t have such review.

Is CISA the Upstream Cyber Certificate NSA Wanted But Didn’t Really Get?

I’ve been wracking my brain to understand why the Intel Community has been pushing CISA so aggressively.

I get why the Chamber of Commerce is pushing it: because it sets up a regime under which businesses will get broad regulatory immunity in exchange for voluntarily sharing their customers’ data, even if they’re utterly negligent from a security standpoint, while also making it less likely that information their customers could use to sue them would become public. For the companies, it’s about sharply curtailing the risk of (charitably) having imperfect network security or (more realistically, in some cases) being outright negligent. CISA will minimize some of the business costs of operating in an insecure environment.

But why — given that it makes it more likely businesses will wallow in negligence — is the IC so determined to have it, especially when generalized sharing of cyber threat signatures has proven ineffective in preventing attacks, and when there are far more urgent things the IC should be doing to protect themselves and the country?

Richard Burr and Dianne Feinstein’s move the other day to — in the guise of ensuring DHS get to continue to scrub data on intake, instead give the rest of the IC veto power over that scrub (which almost certainly means the bill is substantially a means of eliminating the privacy role DHS currently plays) — leads me to believe the IC plans to use this as they might have used (or might be using) a cyber certification under upstream 702.

Other accounts of upstream 702 and CISA don’t account for John Bates’ 2011 ruling

Since NYT and ProPublica caught up to my much earlier reporting on the use of upstream 702 for cyber, people have long assumed that CISA would work with upstream 702 authority to magnify the way upstream 702 works. Jonathan Mayer described how this might work.

This understanding of the NSA’s domestic cybersecurity authority leads to, in my view, a more persuasive set of privacy objections. Information sharing legislation would create a concerning surveillance dividend for the agency.

Because this flow of information is indirect, it prevents businesses from acting as privacy gatekeepers. Even if firms carefully screen personal information out of their threat reports, the NSA can nevertheless intercept that information on the Internet backbone.

Note that Mayer’s model assumes the Googles and Verizons of the world make an effort to strip private information, then NSA would use the signature turned over to the government under CISA to go get the private information just stripped out. But Mayer’s model — and the ProPublica/NYT story — never considered how the 2011 John Bates ruling on upstream collection might hinder that model, particularly as it pertains to domestically collected data.

As I laid out back in June, NSA’s optimistic predictions they’d soon get an upstream 702 certificate for cyber came in the wake of John Bates’ October 3, 2011 ruling that the NSA had illegally collected US person data. Of crucial importance, Bates judged that data obtained in response to a particular selector was intentionally, not incidentally, collected (even though the IC and its overseers like to falsely claim otherwise), even data that just happened to be collected in the same transaction. Crucially, pointing back to his July 2010 opinion on the Internet dragnet, Bates said that disclosing such information, even just to the court or internally, would be a violation of 50 USC 1809(a), which he used as leverage to make the government identify and protect any US person data collected using upstream collection before otherwise using the data. I believe this decision established a precedent for upstream 702 that would make it very difficult for FISC to permit the use of cyber signatures that happened to be collected domestically (which would count as intentional domestic collection) without rigorous minimization procedures.

The government, at a time when it badly wanted a cyber certificate, considered appealing his decision, but ultimately did not. Instead, they destroyed the data they had illegally collected and — in what was almost certainly a related decision — destroyed all the PATRIOT-authorized Internet dragnet data at the same time, December 2011. Bates did permit the government to keep collecting upstream data, but only under more restrictive minimization procedures.

Did FISC approve a cyber certificate but with sharp restrictions on retention and dissemination?

Neither ProPublica/NYT nor Mayer claimed NSA had obtained an upstream cyber certificate (though many other people have assumed it did). We actually don’t know, and the evidence is mixed.

Even as the government was scrambling to implement new upstream minimization procedures to satisfy Bates’ order, NSA had another upstream violation. That might reflect informing Bates, for the first time (there’s no sign they did inform him during the 2011 discussion, though the 2011 minimization procedures may reflect that they already had), they had been using upstream to collect on cyber signatures, or one which might represent some other kind of illegal upstream collection. When the government got Congress to reauthorize FAA that year, it did not inform them they were using or intended to use upstream collection to collect cyber signatures. Significantly, even as Congress began debating FAA, they considered but rejected the first of the predecessor bills to CISA.

My guess is that the FISC did approve cyber collection, but did so with some significant limitations on it, akin to, or perhaps even more restrictive, than the restrictions on multiple communication transactions (MCTs) required in 2011. I say that, in part, because of language in USA F-ReDux (section 301) permitting the government to use information improperly collected under Section 702 if the FISA Court imposed new minimization procedures. While that might have just referred back to the hypothetical 2011 example (in which the government had to destroy all the data), I think it as likely the Congress was trying to permit the government to retain data questioned later.

More significantly, the 2014 NSA, FBI, and CIA minimization procedures contain some version of this language, which appears to be new from the 2011 procedures.

Additionally, nothing in these procedures shall restrict NSA’s ability to conduct vulnerability or network assessments using information acquired pursuant to section 702 of the Act in order to ensure that NSA systems are not or have not been compromised. Notwithstanding any other section in these procedures, information used by NSA to conduct vulnerability or network assessments may be retained for one year solely for that limited purpose. Any information retained for this purpose may be disseminated only in accordance with the applicable provisions of these procedures.

That is, the FISC approved new procedures that permit the retention of vulnerability information for use domestically, but it placed even more restrictions on it (retention for just one year, retention solely for the defense of that agency’s network, which presumably prohibits its use for criminal prosecution, not to mention its dissemination to other agencies, other governments, and corporations) than it had on MCTs in 2011.

To be sure, there is language in both 2011 and 2014 NSA MPs that permits the agency to retain and disseminate domestic communications if it is necessary to understand a communications security vulnerability.

the communication is reasonably believed to contain technical data base information, as defined in Section 2(i), or information necessary to understand or assess a communications security vulnerability. Such communication may be provided to the FBI and/or disseminated to other elements of the United States Government. Such communications may be retained for a period sufficient to allow a thorough exploitation and to permit access to data that are, or are reasonably believed likely to become, relevant to a current or future foreign intelligence requirement. Sufficient duration may vary with the nature of the exploitation.

But at least on its face, that language is about retaining information to exploit (offensively) a communications vulnerability. Whereas the more recent language — which is far more restrictive — appears to address retention and use of data for defensive purposes.

The 2011 ruling strongly suggested that FISC would interpret Section 702 to prohibit much of what Mayer envisioned in his model. And the addition to the 2014 minimization procedures leads me to believe FISC did approve very limited use of Section 702 for cyber security, but with such significant limitations on it (again, presumably stemming from 50 USC 1809(a)’s prohibition on disclosing data intentionally collected domestically) that the IC wanted to find another way. In other words, I suspect NSA (and FBI, which was working closely with NSA to get such a certificate in 2012) got their cyber certificate, only to discover it didn’t legally permit them to do what they wanted to do.

CISA is the new and improved cyber-FISA

And while I’m not certain, I believe that in ensuring that DHS’ scrubs get dismantled, CISA gives the IC a way to do what it would have liked to with a FISA 702 cyber certificate.

Let’s go back to Mayer’s model of what the IC would probably like to do: A private company finds a threat, removes private data, leaving just a selector, after which NSA deploys the selector on backbone traffic, which then reproduces the private data, presumably on whatever parts of the Internet backbone NSA has access to via its upstream selection (which is understood to be infrastructure owned by the telecoms).

But in fact, Step 4 of Mayer’s model — NSA deploys the signature as a selector on the Internet backbone — is not done by the NSA. It is done by the telecoms (that’s the Section 702 cooperation part). So his model would really be private business > DHS > NSA > private business > NSA > treatment under NSA’s minimization procedures if the data were handled under upstream 702. Ultimately, the backbone operator is still going to be the one scanning the Internet for more instances of that selector; the question is just how much data gets sucked in with it and what the government can do once it gets it.

And that’s important because CISA codifies private companies’ authority to do that scan.

For all the discussion of CISA and its definition, there has been little discussion of what might happen at the private entities. But the bill affirmatively authorizes private entities to monitor their systems, broadly defined, for cybersecurity purposes.


(1) IN GENERAL.—Notwithstanding any other provision of law, a private entity may, for cybersecurity purposes, monitor—

(A) an information system of such private entity;

(B) an information system of another entity, upon the authorization and written consent of such other entity;

(C) an information system of a Federal entity, upon the authorization and written consent of an authorized representative of the Federal entity; and

(D) information that is stored on, processed by, or transiting an information system monitored by the private entity under this paragraph.

(2) CONSTRUCTION.—Nothing in this subsection shall be construed—

(A) to authorize the monitoring of an information system, or the use of any information obtained through such monitoring, other than as provided in this title; or

(B) to limit otherwise lawful activity.

Defining monitor this way:

(14) MONITOR.—The term ‘‘monitor’’ means to acquire, identify, or scan, or to possess, information that is stored on, processed by, or transiting an information system.

That is, CISA affirmatively permits private companies to scan, identify, and possess cybersecurity threat information transiting or stored on their systems. It permits private companies to conduct precisely the same kinds of scans the government currently obligates telecoms to do under upstream 702, including data both transiting their systems (which for the telecoms would be transiting their backbone) or stored in its systems (so cloud storage). To be sure, big telecom and Internet companies do that anyway for their own protection, though this bill may extend the authority into cloud servers and competing tech company content that transits the telecom backbone. And it specifically does so in anticipation of sharing the results with the government, with very limited requirement to scrub the data beforehand.

Thus, CISA permits the telecoms to do the kinds of scans they currently do for foreign intelligence purposes for cybersecurity purposes in ways that (unlike the upstream 702 usage we know about) would not be required to have a foreign nexus. CISA permits the people currently scanning the backbone to continue to do so, only it can be turned over to and used by the government without consideration of whether the signature has a foreign tie or not. Unlike FISA, CISA permits the government to collect entirely domestic data.

Of course, there’s no requirement that the telecoms scan for every signature the government shares with it and share the results with the government. Though both Verizon and AT&T have a significant chunk of federal business — which just got put out for rebid on a contract that will amount to $50 billion — and they surely would be asked to scan the networks supporting federal traffic for those signatures (remember, this entire model of scanning domestic backbone traffic got implicated in Qwest losing a federal bid which led to Joe Nacchio’s prosecution), so they’ll be scanning some part of the networks they operate with the signatures. CISA just makes it clear they can also scan their non-federal backbone as well if they want to. And the telecoms are outspoken supporters of CISA, so we should presume they plan to share promiscuously under this bill.

Assuming they do so, CISA offers several more improvements over FISA.

First — perhaps most important for the government — there are no pesky judges. The FISC gets a lot of shit for being a rubber stamp, but for years judges have tried to keep the government operating in the vicinity of the Fourth Amendment through its role in reviewing minimization procedures. Even John Bates, who was largely a pushover for the IC, succeeded in getting the government to agree that it can’t disseminate domestic data that it intentionally collected. And if I’m right that the FISC gave the government a cyber certificate but sharply limited how it could use that data, then it did so on precisely this issue. Significantly, CISA continues a trend we already saw in USA F-ReDux, wherein the Attorney General gets to decide whether privacy procedures (no longer named minimization procedures!) are adequate, rather than a judge. Equally significant, while CISA permits the use of CISA-collected data for a range of prosecutions, unlike FISA, it requires no notice to defendants of where the government obtained that data.

In lieu of judges, CISA envisions PCLOB and Inspectors General conducting the oversight (as well as audits being possible though not mandated). As I’ll show in a follow-up post, there are some telling things left out of those reviews. Plus, the history of DOJ’s Inspector General’s efforts to exercise oversight over such activities offers little hope these entities, no matter how well-intentioned, will be able to restrain any problematic practices. After all, DOJ’s IG called out the FBI in 2008 for not complying with a 2006 PATRIOT Act Reauthorization requirement to have minimization procedures specific to Section 215, but it took until 2013, with three years of intercession from FISC and leaks from Edward Snowden, before FBI finally complied with that 2006 mandate. And that came before FBI’s current practice of withholding data from its IG and even some information in IG reports from Congress.

In short, given what we know of the IC’s behavior when there was a judge with some leverage over its actions, there is absolutely zero reason to believe that any abuses would be stopped under a system without any judicial oversight. The Executive Branch cannot police itself.

Finally, there’s the question of what happens at DHS. No matter what you think about NSA’s minimization procedures (and they do have flaws), they do ensure that data that comes in through NSA doesn’t get broadly circulated in a way that identifies US persons. The IC has increasingly bypassed this control since 2007 by putting FBI at the front of data collection, which means data can be shared broadly even outside of the government. But FISC never permitted the IC to do this with upstream collection. So any content (metadata was different) on US persons collected under upstream collection would be subjected to minimization procedures.

This CISA model eliminates that control too. After all, CISA, as written, would let FBI and NSA veto any scrub (including of content) at DHS. And incoming data (again, probably including content) would be shared immediately not only with FBI (which has been the vehicle for sharing NSA data broadly) but also Treasury and ODNI, which are both veritable black holes from a due process perspective. And what few protections for US persons are tied to a relevance standard that would be met by virtue of a tie to that selector. Thus, CISA would permit the immediate sharing, with virtually no minimization, of US person content across the government (and from there to private sector and local governments).

I welcome corrections to this model — I presume I’ve overstated how much of an improvement over FISA this program would be. But if this analysis is correct, then CISA would give the IC everything that would have wanted for a cybersecurity certificate under Section 702, with none of the inadequate limits that would have had and may in fact have. CISA would provide an administrative way to spy on US person (domestic) content all without any judicial overview.

All of which brings me back to why the IC wants this this much. In at least one case, the IC did manage to use a combination of upstream and PRISM collection to stop an attempt to steal large amounts of data from a defense contractor. That doesn’t mean it’ll be able to do it at scale, but if by offering various kinds of immunity it can get all backbone providers to play along, it might be able to improve on that performance.

But CISA isn’t so much a cybersecurity bill as it is an Internet domestic spying bill, with permission to spy on a range of nefarious activities in cyberspace, including kiddie porn and IP theft. This bill, because it permits the spying on US person content, may be far more useful for that purpose than preventing actual hacks. That is, it won’t fix the hacking problem (it may make it worse by gutting Federal authority to regulate corporate cyber hygiene). But it will help police other kinds of activity.

If I’m right, the IC’s insistence it needs CISA — in the name of, but not necessarily intending to accomplish — cybersecurity makes more sense.

Update: This post has been tweaked for clarity.

Update, November 5: I should have written this post before I wrote this one. In it, I point to language in the August 26, 2014 Thomas Hogan opinion reflecting earlier approval, at least in the FBI minimization procedures, to share cyber signatures with private entities. The first approval was on September 20, 2012. The FISC approved the version still active in 2014 on August 30, 2013. (See footnote 19.) That certainly suggests FISC approved cyber sharing more broadly than the 2011 opinion might have suggested, though I suspect it still included more restrictions than CISA would. Moreover, if the language only got approved for the FBI minimization procedures, it would apply just to PRISM production, given that the FBI does not (or at least didn’t used to) get unminimized upstream production.

Michael Mosman’s Deadlines Raise (More) Questions about the FISC Advocate

In the series of letters purporting to speak for “the judiciary,” Director of the Administrative Office of US Courts John Bates and (after Duff replaced him) James Duff expressed concern about how a FISC amicus would affect the timeliness of proceedings before the court. Bates worried that any involvement of an amicus would require even more lead time than the current one week requirement in FISC applications. He also worried that the presumption an amicus (and potentially tech experts) would have access to information might set off disputes with the Executive over whether they could really have it. Duff apparently worried that the perception that an amicus would oppose the government would lead the government to delay in handing over materials to the FISC.

Which is why I’m interesting in the briefing order Chief FISC Judge Thomas Hogan, signing for Michael Mosman, issued on Wednesday (see below for a timeline).

Back on September 17, Mosman appointed spook lawyer Preston Burton amicus. As part of that order, he gave the government 4 days to refuse to share information with Burton, but otherwise required Burton receive the application and primary order in this docket.

(Pursuant to 50 U.S.C. § 1803(i)(6)(A)(i), the Court has determined that the government’s application (including exhibits and attachments) and the full, unredacted Primary Order in this docket are relevant to the duties of the amicus. By September 22, 2015, or after receiving confirmation from SEPS that the amicus has received the appropriate clearances and access approvals for such materials, whichever is later, the Clerk of the Court shall make these materials available to the amicus.

Yet even after the almost month long delay in deciding to appoint someone and deciding that someone would be Burton, it still took Mosman two weeks after the date when Burton was supposed to have received the relevant information on this issue before setting deadlines. And in setting his deadlines, Mosman has basically left himself only 2 weeks during which time he will have to to decide the issue and the government will have to prepare to keep or destroy the data in question (in past data destruction efforts it has taken a fairly long time). That could be particularly problematic if Mosman ends up requiring the government to pull the data from EFF’s clients from the data retained under their protection order.

On November 28, the order authorizing the retention of this data expires.

To be fair, Mosman is definitely making a more concerted effort to comply with the appearance if not the intent of USA F-ReDux’s amicus provision than, say, Dennis Saylor (who blew if off entirely). And there may be aspects of this process — and FISC’s presumed effort to start coming up with a panel of amici by November 29 — that will take more time than future instances down the road.

Still, it’s hard to understand the almost 3 week delay in setting a briefing schedule.

Unless the government slow-walked giving even a spook lawyer not explicitly ordered to represent the interests of privacy approval to receive and then a packet of documents to review.

I suspect this represents a stall by the government, not FISC (though again, the month long delay in deciding to appoint an amicus didn’t help things, and FISC’s thus far 4 month delay in picking amici likely doesn’t help either). But whatever the cause of the delay, it may indicate a reluctance on someone’s part to use the amicus as intended.


July 27: ODNI declares that “NSA has determined” that “NSA will allow technical personnel to continue to have access to the historical metadata for an additional three months”

By August 20: Government asks for permission to retain data past November 28 (the government must submit major FISA orders at least a week in advance)

August 27: Mosman approves dragnet order, defers decision on data retention

September 17: Mosman appoints Burton and orders the government to cough up its application and the full order

September 21: Last date by which government can complain about sharing information with Burton

September 22: Date by which Burton must receive application and order

October 7: Mosman sets deadlines

October 29: Deadline for Burton’s first brief

November 6: Deadline for Government response

November 10: Deadline for Burton reply, if any

November 28: Expiration of authorization to retain data

Delusional DOJ Claims Documents Declassified, Released Under FOIA Not Declassified, Not Authentic

Screen Shot 2015-08-28 at 11.22.34 AM
Back in March, NYT’s Charlie Savage sued to get the NSA to respond to a FOIA request asking for “copies of — and declassification review of, as necessary” a bunch of things, including IG reports on “bulk phone records collection activities under Section 215 of the PATRIOT Act.”

In late August, they delivered an installment of their response to that suit to him including a series of IG Reports on the 215 program. Among other things, the FOIA response included an August 2, 2010 letter to FISC Judge John Bates referring to a compliance violation in Docket BR 10-10 (the order is dated February 26, 2010). In referring to the caption of that docket (and the caption redactions in other dockets are consistent in size), it named Verizon Wireless.

As I pointed out at the time, this provides Larry Klayman and other Verizon Wireless subscribers challenging the phone dragnet basis to establish standing to sue. While in the Klayman suit, Judge Richard Leon invited Klayman just to add a plaintiff who subscribed to Verizon Business Services, in Northern CA, EFF requested the 9th Circuit take judicial notice of the document.

So now DOJ has gone a bit batshit. (Josh Gerstein first reported on this here.) It mocks that EFF head Cindy Cohn “apparently believes” it fair to conclude Verizon Wireless took part in the phone dragnet because of a reference to “a company name that includes the term ‘Verizon Wireless’ in the caption of a purported FISC filing” that happens to govern the entire phone dragnet. It suggests the accuracy of the document DOJ gave to Savage can be reasonably questioned, apparently disputing its own FOIA response to Savage. And it bitches that EFF “does not contend that this document was declassified,” even though it was given to Savage pursuant to his request for “declassification review [] as necessary.”

In short, in an effort to argue the document doesn’t say what it says (which may, I admit, not mean what it says, but such is the wackiness of the secret FISA Court and the secret phone dragnet), DOJ is saying that DOJ didn’t provide Charlie Savage authentic, declassified documents like he sued to get. DOJ uses words like “purported” to describe DOJ’s own FOIA response.

I mean, I’ll grant you, those of us outside DOJ often doubt the accuracy of their FOIA responses to us. But usually DOJ at least pretends they’re giving us authentic documents.

BREAKING: What emptywheel Reported Two Years Ago

The NYT today:

The National Security Agency has used its bulk domestic phone records program to search for operatives from the government of Iran and “associated terrorist organizations” — not just Al Qaeda and its allies — according to a document obtained by The New York Times.


The inclusion of Iran and allied terrorist groups — presumably the Shiite group Hezbollah — and the confirmation of the names of other participating companies add new details to public understanding of the once-secret program. The Bush administration created the program to try to find hidden terrorist cells on domestic soil after the attacks of Sept. 11, 2001, and government officials have justified it by using Al Qaeda as an example.

emptywheel, 15 months ago:

I want to post Dianne Feinstein’s statement about what Section 215 does because, well, it seems Iran is now a terrorist. (This is around 1:55)

The Section 215 Business Records provision was created in 2001 in the PATRIOT for tangible things: hotel records, credit card statements, etcetera. Things that are not phone or email communications. The FBI uses that authority as part of its terrorism investigations. The NSA only uses Section 215 for phone call records — not for Google searches or other things. Under Section 215, NSA collects phone records pursuant to a court record. It can only look at that data after a showing that there is a reasonable, articulable that a specific individual is involved in terrorism, actually related to al Qaeda or Iran. At that point, the database can be searched. But that search only provides metadata, of those phone numbers. Of things that are in the phone bill. That person, um [flips paper] So the vast majority of records in the database are never accessed, and are deleted after a period of five years. To look at, or use content, a court warrant must be obtained.

Is that a fair description, or can you correct it in any way?

Keith Alexander: That is correct, Senator. [underline/italics added]

Some time after this post Josh Gerstein reported on Keith Alexander confirming the Iran targeting.

The NYT today:

One document also reveals a new nugget that fills in a timeline about surveillance: a key date for a companion N.S.A. program that collected records about Americans’ emails and other Internet communications in bulk. The N.S.A. ended that program in 2011 and declassified its existence after the Snowden disclosures.

In 2009, the N.S.A. realized that there were problems with the Internet records program as well and turned it off. It then later obtained Judge Bates’s permission to turn it back on and expand it.

When the government declassified his ruling permitting the program to resume, the date was redacted. The report says it happened in July 2010.

emptywheel in November 2013:

I’ve seen a lot of outright errors in the reporting on the John Bates opinion authorizing the government to restart the Internet metadata program released on Monday.

Bates’ opinion was likely written in July 2010.


It had to have been written after June 21, 2010 and probably dates to between June 21 and July 23, 2010, because page 92 footnote 78 cites Holder v. HLP (which was released on June 21), but uses a “WL” citation; by July 23 the “S. Ct.” citation was available. (h/t to Document Exploitation for this last observation).

So: it had to have been written between June 21, 2010 and October 3, 2011, but was almost certainly written sometime in the July 2010 timeframe.

The latter oversight is understandable, as this story — which has been cited in court filings — misread Claire Eagan’s discussions of earlier bulk opinions, which quoted several sentences of Bates’ earlier one (though it was not the among the stories that really botched the timing of the Bates opinion).

In September, the Obama administration declassified and released a lengthy opinion by Judge Claire Eagan of the surveillance court, written a month earlier and explaining why the panel had given legal blessing to the call log program. A largely overlooked passage of her ruling suggested that the court has also issued orders for at least two other types of bulk data collection.

Specifically, Judge Eagan noted that the court had previously examined the issue of what records are relevant to an investigation for the purpose of “bulk collections,” plural. There followed more than six lines that were censored in the publicly released version of her opinion.

There have been multiple pieces of evidence to confirm my earlier July 2010 deduction since.

The big news in the NYT story (though not necessarily the NYT documents, which I’ll return to) is that in 2010, Verizon Wireless also received phone dragnet orders. I’ll return to what that tells us too.

But the news that Iran was targeted under the phone dragnet was confirmed publicly — and reported here — in a prepared statement from the Senate Intelligence Chair and confirmed by the Director of National Security Agency a week after the first Snowden leak story.

The Timing of the Contemplated Upstream Cyber-Grab

There’s an aspect missing thus far from the discussion of NSA’s possible bid for a cyber certification under Section 702 for primary use in the collection of attack signatures that could not be attributed to a foreign government.

The timing.

The discussion of creating a new Section 702 certificate came in the aftermath of the 6-month back and forth between DOJ and the FISA Court over NSA having collected US person data as part of its upstream collection (for more detail than appears in the timeline below, see this post). During that process, John Bates ruled parts of the program — what he deemed the intentional collection of US person data within the US — to be unconstitutional. That part of his opinion is worth citing at length, because of the way Bates argues that the inability to detach entirely domestic communications that are part of a transaction does not mean that those domestic communications were “incidentally” collected. Rather, they were “intentionally” collected.

Specifically, the government argues that NSA is not “intentionally” acquiring wholly domestic communications because the government does not intend to acquire transactions containing communications that are wholly domestic and has implemented technical means to prevent the acquisition of such transactions. See June 28 Submission at 12. This argument fails for several reasons.

NSA targets a person under Section 702 certifications by acquiring communications to, from, or about a selector used by that person. Therefore, to the extent NSA’s upstream collection devices acquire an Internet transaction containing a single, discrete communication that is to, from, or about a tasked selector, it can hardly be said that NSA’s acquisition is “unintentional.” In fact, the government has argued, that the Court has accepted, that the government intentionally acquires communications to and from a target, even when NSA reasonably — albeit mistakenly — believes that the target is located outside the United States. See Docket No. [redacted]


The fact that NSA’s technical measures cannot prevent NSA from acquiring transactions containing wholly domestic communications under certain circumstances does not render NSA’s acquisition of those transactions “unintentional.”


[T]here is nothing in the record to suggest that NSA’s technical means are malfunctioning or otherwise failing to operate as designed. Indeed, the government readily concedes that NSA will acquire a wholly domestic “about” communication if the transaction containing the communication is routed through an international Internet link being monitored by NSA or is routed through a foreign server.


By expanding its Section 702 acquisitions to include the acquisition of Internet transactions through its upstream collection, NSA has, as a practical matter, circumvented the spirit of Section 1881a(b)(4) and (d)(1) with regard to that collection. (44-45, 48)

There are a number of ways to imagine that victim-related data and communications obtained with an attack signature might be considered “intentional” rather than “incidental,” especially given the Snowden document acknowledging that so much victim data gets collected it should be segregated from regular collection. Add to that the far greater likelihood that the NSA will unknowingly target domestic hackers — because so much of hacking involves obscuring attribution — and the likelihood upstream collection targeting hackers would “intentionally” collect domestic data is quite high.

Plus, there’s nothing in the 2011 documents released indicating the FISC knew upstream collection included cyber signatures — and related victim data — in spite of the fact that “current Certifications already allow for the tasking of these cyber signatures.” No unredacted section discussed the collection of US person data tied to the pursuit of cyberattackers that appears to have been ongoing by that point.

Similarly, the white paper officially informing Congress about 702 didn’t mention cyber signatures either. There’s nothing public to suggest it did so after the Senate rejected a Cybersecurity bill in August, 2012, either. That bill would have authorized less involvement of NSA in cybersecurity than appears to have already been going on.

With all that in mind, consider the discussions reflected in the documents released last week. The entire discussion to use FBI’s stated needs to apply as backup to apply for a cyber certificate came at the same time as NSA is trying to decide what to do with the data it illegally collected. Before getting that certificate, DOJ approved the collection of cyber signatures under other certificates. It seems likely that this collection would violate the spirit of the ruling from just the prior year.

And NSA’s assistance to FBI may have violated the prior year’s orders in another way. SSO contemplated delivering all this data directly to FBI.

Screen Shot 2015-06-11 at 9.42.56 AM

Yet one of the restrictions imposed on upstream collection — voluntarily offered up by DOJ — was that no raw data from NSA’s upstream collection go to FBI (or CIA). If there was uncertainty where FBI’s targeting ended and NSA’s began, this would create a violation of prior orders.

Meanwhile, the reauthorization process had already started, and as part of that (though curiously timed to coincide with the release of DOJ’s white paper on 702 collection) Ron Wyden and Mark Udall were trying to force NSA to figure out how much US person data they were collecting. Not only did the various Inspectors General refuse to count that data (which would have, under the logic of Bates’ opinions finding that illegally collected data was only illegal if the government knew it was US person data, made the data illegal), but the Senate Intelligence Committee refused to consider reconstituting their Technical Advisory Committee which might be better able to assess whether NSA claims were correct.

Sometime in that period, just as Wyden was trying to call attention to the fact that NSA was collecting US person data via its upstream collection, NSA alerted the Intelligence Committees to further “overcollection” under upstream collection.

2012 Upstream Notice

As I suggested here, the length of the redaction and mention of “other authorities” may reflect the involvement of another agency like FBI. One possibility, given the description of FBI collecting on cyber signatures using both PRTT and (presumably) traditional FISA in the discussions of SSO helping the FBI conduct this surveillance (note, I find it interesting though not conclusive that there is no mention of Section 215 to collect cybersecurity data), is that the initial efforts to go after these signatures in some way resulted in overcollection. If FISC interpreted victim-related data to be overcollection — as would be unsurprising under Bates’ 2011 upstream opinion — then it would explain the notice to Congress.

One more point. In this post, I noted that USA F-ReDux authorized FISC to let the government use data it had illegally collected but which FISC had authorized by imposing additional minimization procedures. It’s just a wildarseguess, but I find it plausible that this 2012 overcollection involved cyber signatures (because we know NSA was collecting it and there is reason to believe it violated Bates’ 2011 opinion), and that any victim data now gets treated under minimization procedures and therefore that any illegal data from 2012 may now, as of last week, be used.

All of which is to say that the revelation of NSA and FBI’s use of upstream collection to target hackers involves far more legal issues than commentary on the issue has made out. And these legal issues may well have been more appropriate for the government to reveal before passage of USA F-ReDux.

Update, 11/6: Some dates added from this opinionContinue reading

1 2 3 10
Emptywheel Twitterverse
JimWhiteGNV RT @emptywheel: 1) I think what happened at Planned Parenthood was terrorism. 2) It was incited. 3) Anti-choicers still have 1A rights to o…
emptywheel 1) I think what happened at Planned Parenthood was terrorism. 2) It was incited. 3) Anti-choicers still have 1A rights to oppose abortion.
emptywheel @noFelld Probably, but not sure. Commentators here keep forgetting that IA exists.
bmaz @freespeechlover @emptywheel Yes. but that is the fun of it.
bmaz @freespeechlover @emptywheel @CalFootball Though Graham not that good of a recruiter or team assembler either; best success w/prior recruits
bmaz @freespeechlover @emptywheel @CalFootball Well, I am happy for them and him. All I know is my coach sucks, especially on game days.
JimWhiteGNV @armandodkos Nooo! I missed that. I'm starting to feel very lonely. @bmaz
emptywheel Really nice D there.
bmaz @JimWhiteGNV I will give you two points for that Tweet!
bmaz @freespeechlover @emptywheel @CalFootball Yes, but Stanford has a real coach (Cal too probably) we have fucking Todd Graham.
JimWhiteGNV I used to think of @bmaz as my friend. Now, I just don't know...
November 2015
« Oct