Group items tagged

This language is broad, circular, and technically incoherent, so it takes some effort to parse appropriately. When the minimization procedures were disclosed in 2013, this language was interpreted by outside commentators to mean that NSA may keep all encrypted data that has been incidentally collected under Section 702 for at least as long as is necessary to decrypt that data. Is this the correct interpretation? I think so. It is important to realize that the language above isn’t just broad. It seems purposefully broad. The part regarding relevance seems to mirror the rationale NSA has used to justify its bulk phone records collection program. Under that program, all phone records were relevant because some of those records could be valuable to terrorism investigations and (allegedly) it isn’t possible to collect only those valuable records. This is the “to find a needle a haystack, you first have to have the haystack” argument. The same argument could be applied to encrypted data and might be at play here.

This exception doesn’t just apply to encrypted data that might be relevant to a current foreign intelligence investigation. It also applies to cases in which the encrypted data is likely to become relevant to a future intelligence requirement. This is some remarkably generous language. It seems one could justify keeping any type of encrypted data under this exception. Upon close reading, it is difficult to avoid the conclusion that these procedures were written carefully to allow NSA to collect and keep a broad category of encrypted data under the rationale that this data might contain the communications of NSA targets and that it might be decrypted in the future. If NSA isn’t doing this today, then whoever wrote these minimization procedures wanted to at least ensure that NSA has the authority to do this tomorrow.

There are a few additional observations that are worth making regarding these nominally new retention guidelines and Section 702 collection. First, the concept of incidental collection as it has typically been used makes very little sense when applied to encrypted data. The way that NSA’s Section 702 upstream “about” collection is understood to work is that technology installed on the network does some sort of pattern match on Internet traffic; say that an NSA target uses example@gmail.com to communicate. NSA would then search content of emails for references to example@gmail.com. This could notionally result in a lot of incidental collection of U.S. persons’ communications whenever the email that references example@gmail.com is somehow mixed together with emails that have nothing to do with the target. This type of incidental collection isn’t possible when the data is encrypted because it won’t be possible to search and find example@gmail.com in the body of an email. Instead, example@gmail.com will have been turned into some alternative, indecipherable string of bits on the network. Incidental collection shouldn’t occur because the pattern match can’t occur in the first place. This demonstrates that, when communications are encrypted, it will be much harder for NSA to search Internet traffic for a unique ID associated with a specific target.

This lends further credence to the conclusion above: rather than doing targeted collection against specific individuals, NSA is collecting, or plans to collect, a broad class of data that is encrypted. For example, NSA might collect all PGP encrypted emails or all Tor traffic. In those cases, NSA could search Internet traffic for patterns associated with specific types of communications, rather than specific individuals’ communications. This would technically meet the definition of incidental collection because such activity would result in the collection of communications of U.S. persons who aren’t the actual targets of surveillance. Collection of all Tor traffic would entail a lot of this “incidental” collection because the communications of NSA targets would be mixed with the communications of a large number of non-target U.S. persons. However, this “incidental” collection is inconsistent with how the term is typically used, which is to refer to over-collection resulting from targeted surveillance programs. If NSA were collecting all Tor traffic, that activity wouldn’t actually be targeted, and so any resulting over-collection wouldn’t actually be incidental. Moreover, greater use of encryption by the general public would result in an ever-growing amount of this type of incidental collection.

This type of collection would also be inconsistent with representations of Section 702 upstream collection that have been made to the public and to Congress. Intelligence officials have repeatedly suggested that search terms used as part of this program have a high degree of specificity. They have also argued that the program is an example of targeted rather than bulk collection. ODNI General Counsel Robert Litt, in a March 2014 meeting before the Privacy and Civil Liberties Oversight Board, stated that “there is either a misconception or a mischaracterization commonly repeated that Section 702 is a form of bulk collection. It is not bulk collection. It is targeted collection based on selectors such as telephone numbers or email addresses where there’s reason to believe that the selector is relevant to a foreign intelligence purpose.” The collection of Internet traffic based on patterns associated with types of communications would be bulk collection; more akin to NSA’s collection of phone records en mass than it is to targeted collection focused on specific individuals. Moreover, this type of collection would certainly fall within the definition of bulk collection provided just last week by the National Academy of Sciences: “collection in which a significant portion of the retained data pertains to identifiers that are not targets at the time of collection.”

The Section 702 minimization procedures, which will serve as a template for any new retention guidelines established for E.O. 12333 collection, create a large loophole for encrypted communications. With everything from email to Internet browsing to real-time communications moving to encrypted formats, an ever-growing amount of Internet traffic will fall within this loophole.

"The presumption of openness requires the presumption of posting," said Archive director Tom Blanton. "For the new generation, if it's not online, it does not exist." The National Security Archive has conducted fourteen FOIA Audits since 2002. Modeled after the California Sunshine Survey and subsequent state "FOI Audits," the Archive's FOIA Audits use open-government laws to test whether or not agencies are obeying those same laws. Recommendations from previous Archive FOIA Audits have led directly to laws and executive orders which have: set explicit customer service guidelines, mandated FOIA backlog reduction, assigned individualized FOIA tracking numbers, forced agencies to report the average number of days needed to process requests, and revealed the (often embarrassing) ages of the oldest pending FOIA requests. The surveys include:

The federal government has made some progress moving into the digital era. The National Security Archive's last E-FOIA Audit in 2007, " File Not Found," reported that only one in five federal agencies had put online all of the specific requirements mentioned in the E-FOIA amendments, such as guidance on making requests, contact information, and processing regulations. The new E-FOIA Audit finds the number of agencies that have checked those boxes is now much higher — 100 out of 165 — though many (66 in 165) have posted just the bare minimum, especially when posting FOIA responses. An additional 33 agencies even now do not post these types of records at all, clearly thwarting the law's intent.

The FOIAonline Members (Department of Commerce, Environmental Protection Agency, Federal Labor Relations Authority, Merit Systems Protection Board, National Archives and Records Administration, Pension Benefit Guaranty Corporation, Department of the Navy, General Services Administration, Small Business Administration, U.S. Citizenship and Immigration Services, and Federal Communications Commission) won their "E-Star" by making past requests and releases searchable via FOIAonline. FOIAonline also allows users to submit their FOIA requests digitally.

Disabilities Compliance. Despite the E-FOIA Act, many government agencies do not embrace the idea of posting their FOIA responses online. The most common reason agencies give is that it is difficult to post documents in a format that complies with the Americans with Disabilities Act, also referred to as being "508 compliant," and the 1998 Amendments to the Rehabilitation Act that require federal agencies "to make their electronic and information technology (EIT) accessible to people with disabilities." E-Star agencies, however, have proven that 508 compliance is no barrier when the agency has a will to post. All documents posted on FOIAonline are 508 compliant, as are the documents posted by the Department of Defense and the Department of State. In fact, every document created electronically by the US government after 1998 should already be 508 compliant. Even old paper records that are scanned to be processed through FOIA can be made 508 compliant with just a few clicks in Adobe Acrobat, according to this Department of Homeland Security guide (essentially OCRing the text, and including information about where non-textual fields appear). Even if agencies are insistent it is too difficult to OCR older documents that were scanned from paper, they cannot use that excuse with digital records.

Key Findings

Excuses Agencies Give for Poor E-Performance

Justice Department guidance undermines the statute. Currently, the FOIA stipulates that documents "likely to become the subject of subsequent requests" must be posted by agencies somewhere in their electronic reading rooms. The Department of Justice's Office of Information Policy defines these records as "frequently requested records… or those which have been released three or more times to FOIA requesters." Of course, it is time-consuming for agencies to develop a system that keeps track of how often a record has been released, which is in part why agencies rarely do so and are often in breach of the law. Troublingly, both the current House and Senate FOIA bills include language that codifies the instructions from the Department of Justice. The National Security Archive believes the addition of this "three or more times" language actually harms the intent of the Freedom of Information Act as it will give agencies an easy excuse ("not requested three times yet!") not to proactively post documents that agency FOIA offices have already spent time, money, and energy processing. We have formally suggested alternate language requiring that agencies generally post "all records, regardless of form or format that have been released in response to a FOIA request."

THE E-DELINQUENTS: WORST OVERALL AGENCIES In alphabetical order

Privacy. Another commonly articulated concern about posting FOIA releases online is that doing so could inadvertently disclose private information from "first person" FOIA requests. This is a valid concern, and this subset of FOIA requests should not be posted online. (The Justice Department identified "first party" requester rights in 1989. Essentially agencies cannot use the b(6) privacy exemption to redact information if a person requests it for him or herself. An example of a "first person" FOIA would be a person's request for his own immigration file.) Cost and Waste of Resources. There is also a belief that there is little public interest in the majority of FOIA requests processed, and hence it is a waste of resources to post them. This thinking runs counter to the governing principle of the Freedom of Information Act: that government information belongs to US citizens, not US agencies. As such, the reason that a person requests information is immaterial as the agency processes the request; the "interest factor" of a document should also be immaterial when an agency is required to post it online. Some think that posting FOIA releases online is not cost effective. In fact, the opposite is true. It's not cost effective to spend tens (or hundreds) of person hours to search for, review, and redact FOIA requests only to mail it to the requester and have them slip it into their desk drawer and forget about it. That is a waste of resources. The released document should be posted online for any interested party to utilize. This will only become easier as FOIA processing systems evolve to automatically post the documents they track. The State Department earned its "E-Star" status demonstrating this very principle, and spent no new funds and did not hire contractors to build its Electronic Reading Room, instead it built a self-sustaining platform that will save the agency time and money going forward.

Let’s Encrypt, which was announced this week but won’t be ready to use until the second quarter of 2015, describes itself as “a free, automated, and open certificate authority (CA), run for the public’s benefit.” It’s the product of years of work from engineers at Mozilla, Cisco, Akamai, Electronic Frontier Foundation, IdenTrust, and researchers at the University of Michigan. (Disclosure: I used to work for the Electronic Frontier Foundation, and I was aware of Let’s Encrypt while it was being developed.) If Let’s Encrypt works as advertised, deploying HTTPS correctly and using all of the best practices will be one of the simplest parts of running a website. All it will take is running a command. Currently, HTTPS requires jumping through a variety of complicated hoops that certificate authorities insist on in order prove ownership of domain names. Let’s Encrypt automates this task in seconds, without requiring any human intervention, and at no cost.

The transition to a fully encrypted web won’t be immediate. After Let’s Encrypt is available to the public in 2015, each website will have to actually use it to switch over. And major web hosting companies also need to hop on board for their customers to be able to take advantage of it. If hosting companies start work now to integrate Let’s Encrypt into their services, they could offer HTTPS hosting by default at no extra cost to all their customers by the time it launches.