![]() ![]() Together with a group of historians, computer scientists, and statisticians, Connelly is developing an ambitious project called the Declassification Engine, which, among other things, employs machine-learning and natural language processing to study the semantic patterns in declassified text. On a cloudless afternoon not long ago, I met with Matthew Connelly, a Columbia history professor, outside the National Archives in Washington, D.C. Slowly, an attentive researcher can chip away at the blacked-out parts of a document, building context and fuelling further excavation.īut Trachtenberg’s techniques, though fundamentally sound, are slow, and naturally other researchers have taken up the task of trying to automate the process, at least in part. The guidelines for declassification vary across agencies and offer room for interpretation, so that a sanitizer responding to a Freedom of Information Act request in 1992 might redact a document differently from a sanitizer in 2012, creating advantageous inconsistencies. “The basic idea is to exploit the fact that documents-the same documents-are declassified in different ways in different repositories,” revealing different information in each version, he said. Trachtenberg, who has refined old methods for analyzing the redacted segments in declassified texts, utilizes a sort of comparative reading to see beneath the black. “Once a user enters data into the document, effectively removing it can be difficult.” “Complex file formats offer substantial avenues for hidden data,” it warns. The manual emphasizes a new set of actions when working with a word processor, sanitizers must delete sensitive content, replace it with “innocuous text” to preserve formatting, and only then cover the innocuous text with a digitally drawn black-or, as it recommends, gray-box. released an updated version of “Redacting with Confidence,” its how-to guide for the declassification of digital documents. The rise of born-digital documents has brought new challenges: in 2009, the N.S.A. Occasionally, a sanitizer simply covers text with another piece of paper when he photocopies the document. You can usually identify the tool by the marks it leaves behind: the pen-redacted page is filled with heavy, imperfect lines, while the razor knife and opaque tape both leave sharp edges (though the photocopier gives a knifed-out block a mottled, grayish hue). Sanitizers also employ opaque tape and razor knives, cutting out the sensitive content from a copy of the page. ![]() But it can be sloppy, and the sheen of a photocopy sometimes reveals the letters beneath the ink. The black marker pen is the sanitizer’s most basic tool. For an act so often associated with the anonymous, passionless churning of the government machine, redaction betrays a striking individualism in the choices about what to leave visible and what to obscure, and in the shapes of the black bars themselves. Most government agencies that handle classified information have dedicated sanitizers. (Think of the cables released by WikiLeaks in 2010, some of which didn’t reveal sensitive information but were merely unflattering.) Government officials frequently perpetuate this culture by invoking national security, but Marc Trachtenberg, a Cold War historian at U.C.L.A., told me that “the function of declassification is much broader than keeping information from the enemy.” Often documents remain classified simply to save face. A 2012 report by the Public Interest Declassification Board, a government-funded advisory group, found that a “culture of caution” among executive-branch agencies had lead to chronic over-classification and, in turn, has “compromised” the entire classification system. ![]() government has a tendency toward over-classification. ![]()
0 Comments
Leave a Reply. |