Slashdot Mirror
NSA Makes Contribution To Apache Hadoop Project
An anonymous reader writes "The National Security Agency has submitted a new database, Accumulo, to the Apache Foundation for incubation. Accumulo is based on the original BigTable paper with some extensions such as the ability to provide cell-level security. It appears there are some hurdles that must be cleared concerning copyright before the project could be accepted."
Apache Brand
Our interest in releasing this code as an Apache incubator project is due to its strong relationship with other Apache projects, i.e. Accumulo has dependencies on Hadoop, Zookeeper, and Thrift and has complementary goals to HBase.
'The tyrant will always find pretext for his tyranny.' - Aesop's Fables
But other companies and individuals that produce works do get copyright. While they may give the government (and even the NSA) a license to use their works, the government can't just donate those works off to Apache without clearing it first. That means any code the NSA didn't write themselves needs to be removed, replaced, or also donated by the owner.
You do not have a moral or legal right to do absolutely anything you want.
NSA has been trying for decades to get vendors to get serious about security, without much success. One of NSA's units is the Central Security Service, the defensive side, which develops and tests security technologies for Government and military use. They have people testing safes and locks, for example.
Back in the 1980s, NSA tried applying that approach to computing, with the Trusted Computer System Evaluation Criteria. Systems were classified from A1 down to D. A very few specialized systems made it to an A level, but most commercial systems couldn't come close.
Manufacturers hated the testing procedure. Software vendors are used to controlling their own Q/A process. The NSA approach came from the test procedures for safes and padlocks - vendors could submit something, and it was tested by NSA personnel against NSA criteria. If it failed, the manufacturer got a list of defects, which was not necessarily complete. The manufacturer could resubmit the product, and NSA would retest it, on a strictly pass/fail basis. No third try was allowed, and failure was publicly announced by NSA.
After a decade of screaming and foot-dragging by vendors, the "common criteria" security scheme replaced the TCSEC in 2002-2005. This is much more "vendor friendly". The most strict levels of the TCSEC criteria were removed. Security evaluation is mostly done by outside labs, not NSA, and the vendor pays for and controls the process. The vendor can keep trying to pass as many times as they want. Failure is not publicized.
A reasonable number of systems meet some levels of the common criteria, but nothing below EAL5 really means much. Windows XP made it to EAL4.
NSA has tried, with NSA Secure Linux, to get people to take mandatory security seriously. NSA Secure Linux has "mandatory security", where there are levels and compartments which create boundaries data is not allowed to cross. Think of everything being in its own sandbox, with limited and tightly controlled intercommunication between sandboxes.
The point of that is not that NSA Secure Linux is a highly secure implementation of mandatory security. It was to get people to implement, modify, and partition applications so that they could work under a mandatory security model. A web browser, for example, would have to be structured so that the parts which could open local files were completely separated from the parts that communicated with the untrusted outside world. This didn't catch on in the browser world, although finally, a decade or so too late, browsers are starting to to run Flash in sandboxes.
NSA keeps trying. This new database is one for which fine-grained access control is possible. The challenge is to write apps that can live with such tight controls. They're trying to get people to get serious about security.
(It's been a long time, but I used to work on this stuff.)