NSA reveals its secret: No backdoor in encryption standard
Agency official details for the first time NSA's role in developing DES
- By William Jackson
- Feb 16, 2011
SAN FRANCISCO — The National Security Agency made changes in the proposed design of the Data Encryption Standard before its adoption in 1976, but it did not add any backdoors or other surprises that have been speculated about for 35 years, the technical director of NSA’s information assurance directorate said Wednesday.
“We’re actually pretty good guys,” said Dickie George. “We wanted to make sure we were as squeaky clean as possible.”
Besides, “I don’t think we were good enough to sneak things in that you guys wouldn’t have found,” he told a crowd of crypto professionals and security officials.
George detailed for the first time, in a presentation at the RSA Security Conference, the role of the NSA in developing and vetting DES, the nation’s first commercial encryption standard. Changes were made by the NSA in the Substitution boxes, or S-boxes, which provide the core security for the algorithm that was submitted for approval as a standard by IBM.
“We had to change the S-boxes,” to make them strong enough to withstand practical attacks on the cryptography, George said. “We didn’t see any need to change anything else.”
3 basic steps to thwart most cyberattacks, courtesy of NSA
FBI accused of installing backdoor in OpenBSD operating system
The agency’s goal was to make the algorithm as strong as possible, he said. “This is going to be used by the U.S. government and the U.S. banks,” and not by the Russians, he said. “It had to be as good as advertised.”
It was good enough to survive as a standard, in the form of Triple DES, until it began being replaced by the Advanced Encryption Standard that was adopted in 2001, which was much longer than any of the teams evaluating it expected it to last, George said.
“They did not believe DES would be the answer for as long as it was the answer,” he said, so provisions were not made to update it by lengthening its relatively weak 56-bit key.
George was a member of one of the evaluation teams that looked at DES in the mid 1970s, and spent the last year putting together the story of NSA’s involvement from contemporary documents and notes. The job was a challenge and produced surprises, he said.
“It was amazing how many people were on teams that I didn’t know about,” he said. “Not all of the notes on the different teams agreed.”
But the basic story is that the National Bureau of Standards, the forerunner of the National Institute of Standards and Technology, decided in 1972 that a commercial cryptographic standard was needed, and asked NSA for help because the agency basically had a monopoly on serious cryptographic expertise.
NSA had two goals in vetting submitted algorithms: To assure that there were no attacks against them that were better than brute force exhaustion, and to make no changes that were not necessary. It was decided early on that NSA should not submit a candidate of its own.
“We felt there wouldn’t be a lot of trust,” in an NSA algorithm, a concern that was borne out by reaction to NSA’s involvement in the process.
The initial call for proposals resulted in only a handful of requests for grants to study the problem. A second request produced three submissions, only two of which were acceptable. One of those accepted was based on a one-time pad, which would have been secure, “but the key management would have been horrendous,” George said. “So at that point we had a winner,” with a few necessary changes.
One of the major contributions of DES was the stimulus it gave to the development of the non-government cryptographic community, George said. After DES, NSA no longer had a monopoly on cryptography.
“We greatly underestimated how the Internet would allow a virtual cryptographic mass to develop” he said. “We benefitted greatly by seeing all the work that was done on the outside” in evaluating and trying to break DES.
That marked a major shift in the use of cryptography in the civilian world and a shift in NSA policy toward it.
“For a politician it was probably very scary,” George said of the shift. “For a mathematician it was a lot of fun. DES presented a problem for the world to work on,” and gave outside analysts their first big opportunities. And it is not dead yet. “DES is still a viable problem,” for analysts, he said.
William Jackson is a senior writer of GCN and the author of the CyberEye blog.