Crypto gurus: The government's key escrow plan won't work

U.S. government officials have been increasingly hostile over the past year to the widespread use of encryption on mobile phones and online communications, arguing that a way needs to be found to provide law enforcement and intelligence agencies with lawful interception capabilities.

In response, security experts warned that building "back doors" into cryptographic systems in order to provide governments with access to data would be dangerous because it would create vulnerabilities that could later be exploited by hackers too.

"I don't want a back door," said U.S. National Security Agency Director Michael Rogers during a recent speech at Princeton University, the Washington Post reported. "I want a front door. And I want the front door to have multiple locks. Big locks."

Rogers suggested that technology companies could create a master key and then split it up into pieces that would be distributed to multiple parties so that no one person or agency could by itself decrypt communications or data stored on devices.

There are many cryptographic engineering problems with implementing such a "key escrow" system, including making complex changes to the already fragile cryptographic APIs and libraries in most operating systems, experts said.

Some of the issues are described very well in a blog post by Matthew Green, a research professor at Johns Hopkins University, said Ron Rivest, one of the inventors of the RSA cryptosystem, Tuesday during a discussion panel at the RSA security conference.

One problem that hasn't received a lot of consideration yet is that if the U.S. government builds such a system, other governments will want similar "doors" to access the data, Rivest said.

The system would be "like a house with many, many doors, with keys held by many, many parties," he said. "It's just not going to work."

"There really is no difference between back doors and front doors," said Adi Shamir, another co-inventor of the RSA algorithm and speaker on the same panel. "The only difference is that the NSA will have to take your house and turn it around."

"Technically speaking, I think there's some misunderstanding about the secret sharing aspect of key escrow" in the plan suggested by the U.S. government, Shamir said, adding that he's all in favor of his old ideas being put into practice. "But, I think in this particular case, the head of the NSA is just misusing the idea."

"I see no security benefit for taking a key and breaking it up in 10 different shares, if all those shares will be revealed all at once, or none at all," he said.

Whitfield Diffie, co-inventor of public-key cryptography, highlighted a different reason why he thinks such a system is not practical -- "not to disparage the other ways in which it's not going to work," he said to the audience's amusement.

With the current state of technology it's much easier to pre-encrypt what is sent into the encrypted channel than it was when the same problem was raised 20 years ago, Diffie said. In such a case, the government is going to exercise a warrant and break the outer encryption layer that is accessible to them and then find out that you've encrypted it in some other way, he said.

"In that case they'll do what they might have had to do anyway: come down on you with a bench warrant or something and order you to tell them how to read it," Diffie said.

He suggested another approach: a crypto system whose output is not impossible to break, but that would require a lot of resources to do so, making the process expensive. That would allow governments to recover specific encrypted messages in a targeted manner, but would prevent them from doing it on a large scale.

Ed Giorgio, who served as chief code maker and code breaker for the NSA in his 30 years career at the agency, thinks that governments from around the world won't give up on their lawful interception authorities, so a solution will need to be found.

"It will be an ongoing negotiation," Giorgio said on the RSA panel. "It might result in key escrow 2.0, but I'm sure we'll see various versions of it."