>...but seriously won't the FBI just cite
>something like "national security" or some other
> phrase to get what they want when they haven't
> got a valid case?
--- No, the 4758 is carefully designed so that no one - even the designers - can get into it if properly set up. One of the design principles of a good security product is that knowledgeable insiders - including the engineers who designed it, or the programmers who wrote the software - should be unable to break in.
It's a little misleading to say the 4758 is "designed to destroy itself" when it detects a tamper attempt. What it does is immediately destroy all sensitive data stored inside the secure module. Thus, any data inside the card (like keys) cannot be obtained by anyone attacking the card. There are a number of sensors and barriers to detect many different hi-tech attempts to break in or extract data, and all of them trigger hardware-based tamper response mechanisms that destroy data before the attacker could get inside.
Note that with the 4758, it is entirely possible to have the card generate its own RSA key pair, then keep the private key inside, so that it can never be compromised by getting at a backup, or at another card with the same key. The standard 4758 CCA software provides functions to do that, as it is a requirement by some applications. (Yes, they realize that the keys are lost if the hardware fails, is stolen, etc.)
IBM promptly fixed the problems identified by the Cambridge researchers, and the fixes have been on the IBM 4758 web site for some time. (http://www.ibm.com/security/cryptocards). Note that the problems were not exploitable in most real-world systems, and there were no known exploiters of this attack.
Slashdot Mirror
>...but seriously won't the FBI just cite >something like "national security" or some other > phrase to get what they want when they haven't > got a valid case? --- No, the 4758 is carefully designed so that no one - even the designers - can get into it if properly set up. One of the design principles of a good security product is that knowledgeable insiders - including the engineers who designed it, or the programmers who wrote the software - should be unable to break in.
It's a little misleading to say the 4758 is "designed to destroy itself" when it detects a tamper attempt. What it does is immediately destroy all sensitive data stored inside the secure module. Thus, any data inside the card (like keys) cannot be obtained by anyone attacking the card. There are a number of sensors and barriers to detect many different hi-tech attempts to break in or extract data, and all of them trigger hardware-based tamper response mechanisms that destroy data before the attacker could get inside. Note that with the 4758, it is entirely possible to have the card generate its own RSA key pair, then keep the private key inside, so that it can never be compromised by getting at a backup, or at another card with the same key. The standard 4758 CCA software provides functions to do that, as it is a requirement by some applications. (Yes, they realize that the keys are lost if the hardware fails, is stolen, etc.)
IBM promptly fixed the problems identified by the Cambridge researchers, and the fixes have been on the IBM 4758 web site for some time. (http://www.ibm.com/security/cryptocards). Note that the problems were not exploitable in most real-world systems, and there were no known exploiters of this attack.