PGP CTO Blog

Hardware Encryption is Coming - and Not a Moment Too Soon
12 Apr 2006

I've written and spoken before about the trend toward pervasive encryption. One component of that trend is that more hardware devices have on-board encryption. This development also ties into another recurring theme of mine: the need to mix security and reliability.

There are new announcements almost daily. Seagate is putting encryption into the drive itself. Trusted Platform Module (TPM) chips put a smart card onto the motherboard of PCs. Chipmakers such as VIA Technologies are putting cryptographic algorithms in the CPUs. Intel and AMD have made their own plans, and IBM has announced its SecureBlue mechanisms that it hopes to license to chipmakers.

For us in the security software business, this trend is good news. We at PGP Corporation support between one and two dozen different hardware implementations of the core crypto in our products. The more we can rely on a given device having encryption, the easier it is for us to solve the hard problems that are important to you, the people who will be using these systems.

Encryption is the easy part of any of these systems, and even that is harder than you'd expect. It will be used by individual end users as well as people in small and large enterprises. There is the problem of backups and, even more important, there is the problem of restores. Most important of all, there is the problem of ease of use. The more encryption becomes part of everything, the more it will be used by people who shouldn't have to understand it at all.

There are many parts of these systems that will have to be improved over the next few generations of hardware. The first-generation systems that are appearing are like the proverbial dancing bear ^*. These first-generation hardware systems are not as good as the n-th generation software systems. This assessment isn"t an insult, and it's not dismissive. There"s simply no way the new systems could be as good as the established ones. What is exciting is that they are here in a form that can be used for several things. As we start integrating hardware encryption into the security platform, it will let product designers worry less about the nuts and bolts of security and more about the real goal--securing everything in a way that real people can use.

There are other technologies that have even more promise, but "promise" is a way of saying "future." We are still years away from widespread core cryptographic support in commonly used CPUs. VIA has it now, we'll get it eventually from Intel and AMD, and then you, the customer, will slowly get it as well. At what time should we start supporting it? The answer involves a huge chicken-and-egg problem for everyone. The hardware makers drag their feet in implementing something if they don't think the software will support it. We software people drag our feet because what we really want is the second-generation (or third-generation) system, and the software we have now is likely as good as the first-generation hardware.

As a case in point, Intel put random number generation in the support chips for the Pentium III and did a wonderful job of working with us software people. PGP® software supported it as the first chips were delivered. That hardware support didn't appear in all Pentium III systems, however, and was dropped from the Pentium IV. PGP software still supports the Pentium III random number generator, but you probably got rid of your Pentium III about 4 years ago. We would be delighted to have this functionality back. It helps us tremendously.

The good news is that hardware support for encryption is on its way. Some improvements will be here soon, some years from now. In the long run, these will make for better, faster, easier products. Unfortunately, there is still a lot of work to do in building, integrating, and making these products work with the software you use.