The consequences of infraction are becoming more serious as regulators and legislators increase the protection afforded to citizens' records. In the U.S., the Gramm-Leach-Bliley (GLB) Act requires all financial institutions to "design, implement and maintain safeguards to protect customer information." European data protection law stipulates that "appropriate care" must be taken with data.

In California, the state's 2003 Database Breach Act (DBA) requires organizations to disclose publicly any breach of personal data security and to individually contact the affected parties. Such provisions, carrying with them the threat of high-profile media coverage, are perhaps the greatest impetus of all toward the securing of sensitive data.

It is easy to assume that security risks are largely associated with the Internet and the transmission of data. The fact is, however, that problems are just as often encountered because of the highly portable nature of today's data-storage devices, including laptops and various hand-held products.

In addition to imposing obligations, many of the relevant rules and regulations also suggest a solution to the problem: data encryption.

Consequently, manufacturers of external disk drives for business or home use have begun to incorporate enhanced data security features into their products. Increasingly, hardware manufacturers are looking to incorporate hardware-encryption schemes into the drive interface itself. This development means that what is stored on the disk is encrypted. To an authorized user, it appears to be a normal drive, but to a data thief or hacker, the coding makes it indistinguishable from an unformatted disk.

This hardware-based approach can produce a completely "transparent" encryption system operating at full disk data rates. Just as important, the operation of the drive and security system can, with the right authentication scheme, be made more or less independent of the hardware platform and operating system with which it will be used.

Architecture for implementing software password authentication. Source: Oxford Semiconductor

Manufacturers of drive interface devices are now beginning to offer chip-level solutions for such encryption tasks. Some products, for example, now include USB2.0-to-dual-SATA and Firewire/USB2.0-to-dual-SATA devices with embedded real-time 128-bit AES encryption. These products are often certified to National Institute of Standards and Technology (NIST) standards, allowing them to be used in FIPS-140-compliant encryption products for U.S. federal government agencies.

Some of the products on the market today are hardware-based, which means that devices can easily keep up with the data-transfer rates of Firewire 800. In some cases, it is possible to encrypt all of the data on the disk (so-called full disk encryption, or FDE), a scheme with an inherently higher level of security than, for instance, drives that are simply deactivated until the correct password is entered. It is the transparent nature of the system that additionally makes it possible to use hardware-encrypted drives in RAID architectures that utilize disk striping, disk mirroring and disk spanning.

However, the ready availability of a hardware-encryption implementation embedded in the controller device is only half the story. The finished drive must also include an appropriate authentication system. There are several products on the market that can support three distinct approaches: secure token, biometrics and password control.

Architecture for providing biometric authentication. Source: Oxford Semiconductor

This situation presents drive manufacturers with a further challenge. If encryption technology is one step removed from their core competence, then the skills and knowledge required to design an image-capture system for fingerprints, or to write a host application for password verification, are a giant leap away.

As a result, manufacturers are now providing a high level of support, from low-level software to fully featured reference designs. Frequently, these developments are the fruit not just of design and programming expertise, but also of inputs from partner companies that can supply the relevant products and integration knowledge.

Secure token authentication, while still providing a high level of security, is probably the simplest approach of all: There are solutions for such architecture that have been built in conjunction with iButton. The drive is enabled and ready for use as soon as the token (essentially an electronic key) is inserted; no special drivers or applications on the host are required. Disk utility software, such as partition managers, disk imagers and defragmenters, will also work transparently.

Token systems are commonly found in end uses, such as electronic point-of-sale terminals and tills. The 128-bit security key is embedded in the token and cannot be extracted other than by brute-force attack, a process that could reasonably be expected to take upwards of 1,011 years; furthermore, there is no information about the key stored on the drive.

Although less secure than token-based authentication, biometric techniques, particularly fingerprint recognition, have their own unique attractions. In addition to engendering a high level of user confidence, fingerprints cannot be lost as keys can, or forgotten as passwords can. Biometric techniques are also (theoretically) unique, which is a great advantage in systems designed to only give access to a restricted number of authorized users.

IC manufacturers have collaborated with manufacturers of fingerprint-recognition modules to produce a complete reference design for drive manufacturers. Again, the systems-integration task is multidisciplinary. A host application is needed to enroll users, although after training, the system is host- and OS-independent. In use, fingerprints are scanned via a sensor and digitized. Software then identifies key features in the digitized image and attempts to produce a match with known, authorized prints stored within the sensor's companion chip.

Fingerprint recognition can also be supplemented with the third form of authentication: password control. This process can be used to cover the possibility that prints can be temporarily or permanently compromised by, for instance, wear caused by manual labor. Moreover, the inherent trade-off between false reject rate (FRR) and false acceptance rate (FAR) can lead to the system rejecting genuine users. The FAR itself gives a measure of the level of security the system achieves, a measure that is typically around one in 10,000.

Password-only authentication continues to grow in popularity, largely because it is easy for users to understand and relatively simple (and cost-effective) to implement; it also provides a level of security that is acceptable for most applications. On the other hand, it is dependent on an application run on the host. As with the other forms of encryption, many companies now offer a reference design for password authentication.

One significant feature of any password-based system is its ability to withstand dictionary attacks. This feature can be enhanced by using a password-based key-derivation function, rather than using the password directly as the security key. Such functions combine the password with a 64-bit random number to generate the 128-bit key. The randomization process means that an attacker cannot prepare a list of key combinations to try to break in; this obstacle in turn increases the time needed for each password "try" to the order of seconds, effectively making it impractical to mount an attack.

The use of encryption to ensure the integrity of personal data looks certain to increase. In addition to its existing products and reference designs, there are companies on the market today that have announced plans for next-generation disk interface chips to be equipped with encryption. n

Jacqui Adams (Jacqui.adams@ oxsemi.com) is marketing manager for storage solutions at Oxford Semiconductor (Abingdon, U.K.). She holds a BEng in electronic systems and microcomputer engineering.