backtop


Print 37 comment(s) - last by Jack Ripoff.. on Feb 6 at 9:35 AM

And so they forged the encryption of power, one encryption to rule them all

With reports of sensitive and, at times, top-secret information being lost on the hard drives of notebook computers, keeping data safe is one of the most important things for business and consumers today. One problem is that hard drive makers often used their own encryption format, which made things confusing for the consumer.

ComputerWorld reports that hard drive makers have now agreed to use the same encryption method for full-disk encryption (FDE) that can be used across all brands of hard drives and SSDs. When FDE is enabled, the computer requires a password before it will boot and all data on the drive is encrypted.

The final specifications for the encryption standard were published this week by the Trusted Computing Group (TCG) and cover specs for FDE in notebooks, desktop and server applications. Robert Thibadeau from Seagate said, "This represents interoperability commitments from every disk drive maker on the planet. We're protecting data at rest. When a USB drive is unplugged, or when a laptop is powered down, or when an administrator pulls a drive from a server, it can't be brought back up and read without first giving a cryptographically-strong password. If you don't have that, it's a brick. You can't even sell it on eBay."

Settling on one single encryption standard will allow all drive makers to build security into all products, which will lower the cost of production and make it easier for user to secure the data on their computers.

This is big news for enterprise environments where a standard encryption protocol means less configuration and less hassle during installation along with less management down the road. The specifications allow encryption to be set by administrators and can’t be turned off by end-users.

One very important factor is that modern FDE has come a long way and now only marginally effects read-write speeds of hard drives. Writing data to an encrypted drive is almost as fast as writing data to a non-encrypted drive. The companies that are members of the TCG include Fujitsu, Hitachi GST, Seagate Technology, Samsung, Toshiba, Western Digital, Wave Systems, LSI Corp., ULink Technology, and IBM.

Analyst Jon Oltsik from Enterprise Strategy Group said, "In five years time, you can imagine any drive coming off the production line will be encrypted, and there will be virtually no cost for it."

The three specifications for FDE includes the Opal spec for outlining minimum requirements for a storage device in a PC or laptop. The Enterprise Security Subsystem Class Specification is aimed at drives in data centers where minimum security configuration is needed during install. The final spec is the Storage Interface Interactions Specification, which details how the specifications interact with other standards for storage interface.

The specification supports PATA and SATA, SCSI SAS, Fibre Channel, and ATAPI. The three larger members of the group -- Seagate, Fujitsu, and Hitachi -- are already producing drives that support the standard. The specifications call for vendors to choose to use either AES 128-bit or AES 256-bit keys depending on the level of security wanted. The group points out that neither of these standards has been broken.



Comments     Threshold


This article is over a month old, voting and posting comments is disabled

speeds
By Screwballl on 1/29/2009 10:21:29 AM , Rating: 2
quote:
One very important factor is that FDE doesn't affect the performance of the hard drive. Writing data to an encrypted drive is as fast as writing data to a non-encrypted drive.


What? I have never seen anything shy of 32 or 64-bit real time encryption with speeds anywhere close as non-encryption... especially with 128 and 256-bit encryption schemes, there is ALWAYS some performance reduction, and many times the impact is over 30%.




RE: speeds
By JasonMick (blog) on 1/29/2009 10:32:07 AM , Rating: 2
I think what he meant is that a lot of drive manufacturers advertise encrypted drives as being "virtually as fast" as nonencrypted drives or other sticky phrases. Its typical marketing speak. For example NVIDIA bragged about how great their new 3D gaming vision is, and how its "easily supported" and it was only when I prodded them that they admitted it yields a 30 percent performance hit.

However, in this case, I think the benefits of encryption far outweigh performance drops. In general, write speeds do impact performance marginally, but compared to graphics cards and processors, HDDs are seldom the bottleneck in the system. However, failing to secure your data can be very costly on the other hand, and I wouldn't want to give up a few frames-per-second for that.


RE: speeds
By jmke on 1/29/2009 10:40:40 AM , Rating: 5
quote:
However, in this case, I think the benefits of encryption far outweigh performance drops. In general, write speeds do impact performance marginally, but compared to graphics cards and processors, HDDs are seldom the bottleneck in the system.


what?? HDDs are the major bottleneck of systems! Most likely you'll want your laptop's HDD encrypted; in laptops the speed of the HDD pretty much dominates its performance level, adds encryption and it will run remarkably slower; add a faster HDD and it will work remarkably snappier and faster.

HDDs are pretty much the only component in current gen systems which is not as fast as it should be...


RE: speeds
By bobsmith1492 on 1/29/2009 10:43:58 AM , Rating: 4
Although with Vista's memory management, if you have enough RAM, it's only an issue on startup. Otherwise most of your applications and data can be cached in RAM. I'm sure Windows 7 will be the same or better.


RE: speeds
By Oregonian2 on 1/29/2009 10:46:19 AM , Rating: 5
Encryption will only be "remarkably slower" if it's done in software. The HDD makers will do it in hardware which will easily work in real time with gates and flops tied behind it's back.


RE: speeds
By BladeVenom on 1/29/2009 11:43:07 AM , Rating: 2
Even software encryption isn't that big of a penalty. Here's a recent test at Tom's of Truecrypt. http://www.tomshardware.com/reviews/truecrypt-secu...


RE: speeds
By JasonMick (blog) on 1/29/2009 10:50:06 AM , Rating: 3
I understand what you're saying but that's highly subjective and depends on your application. It really depends on your application. In gaming, on laptops HDDs are definitely NOT the bottleneck. Perhaps they are a considerable one, but by far the underpowered graphics cards on most laptops are the biggest bottleneck. Laptop graphics are only now FINALLY starting to get up to snuff. Most people don't realize that the average card features a mere half to a fourth of the processing units of its similarly named PC brethren.

I'm sure HDD performance is much more critical for database or other storage-intensive apps, though (I'd say you're 100 percent right if you meant this). I would agree that laptops are most vulnerable to encryption performance hits, but on business notebooks, these losses aren't as critical, but protection is.

PC HDDs are of course faster, as are ones in server setups so losses are smaller and less of a concern.


RE: speeds
By icanhascpu on 1/29/2009 4:03:42 PM , Rating: 3
quote:
PC HDDs are of course faster, as are ones in server setups so losses are smaller and less of a concern.


Yet still by far the biggest bottleneck in either setup. It does not depend on the application. It does not depend on anything. Why? Becuse as soon as a system needs to access the hard drive, the bottleneck presents itself. Thats exactly the point. I dont know what youre talking about or if youve had one too many appletinies.


RE: speeds
By gss4w on 1/30/2009 3:09:12 AM , Rating: 1
HDD are certainly a bottleneck in some scenarios, particularly loading applications. However, much of the problem is caused by rotational and seek latency, and adding encryption will not increase the latency.

Also with software encryption such as Bitlocker on Vista the data is read from the HDD and then decrypted in memory. So the performance of reading from the HDD is not degraded, although there is some additional delay to do the encryption/decryption. However, the encryption operations in memory are so much faster than the HDD that it ends up being negligible.

In summary, yes the HDD is a major bottleneck, but adding encryption does not really make the bottleneck any worse because the encryption operations are so much faster than the R/W from the HDD in the first place.


RE: speeds
By SilentSin on 1/29/2009 10:43:23 AM , Rating: 2
quote:
n general, write speeds do impact performance marginally, but compared to graphics cards and processors, HDDs are seldom the bottleneck in the system.


??? HDDs are by far the slowest internal part of the system and therefore the bottleneck. Why do you think there is such a big push to move towards SSDs with far higher bandwidth? Your case may work for gaming where there is very little writing being done, but as for just about anything else that's just not true. Even systems with real time virus checking have very noticeable slowdown when you're installing a big application, encryption would just add to that delay. I'm not saying encryption is unusable and it definitely should be used for any sensitive data regardless of performance, but to say that a GPU or CPU is a "slower" part in that chain is just false.


RE: speeds
By JasonMick (blog) on 1/29/09, Rating: 0
RE: speeds
By TomZ on 1/29/2009 11:17:30 AM , Rating: 3
Yes of course HDDs are the slowest component by design, but they are also a bottleneck in many typical computer operations. The OP is right about that. It makes sense to pay attention to HDD performance if you are looking to speed up your machine overall.


RE: speeds
By ImSpartacus on 1/29/2009 10:33:17 AM , Rating: 2
Yeah, I want to see some third party numbers on this.


RE: speeds
By Oregonian2 on 1/29/2009 10:42:29 AM , Rating: 2
Why should hardware based real-time encryption in the drives slow down drive performance at all (which certainly is what they are doing, else the hard drive makers themselves would not have to be involved).

It's just adding a "streaming" encryption/decryption unit to the controller IC with means of programming in the key. The hardware involved isn't all that difficult or even all that much circuitry in modern terms. The only thing that needs to be standardized (what they've done) is the means to control the mechanism (putting in the key, for instance) -- how it is done internally doesn't otherwise matter.


RE: speeds
By ninus3d on 1/29/2009 10:50:33 AM , Rating: 2
If that is the case, isnt a little circuitry changes all that would be needed to work around the security?
Or, if you meant that this would "randomize" the data on the drive and would need to decrypt it with the actual circuit intact then this would atleast require some calculation and this would cause further strain on the maximum performance a HDD otherwise could give.


RE: speeds
By TomZ on 1/29/2009 10:58:35 AM , Rating: 2
Huh? Think of it this way. Unecrypted data comes into the microcontroller/chip via SATA. The chip encrypts the data (on-chip), then writes encrypted data to the platters.

The point really is that the microcontroller performs the encryption/description via hardware (fast) instead of by software (slow).


RE: speeds
By Oregonian2 on 1/29/2009 11:51:56 AM , Rating: 3
It just means that the (probably) same key needs to be programmed in for the reading (at boot time) that was there when the disk was writing that data. Once the password derived key is programmed into the drive, the drive will act like normal unencrypted ones -- but until that time, it'd be useless.

Encryption hardware would not have to "stop data and calculate", it's a pile of flops in a pipelined pile of circuitry (I'm a hardware engineer of over thirty years experience -- I've designed FPGA/ASIC stuff using synthesis languages (like Verilog or VHDL) where 100% of everything is real time because it's hardware, not software).

Think of having a black box that takes data in and out at the same "full speed" rate, but with only a latency delay that's utterly insignificant compared to other delays in the system.


RE: speeds
By PrinceGaz on 1/29/2009 11:56:56 PM , Rating: 2
Exactly. I expect these drives, together with suitably aware BIOS or EFI, will prompt the user for the HD password on bootup so that it can read it (there might be the option for the password to be stored by the mobo for future bootups, though that obviously compromises security but for most users would be preferable).

In terms of speed, drives with this encryption would be as fast as non-encrypted drives. The overhead imposed by hardware encryption is negligible compared with the access times of magnetic hard-drives.

So long as the extra hardware involved isn't too expensive (it should be fairly trivial) and the encryption is optional, it's hard to find a downside except perhaps that should data need to be recovered from a drive which has failed, it could be more difficult.


RE: speeds
By TomZ on 1/29/2009 10:48:33 AM , Rating: 2
quote:
What? I have never seen anything shy of 32 or 64-bit real time encryption with speeds anywhere close as non-encryption... especially with 128 and 256-bit encryption schemes, there is ALWAYS some performance reduction, and many times the impact is over 30%.
There are many embedded microcontrollers that have cryptography accelerators implemented in hardware. And I would guess with this degree of industry-wide standardization, that more microcontrollers will be developed that specifically optimize the type of encryption/decryption needed for this job.

Bottom line is that this is not a hard problem to solve, especially for an industry that purchases tens or hundreds of millions of microcontrollers every year.


RE: speeds
By Oregonian2 on 1/29/2009 11:59:26 AM , Rating: 2
No real need for it to use a microcontroller, it'd probably be pure hardware. The acceleration add-on that many microcontrollers have (I've worked on projects using those processors) is I think done that way because the datastream is "passing through" software, so to speak, and is being stripped of protocol overhead with only the "proper" portion being sent (by software) through the decryption. In terms of the hard disk, the protocol is fixed, probably simple, and would best be implemented in a hw design for least cost, probably integrated into existing controller ICs.


RE: speeds
By TomZ on 1/29/2009 3:28:17 PM , Rating: 2
quote:
No real need for it to use a microcontroller, it'd probably be pure hardware.
I'm not a HDD designer, but as I understand it, they already are using MCUs to interpret the ATA commands and to control the other drive electronics. After all, HDDs wouldn't need "firmware" updates if they didn't currently use MCUs, would they?

But I'm just guessing, since as I said, I don't work in that field.


RE: speeds
By Oregonian2 on 1/30/2009 3:16:23 AM , Rating: 3
Things like ATA commands are transaction level things. Things on the control plane (I've been in networking hardware design for quite some time -- I'm not sure if hard drive designers use the same terminology). The encryption itself is in the data plane. Encryption is done on the data at it's speed which is a lot faster than control plane things that happen much less often. That makes the MCU's more appropriate for the use you mention and hardware implementation for things like encryption and error correction.


RE: speeds
By OrSin on 1/29/2009 12:49:51 PM , Rating: 2
30%? are you kiding. I have test at least a half dozen produces and not a single one showed more then 6% performance shot. The HHD is a preformance bottleneck, but its the hard drive speed not the system processing it.
All the encrytion and decrytion is done by the CPU, so the data read and wrote is not slowed down by the HHD speed at all. People throw out numbers and have not clue what they are talking about. Anyone that says other wises, name a single product they notice this 10-30% slow down in. People are so full of crap.


RE: speeds
By gstrickler on 1/29/2009 2:34:59 PM , Rating: 2
quote:
...especially with 128 and 256-bit encryption schemes, there is ALWAYS some performance reduction, and many times the impact is over 30%.
If you had bothered to check any of the third party benchmarks that have been done on the existing Seagate or Hitachi FDE drives, you would know that there is no measurable performance penalty with the hardware encryption on these drives. Technically, there might be a microsecond of additional latency, but compared to the typical multi-millisecond latency in a HD system, the result is not measurably different than the same drive without encryption.

Further, with modern CPUs, even software encryption can be done with only 2%-5% performance penalty. Go check out reviews of TrueCrypt.


"It looks like the iPhone 4 might be their Vista, and I'm okay with that." -- Microsoft COO Kevin Turner

Related Articles













botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki