Now that should scare everyone who uses flash based storage, and who has important data to destroy. whatever their reason. Oh, well use a hammer then, someone will glibly say. Trouble with that is that most people don't know that the problem even exists and will rely on a single whole drive overwrite, which is usually effective on hard drives. But on SSDs and their ilk, it just isn't..
The complete solution for SSDs (and for hard drives, come to that) is full disk encryption, from the outset. Given that many modern processors have hardware support for AES, and even those that don't have cycles to spare, there's no reason not to. When you buy a computer, turn on FDE before you put any of your own data onto it. The configuration of that FDE might get more exotic as you consider more exotic adversaries, but if you've got data with those sorts of data at rest security requirements you've also got people who know what they're doing to help you get it right.
However, "you can extract all the erased data from an SSD" is this year's version of the usual obfuscation. It simply isn't true. The alleged threat is that SSDs over provision, and silently replace blocks that appear to be approaching their limit with blocks drawn from the list of spares. The rate at which this happens is relatively low, but is distinctly non-zero, so an adversary who has physical access to the drive and is willing to dismantle it and install additional hardware --- ie, a motivated, resourced adversary, not a bloke buying drives off EBay --- can recover portions of the data by looking at the original blocks that have subsequently been remapped. Simply over-writing the drive from the OS can't erase those blocks, as they are no longer part of the addressable list of blocks.
However, in 2013 that's a partial account. Leave aside ATA Secure Erase, on the grounds that it's possible that people don't or can't use it. Leave aside the fact that the over-provisioning is around 10% and will only be used over the later part of the lifetime of the drive. Every drive sold today, and every modern operating system, supports TRIM. I'm guessing the document you're talking about is [1] which builds on [2]: the latter paper dismisses TRIM in a single footnote, which is ambitious at best. TRIM is being worried about as the end of digital forensics in the very document I think you're citing, because it provides a relatively secure overwrite shortly after deletion.
So if you happened to get an unencrypted SSD on the last day of its useful life, which has been overwritten, then after physically disassembling the drive you you might be able to recover up to 10% of its capacity, which will be drawn from the more recent overwrites, provided TRIM --- which is supported by every modern drive, and every modern OS --- hasn't been used. It's a substantial risk, and one which makes OS-level FDE very attractive, but it's a long way from "the shadowy organisations can recover all your data even if it's been overwritten".
The complete solution for SSDs (and for hard drives, come to that) is full disk encryption, from the outset.
Microsoft promote their BitLocker technology for servers and clients for that very purpose: Force all PCs to have encrypted disks via Group Policy, and then just delete the encryption keys prior to hardware disposal.
Microsoft promote their BitLocker technology for servers and clients for that very purpose: Force all PCs to have encrypted disks via Group Policy, and then just delete the encryption keys prior to hardware disposal.
That's sort of what we do in our school with laptops provided to teachers. The whole hard drive is encrypted using Compusec, and only we (the technicians) know who has what password. However when a drive goes pop, there's absolutely no way back as we can't retrieve any of the data, but it makes the teachers learn a lesson...don't have just one copy of your data.
That's sort of what we do in our school with laptops provided to teachers. The whole hard drive is encrypted using Compusec
Why not Bitlocker? It's free, it makes effective use of a TPM if fitted and it's accredited for government IL3, which is substantially above anything a school could possibly have.
Why not Bitlocker? It's free, it makes effective use of a TPM if fitted and it's accredited for government IL3, which is substantially above anything a school could possibly have.
BitLocker is only in Windows 7 Ultimate and Enterprise which not everyone uses. It also needs a good Group Policy infrastructure to work properly.
Why not Bitlocker? It's free, it makes effective use of a TPM if fitted and it's accredited for government IL3, which is substantially above anything a school could possibly have.
Because a lot of laptops use are still on Windows XP. Plus we've only got licenses for Windows 7 Professional for desktops and laptops.
It's a shame that this thread can never be fully deleted from the entire WWW
No, it would be a shame if it was.
DS computer forum can get dull at times and threads like this can bring it back to life.
This post from a related topic had me laughing aloud very early this morning and I'm usually a grump in the morning ......
RobertCrowther, he's the poster that just keeps giving! ..... secret agent, web designer, professional witness, secret agent, IT tech support, data recovery expert, security consultant etc etc......
Oh my, what happened to RobertCrowther. He was a tech God and helped me iat least twice. I'd adopt him if I could, so that he could tell me everytime I was wrong!
Oh my, what happened to RobertCrowther. He was a tech God and helped me iat least twice. I'd adopt him if I could, so that he could tell me everytime I was wrong!
I'm sure if he were here, he would take great pains to point out that it is 'robertcrowther'.
Failure to note the appropriate case indicates a lack of IT/CS ability.
I'm sure if he were here, he would take great pains to point out that it is 'robertcrowther'.
Failure to note the appropriate case indicates a lack of IT/CS ability.
Ok robertcrowther. I apologise I didn't look up his name properly and typed it with standard capitalization. I don't mind admitting my lack of IT/CS ability, whatever that means. That's why I come on 'ere for advice.
Ok robertcrowther. I apologise I didn't look up his name properly and typed it with standard capitalization. I don't mind admitting my lack of IT/CS ability, whatever that means. That's why I come on 'ere for advice.
As bob would no doubt be quick to point out, you actually used Pascal case. Normal capital case would demand a gap between the two words.
The complete solution for SSDs (and for hard drives, come to that) is full disk encryption, from the outset. Given that many modern processors have hardware support for AES, and even those that don't have cycles to spare, there's no reason not to. When you buy a computer, turn on FDE before you put any of your own data onto it. The configuration of that FDE might get more exotic as you consider more exotic adversaries, but if you've got data with those sorts of data at rest security requirements you've also got people who know what they're doing to help you get it right.
However, "you can extract all the erased data from an SSD" is this year's version of the usual obfuscation. It simply isn't true. The alleged threat is that SSDs over provision, and silently replace blocks that appear to be approaching their limit with blocks drawn from the list of spares. The rate at which this happens is relatively low, but is distinctly non-zero, so an adversary who has physical access to the drive and is willing to dismantle it and install additional hardware --- ie, a motivated, resourced adversary, not a bloke buying drives off EBay --- can recover portions of the data by looking at the original blocks that have subsequently been remapped. Simply over-writing the drive from the OS can't erase those blocks, as they are no longer part of the addressable list of blocks.
However, in 2013 that's a partial account. Leave aside ATA Secure Erase, on the grounds that it's possible that people don't or can't use it. Leave aside the fact that the over-provisioning is around 10% and will only be used over the later part of the lifetime of the drive. Every drive sold today, and every modern operating system, supports TRIM. I'm guessing the document you're talking about is [1] which builds on [2]: the latter paper dismisses TRIM in a single footnote, which is ambitious at best. TRIM is being worried about as the end of digital forensics in the very document I think you're citing, because it provides a relatively secure overwrite shortly after deletion.
So if you happened to get an unencrypted SSD on the last day of its useful life, which has been overwritten, then after physically disassembling the drive you you might be able to recover up to 10% of its capacity, which will be drawn from the more recent overwrites, provided TRIM --- which is supported by every modern drive, and every modern OS --- hasn't been used. It's a substantial risk, and one which makes OS-level FDE very attractive, but it's a long way from "the shadowy organisations can recover all your data even if it's been overwritten".
My point has been that most people don't know most of what you and I have written and will tend to use SSDs like hard drives when wanting to securely erase data. Whether or not over-writing even a complete drive works effectively seems to depend very much on pot luck, no matter how new or old the drive is (not just old ones as you claim).
The biggest problem with flash storage (including SD cards and usb sticks) is that there are gaps in the standards, some standards aren't universally applied, their workings are complex and act as a barrier to understanding. Flash-type storage like usb sticks and SD cards do not usually support the secure erase commands, or TRIM. Neither do some O/Ss or FAT32 etc! Even on SSDs, the security commands are sometimes not implemented correctly, according to the usenix.org paper.
Your point about extracting all the erased data is moot - because I haven't said that or anything like it. I've said that experts can often, with somewhat complex but fairly inexpensive tools, extract some useable data following a full-disk (flash disk) over-write, the amount recovered being unpredictable and seemingly depending on pot luck. SD cards and usb sticks (the subject of this thread) seem to be by far the least secure but fully over-writing them at least twice (from the usenix.org paper) usually works well. Or just use a hammer.
It is interesting to note from the forensics paper you linked, that forensics too can have problems with SSDs - the data *may* be completely gone and unrecoverable - but equally, from the other paper, some of it may be still there when someone *thinks* it's securely gone... and therein lies the problem. It's all a bit of a mess in fact, but encryption from the word go would as you suggest solve the problem.
Very few people will be able to recover data after even a single "complete drive" over-write, however, if you are are bothered enough to want to protect against experts who can pull it apart and use specialist gear:
Over-write the entire drive (in practice a few big files then smaller and smaller ones until it's filled), permanently delete that, fully over-write it again and then, in most cases apparently, it will all be gone for ever.
Or if it's a "proper" internal SSD supporting the extra commands, use its secure delete, make sure it's using TRIM (if the O/S and interface supports it) and hope that the commands were implemented properly by the drive makers.
Best to do both if you can, for anti-expert protection. Or physically destroy it.
Flash-type storage like usb sticks and SD cards do not usually support the secure erase commands, or TRIM.
That's right, and even if the devices did support them, the API for them would be non-standard in the extreme. However, if you're worried about people recovering data from your discarded SSD or by people dismantling your machine while you aren't looking, surely it goes without saying that you're encrypting external removable devices, either by using OS disk encryption (OSX Filevault will encrypt USB sticks, and I assume it's the same for Windows, and there's always Truecrypt) or by using equipment with hardware encryption provided.
Although the early ones were hideously insecure, the recent versions of the Corsair Padlock 2 look very sound if you need a USB stick to be secure and you can't rely on the OS facilities.
I've found a couple of really, really weird edge cases where they aren't a drop-in replacement for an unencrypted drive (I couldn't boot an old Sun X2200 and install Solaris 11 off one, for example) but I use one for shifting presentations onto projectors that want a FAT32 filesystem, and it works perfectly.
robertcrowther in post 88 wrote :-
All software that is available to the public that erases data leaves a fingerprint (id) on the hard disk/usb stick, this is a requirement in law and any software that does not is not permitted.
Errr so if I used software that was illegal because it didn't leave a "fingerprint" then I would be prosecuted for so doing. How could it be proved that the software that leaves no trace had been used?????
robertcrowtherin post 88 wrote :-
All software that is available to the public that erases data leaves a fingerprint (id) on the hard disk/usb stick, this is a requirement in law and any software that does not is not permitted.
Errr so if I used software that was illegal because it didn't leave a "fingerprint" then I would be prosecuted for so doing. How could it be proved that the software that leaves no trace had been used?????
Please don't start that up again :eek: as it was knocked well and truly on the head pages ago!
However, Jason JAG Law has told me that, though it's out of his area of specialisation, he can't imagine such a UK legislation to exist but he's be happy to be proven wrong.
Comments
The complete solution for SSDs (and for hard drives, come to that) is full disk encryption, from the outset. Given that many modern processors have hardware support for AES, and even those that don't have cycles to spare, there's no reason not to. When you buy a computer, turn on FDE before you put any of your own data onto it. The configuration of that FDE might get more exotic as you consider more exotic adversaries, but if you've got data with those sorts of data at rest security requirements you've also got people who know what they're doing to help you get it right.
However, "you can extract all the erased data from an SSD" is this year's version of the usual obfuscation. It simply isn't true. The alleged threat is that SSDs over provision, and silently replace blocks that appear to be approaching their limit with blocks drawn from the list of spares. The rate at which this happens is relatively low, but is distinctly non-zero, so an adversary who has physical access to the drive and is willing to dismantle it and install additional hardware --- ie, a motivated, resourced adversary, not a bloke buying drives off EBay --- can recover portions of the data by looking at the original blocks that have subsequently been remapped. Simply over-writing the drive from the OS can't erase those blocks, as they are no longer part of the addressable list of blocks.
However, in 2013 that's a partial account. Leave aside ATA Secure Erase, on the grounds that it's possible that people don't or can't use it. Leave aside the fact that the over-provisioning is around 10% and will only be used over the later part of the lifetime of the drive. Every drive sold today, and every modern operating system, supports TRIM. I'm guessing the document you're talking about is [1] which builds on [2]: the latter paper dismisses TRIM in a single footnote, which is ambitious at best. TRIM is being worried about as the end of digital forensics in the very document I think you're citing, because it provides a relatively secure overwrite shortly after deletion.
So if you happened to get an unencrypted SSD on the last day of its useful life, which has been overwritten, then after physically disassembling the drive you you might be able to recover up to 10% of its capacity, which will be drawn from the more recent overwrites, provided TRIM --- which is supported by every modern drive, and every modern OS --- hasn't been used. It's a substantial risk, and one which makes OS-level FDE very attractive, but it's a long way from "the shadowy organisations can recover all your data even if it's been overwritten".
[1] http://forensic.belkasoft.com/en/why-ssd-destroy-court-evidence
[2] http://static.usenix.org/events/fast11/tech/full_papers/Wei.pdf
Microsoft promote their BitLocker technology for servers and clients for that very purpose: Force all PCs to have encrypted disks via Group Policy, and then just delete the encryption keys prior to hardware disposal.
Why not Bitlocker? It's free, it makes effective use of a TPM if fitted and it's accredited for government IL3, which is substantially above anything a school could possibly have.
No, it would be a shame if it was.
DS computer forum can get dull at times and threads like this can bring it back to life.
This post from a related topic had me laughing aloud very early this morning and I'm usually a grump in the morning ......
The sooner RC is back the better
Failure to note the appropriate case indicates a lack of IT/CS ability.
Ok robertcrowther. I apologise I didn't look up his name properly and typed it with standard capitalization. I don't mind admitting my lack of IT/CS ability, whatever that means. That's why I come on 'ere for advice.
As bob would no doubt be quick to point out, you actually used Pascal case. Normal capital case would demand a gap between the two words.
Alright I'm gonna sayI forgit to press space. My touch type lacks much.
Who's bob bTw?
Must be Robert.
Duurrrrrrrrrrrrr. I was like Bob, bob who, bob? Bob??
Robert! I didn't make the connection!
My point has been that most people don't know most of what you and I have written and will tend to use SSDs like hard drives when wanting to securely erase data. Whether or not over-writing even a complete drive works effectively seems to depend very much on pot luck, no matter how new or old the drive is (not just old ones as you claim).
The biggest problem with flash storage (including SD cards and usb sticks) is that there are gaps in the standards, some standards aren't universally applied, their workings are complex and act as a barrier to understanding. Flash-type storage like usb sticks and SD cards do not usually support the secure erase commands, or TRIM. Neither do some O/Ss or FAT32 etc! Even on SSDs, the security commands are sometimes not implemented correctly, according to the usenix.org paper.
Your point about extracting all the erased data is moot - because I haven't said that or anything like it. I've said that experts can often, with somewhat complex but fairly inexpensive tools, extract some useable data following a full-disk (flash disk) over-write, the amount recovered being unpredictable and seemingly depending on pot luck. SD cards and usb sticks (the subject of this thread) seem to be by far the least secure but fully over-writing them at least twice (from the usenix.org paper) usually works well. Or just use a hammer.
It is interesting to note from the forensics paper you linked, that forensics too can have problems with SSDs - the data *may* be completely gone and unrecoverable - but equally, from the other paper, some of it may be still there when someone *thinks* it's securely gone... and therein lies the problem. It's all a bit of a mess in fact, but encryption from the word go would as you suggest solve the problem.
Over-write the entire drive (in practice a few big files then smaller and smaller ones until it's filled), permanently delete that, fully over-write it again and then, in most cases apparently, it will all be gone for ever.
Or if it's a "proper" internal SSD supporting the extra commands, use its secure delete, make sure it's using TRIM (if the O/S and interface supports it) and hope that the commands were implemented properly by the drive makers.
Best to do both if you can, for anti-expert protection. Or physically destroy it.
That's right, and even if the devices did support them, the API for them would be non-standard in the extreme. However, if you're worried about people recovering data from your discarded SSD or by people dismantling your machine while you aren't looking, surely it goes without saying that you're encrypting external removable devices, either by using OS disk encryption (OSX Filevault will encrypt USB sticks, and I assume it's the same for Windows, and there's always Truecrypt) or by using equipment with hardware encryption provided.
Although the early ones were hideously insecure, the recent versions of the Corsair Padlock 2 look very sound if you need a USB stick to be secure and you can't rely on the OS facilities.
I've found a couple of really, really weird edge cases where they aren't a drop-in replacement for an unencrypted drive (I couldn't boot an old Sun X2200 and install Solaris 11 off one, for example) but I use one for shifting presentations onto projectors that want a FAT32 filesystem, and it works perfectly.
All software that is available to the public that erases data leaves a fingerprint (id) on the hard disk/usb stick, this is a requirement in law and any software that does not is not permitted.
Errr so if I used software that was illegal because it didn't leave a "fingerprint" then I would be prosecuted for so doing. How could it be proved that the software that leaves no trace had been used?????
Please don't start that up again :eek: as it was knocked well and truly on the head pages ago!
We took that particular topic over into another thread:
http://forums.digitalspy.co.uk/showthread.php?t=1807345
However, Jason JAG Law has told me that, though it's out of his area of specialisation, he can't imagine such a UK legislation to exist but he's be happy to be proven wrong.
He's been sent on 'holiday'.