I think software devs could learn something from web security then..
the only way to resolve this is for software to stop storing keys in memory; but will need some sort of session identifier in order to maintain security etc.
Interesting times!
This story was updated on Tuesday 1st April 2008 to correct inaccuracies about DaisyDukes. It works on memory dumps or live memory. At the moment, it is not memory dumper. Building off recent research that showed how to extract encryption keys from a computer's memory, a penetration testing company has unveiled a prototype of a …
So, lets see...
I now have to set the following BIOS fields:
Chassis Intrusion Detection = Enabled
Chassis Intrusion Action = Reboot
Reboot RAM test = Aggressive
Boot from USB = Disabled
BIOS Password = Enabled
BIOS Password Strength = Very Strong
But, I still have no method of stopping someone simply pulling the plug then getting the RAM out and into another machine before it fades. Time for an on-board battery that is used to wipe the RAM if power is lost.
What about PIE SSP?
If every stack is loaded into a different place and compiled as such then surely this kind of attack cannot work.
Admittedly you'd have to use a FUSE based file system (so much for a monolithic kernel being good) but you can maintain security against this and many other attacks.
Anyone who has suggested disabling USB booting is missing the point: the program could as easily be run from a bootable CD. Possibly even a floppy.
Clearing/over-writing the RAM is just masking the problem - certainly not a 'long term solution'. Passwords simply shouldn't be stored in memory. Jeff may be right on this one, Interesting times...
Er, how?
You read a disk sector, not individual bytes. I don't think you can keep that chunk of data out of RAM without a fundamental redesign of the whole chain from disk platter to CPU registers.
When better physical security, and some BIOS changes, could stop such attacks, why go to such lengths.
Possible the big risk is the inside job?
Jeff - if you look at the actual security risks this sort of attack adds (disregarding physical security), it's primarily drive encryption that can be compromised. And drive encryption requires the key in memory, and somewhere the software can find it.
BIOSes that clear RAM ('full memory test' in POST should do it) would prevent this particular attack.
1. why there is a distinct signature for passwords in memory.
2. why passwords aren't wiped from memory when I lock the machine.
3. why el Reg comments don't provide a 'refer to previous comment' option.
4. why my local supermarket doesn't have fresh bread on Mondays.
1. so your computer can find them.
2. a.Because they will be needed to restart the session.b. Because it's being turned off why bother they will fade anyway slow/hung shutdown can also be a problem.
3. Because we are a pain in their collective asses, and we don't need encouragement.
4. Because they always sell their fresh bread to convenience stores first next day they sell whats left over to the grocery stores. The reason is they get more money for their product at convenience stores and restaurants. The account is worth more gets quicker service. Go to a 7-11 and get that $8.00 loaf of white bread it's fresh.