@Secure (good to see you are happy to use your real name with a URL)
Some good calculations, but unfortunately it misses some key points. I feel like I must have glossed over some “assumed” logic so I’ll explain more in detail.
Before I start with my rant, I’ll just ask the simple question – are you arguing that a slower encryption IS more secure, but the difference doesn’t matter because it would take so long to crack the “less secure” algorithm that it doesn’t matter? If so then we disagree only on time frame of computers and hacking. If not then we disagree on everything.
First of all the bit I agree with – that big $$ helps, but even then it will only help to an extent. The big $$ is for more processing power by forming a grid of computers, where each computer hacks a different part of the encryption (different key range). Every computer you bring on-line effectively reduces the bit range of the encryption. That is - if you only had 2 computers and 32 passwords to check, 1 computer would check the first 16 while the other would check the next 16.
Therefore we have a distinction between actual bit rate of the encryption and effective bit rate.
1 computer = full effective bit encryption
2 computers = minus 1 effective bit encryption
4 computers = minus 2 effective bit encryption
8 computers = minus 3 effective bit encryption
and so on so that with 1,024 computers the bit range has been reduced by 10, taking the encryption to effectively 118 bit encryption.
Obviously adding somewhere between these values reduces the time, but doesn’t calculate as an exact bit reduction.
In historic tests, approximately 100,000 computers were used to crack DES (it took about 22 hours but that was a decade ago).
Let’s round that off to 131,072 as that reduced security bits by 17, taking the 128 bit encryption down to 111 (it reduced DES to effectively 39 bit encryption).
This is the way to speed up encryption “right now”… and this is where either big $$ or zombie PCs would come into the picture. Even then, it is unlikely to be successful with anything but a dictionary or close to dictionary attack.
But secondly, and most importantly, you are still forgetting or leaving out the fact that computers are getting faster. The computer I’m using has 4 cpus, which can process four independent attacks (4 passwords at once) AND computers are faster AND cheaper AND they are bringing in new technology, such as GPU which can offer many-many-many times more parallel hack attempts.
But let’s stick with an approximation that flops double every 18 months… that’s exponential computer speed growth! Exponential growth is scary and should be on the minds of all those who need to protect data. It’s scary because it grows really fast constantly making unthinkable processing tasks possible.
If a computer doubles in speed every 18 months, then the effective bit rate of the encryption will decrease by 1 every 18 months. This means that the time taken to crack a 55 bit encryption this year equals the time it will take to crack 56 bit encryption in 18 months time.
Encrypted data that took 24 hours to crack 18 months ago will only take 12 hours to crack today and 6 hours in 18 month’s time. That’s a very sharp reduction in just 3 years, going from 1 day to 6 hours to crack an encryption.
So in 18 months 128 bit encryption becomes effectively 127, in 15 years it becomes 110 bit. In 150 years time it’s effectively 28 bits (comparable to 28 bit encryption by today’s standards). And with my back-of-the-envelope calculations, that would take about 5 seconds to crack. Obviouly that first point is still an issue - crackers with large arrays of computers can still get the data much sooner than 1 computer processing away.
Anyway - it would still be 128 bit encrypted data, but it would only take 5 seconds to decrypt, which is as good as 28 bit encryption on today’s computers. At the same point in time the 192 bit encryption would still take over 8 hours and 256 bit would take until the end of time…. 1.799*10^23 years… well at least for a couple of years until computers are fast enough again.
But all of this is using the DES calculation times, using a faster algorithm reduces the power of the encryption because that 8 hours with 192 bit encryption would be far less. In fact, the faster the encryption, the sooner that it will be viable to crack the encryption cheap and quickly.
That is to say that a slower encryption method IS more secure key bit for key bit, and the results of this security difference WILL be seen within our lifetimes.
To re-iterate my main point, the first question should be how long in time data should remain secret. Keeping in mind that any encryption halves in power ever 18 months or so, calculations should be made as to how long it will take to crack the encryption at various points in time until the privacy of the data is no longer a concern.
To meet that question you need to be aware of the bit rate AND how fast the encryption algorithm itself takes. Yes it DOES mean approximating the future speed of computers, which is NOT going to be accurate, but at least it is better than just wrongly assuming that a fast encryption method is necessarily better than a slow one given the same bit rate.
