Score:2

What is the global resource expense (financial, energy, computing power & time) due to the need for cryptography?

vn flag

Classic disclaimer: there may be a better place to ask this question, if so comment and I will be happy to move it.

An example I'm looking for is related to recent scrutiny over energy consumption of bitcoin, such as here although there is no math in this example.

An example similar in nature is the math behind the question, "how long would it take to crack 128 AES"? -- Example 1 -- Example 2- from this forum

I am curious about the math, the numbers on this. In a perfect world with absolutely zero criminals, zero cyber crime, no identity theft, no data theft or sale, privacy breach etc, ad infinitum, we would not need any cryptography whatsoever. Of course a world like above is not possible to achieve, but it does beg the question, in an ever growing energy starved and energy concerned/paranoid society such as earth:

As of the year 2021 proceeding forward (as many years as you feel to calculate into the future), on a yearly statistical basis:

What is the expenditure of cryptography worldwide in each of the following categories:

  • Electrical energy consumption
  • Electrical energy cost in USD
  • Total time "wasted". Measured as non-concurrent cpu time.

In your calculation do not include (unless separate) any cryptocurrency numbers as this is dis-related to the pure question: in a perfect world with no need for cryptography relating to information tech.

Justifying "3 answers". Since arriving at one stable datum of calculation for the above, such as Total Time one could easily corollary or extrapolate the other two. I think posting a separate question is redundant and these give further data as well as further answer the question: What is the expenditure of cryptography worldwide...

UPDATED- simplified the question a little & justify how "3 answers" are extrapolations/corollary of one answer.

fgrieu avatar
ng flag
The wording looks much like a copy/paste of an interesting assignment. That would make the question off-[topic](https://crypto.stackexchange.com/help/on-topic), and likely a violation of local academic code of conduct. I suggest the OP at least appends to the Q the state of their current research, and points what's blocking their progress.
RobbB avatar
vn flag
I am not a student in this area, I am a software engineer in training and this is purely a question of personal interest :) I do not have the scientific or mathematical capacity to answer this so I'm appealing to the community of geniuses ;)
kodlu avatar
sa flag
so what do you understand to be the difference between computing time and actual production time? what on earth are units of production time? what does it even mean?
RobbB avatar
vn flag
Sorry, that was tough to explain. What I mean is that, since multi-core processing can simultaneously compute concurrently, I would like to have a contiguous time measurement of cpu time. I guess this also translates directly into total time, for actual production time I mean not including the stated concurrency of cpu time. i.e. actual time. I will update and try to explain better.
RobbB avatar
vn flag
If you can advise me on better terminology I would appreciate it.
Maarten Bodewes avatar
in flag
I think it is an interesting question, but since a lot of computation is not even visible to everybody, I wonder how the heck we're supposed to answer it. It might be that mining is such a huge energy hog that it dwarves the rest, but since almost every data stream is encrypted, I kinda doubt that. I mean, I had to look up the non-encrypted mp3 stream for my fave radio station because even those are now HTTPS, and video streams are generally encrypted as well.
us flag
@MaartenBodewes, honestly, I think there are some possible and interesting answers that do some back-of-the-envelope maths. For instance, you could assume that all HTTP traffic is TLS-1.3-encrypted, and then compute that for a modern-ish server CPU. That alone sounds like an interesting answer to me. However, there's the "due to need for cryptography" in the question, which makes it a bit subjective, so maybe the question is not strictly answerable.
Maarten Bodewes avatar
in flag
As it stands you are asking four very broad questions; if this is just about approx. energy consumption for 2021 then the question may have a better chance of surviving. Questions that require active research are generally not considered on topic.
RobbB avatar
vn flag
What research would need to be done? Maybe a little for baselines of data in transit averages and approximate data in storage worldwide. Calculate computing times and go from there, I think I will make a go at this. As an example also, see how some geniuses have calculated energy and time for cryptographic key brute force attacks of current computing power worldwide. There is math that’s over my head there and why I’m not the right person for the job. I will still make a go at it
b degnan avatar
ca flag
@RobbB It generally costs me 50x power overhead at a minimum (in circuits) when compared to a non-cryptography implementation. AES256 is about 50x the power just on register costs. The power fluctuates with keys and data. If you want to dual-rail encode it to protect it from attacks, it's 100x. You can make a pretty good argument in power constrained computing to encrypt nothing.
Score:0
vn flag

Taking a .50 BMG shot at this one with my eyes half open. My math is ballpark at best, and highly elementary. I'm not a mathematician in the slightest. I would still like to see genius level answers here. But to show that its relatively simple I'm shooting for this, prove me wrong ;)

Reference: percentage of web traffic that's encrypted as of 2020 - 80-90%

Reference: global IP data per month as of end 2021 - 278,108 petabytes

Reference: time to encrypt 1GB with AES 128 - 0.634 seconds

Parameters I'm using:

Web traffic calculations only, on a per month basis

85% of all web traffic, which is 236391.8 petabytes

0.634 seconds to encrypt 1GB

Assuming a baseline of AES 128

2.2 Ghz CPU @ 37 watts

I am not including any data at rest/client side cryptographic actions

Calculations:

41631222.55 hours to encrypt one months worth of traffic only once.

1.54 Gigawatt hours of energy

$184,800 of power cost

Multiplied twice since the traffic must be encrypted and decrypted:

83262445.1 hours to encrypt two-way

3.8 Gigawatt hours

$369,600 of power cost

Per Year globally:

999149341.2 hours to encrypt two-way

45.6 Gigawatt hours

$4,435,200 of power cost

What it translates to per person:

.19 hours of encryption/person per year

8.82 watts/person per year

$0.00085 per person/year

Conclusion

Not bad at all, according to my very very ballpark parameters and math. Compared to bitcoin/month at 7583.33 gigawatts its like a needle in a haystack...

mangohost

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.