SoylentNews is people
SoylentNews
Four years ago, I did a back of the envelope calculation on how much it would cost to store audio recordings of all the phone calls in Norway, and came up with NOK 2.1 million / EUR 250 000 for the year 2013. It is time to repeat the calculation using updated numbers. The calculation is based on how much data storage is needed for each minute of audio, how many minutes all the calls in Norway sums up to, multiplied by the cost of data storage.
[...] Both the cost of storage and the number of phone call minutes have dropped since the last time, bringing the cost down to a level where I guess even small organizations can afford to store the audio recording from every phone call taken in a year in Norway. Of course, this is just the cost of buying the storage equipment. Maintenance, need to be included as well, but the volume of a single year is about a single rack of hard drives, so it is not much more than I could fit in my own home.
(Score: 4, Interesting) by JoeMerchant on Tuesday November 26 2019, @10:02PM (6 children)
I assume Norwegians are still communicating, just using other methods.
Also, food for thought, the US population is only ~62x that of Norway, anyone care to bet on whether or not the FBI is willing to spring $3M per year to store all domestic and international US phone calls?
Next question: what's the cost to run voice recognition on 445TB worth of voice calls and store that in a cross-indexable database? Also seems like something that would be more than "cost effective" in a post-911 security community operation.
🌻🌻 [google.com]
(Score: 1, Informative) by Anonymous Coward on Wednesday November 27 2019, @12:44AM (1 child)
The FBI don't need to. The NSA's PRISM facility is already recording all calls made in most of the western world. It isn't just phone companies in the US forwarding everything, but internationally.
(Score: 5, Interesting) by edIII on Wednesday November 27 2019, @06:44AM
Nobody needs to do so. The FBI with CALEA is already able to intercept all calls in real-time. All the high level carriers for the PSTN/Cellular networks are running mediation switches that allow such intercepts. The FBI's DCSNet [wired.com] manages the whole thing.
There's already systems in place to funnel the data. Speech to text has advanced very significantly, especially when you can afford the best AI. The data storage requirements for the metadata of processed audio are a heck of a lot less. Anything interesting that causes the conversation to be flagged can also have the raw data preserved. A couple hundred well placed servers could probably process and compress all the traffic for the entire United States. Especially since these would heavily leverage TPU's, which might even bring real time transcription to the table. If it doesn't already exist as a feature.
If you're speaking into any system that is connected to the PSTN, you can absolutely assume that you were monitored. The primary operational paradigm of these systems seems to be that your rights aren't being violated as a human being hasn't seen the data yet. Until then it's innocent moving, processing, and copying of the data to temporary locations.
They're very good at it, and when done at the Layer 1 level there's practically zero way for anybody to tell what is going on with their equipment. It's so easy actually, that the US government isn't the only one monitoring :) Why would the US spend any money when it can just nudge another member of the Five Eyes to do it and share the data?
If there's anything they will have a challenge with it's bringing that same level of surveillance to Internet communications. Thankfully, it's a problem several orders of magnitude bigger.
Technically, lunchtime is at any moment. It's just a wave function.
(Score: 0) by Anonymous Coward on Wednesday November 27 2019, @05:53AM
Check out the Utah Data Center [wikipedia.org]. Its storage capacity is estimated to be at least on the scale of exabytes. Probably the most relevant quote there is:
4 years later PRISM was leaked.
(Score: 0) by Anonymous Coward on Wednesday November 27 2019, @11:50AM (2 children)
In his book "Culture and Empire", Pieter Hintjens makes the point that within decades it will cost governments less than $100 per person to store all audio and video surveillance of each person within their borders for an entire year.
(Score: 0) by Anonymous Coward on Wednesday November 27 2019, @01:05PM
Several of Hintjens book are available for download directly from his old web site, which is still up. "Culture and Empire" is also available for download:
https://legacy.gitbook.com/download/pdf/book/hintjens/culture-empire [gitbook.com]
It's is quite a good read, as are his other books.
(Score: 2) by takyon on Wednesday November 27 2019, @01:43PM
He wrote the book in 2013. $100 can buy about 6 TB of disk storage today. That's enough to hold about 2 months of 10 Mbps bitrate video. So if $16.70/TB declines to $2.78/TB, i.e. 36 TB for $100, you're there. Seagate is planning 50 TB by 2026, so the prediction could become reality by 2030.
You can play with those numbers by lowering the bitrate, using bulk prices for the millions of drives needed, assuming NAND will become cheaper than disk, using magnetic tape, adding in other program overhead, etc.
A $100/person surveillance program would cost over $35 billion per year for the U.S. Feasible but big and costly, with a physically large footprint if billions of drives are needed. However, the next 20 years is plenty of time for a storage breakthrough that crushes HDDs, NAND, and tape. $10 per person-year would make the surveillance a lot more attractive. That will happen "within decades" and there will be nice microdrones to capture all the footage.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]