I think you highlight the exact issue rather well... fixed-point DSP instructions are awkward to use. The FPU and the double-precision hardware with its baked operations work with IEEE floats out of the box, so the programmer can be "lazy." A thousand new programmers can write something like `* 0.7` in C++ inside a tight loop without it stealing 200 instructions from the timing of the rest of the program.
Downside is it occupies a CS on the QSPI controller, presumably bonding to the same pads as the QSPI pins on the package, so now you only have one external memory IC. It's a very small tradeoff all things considered, but is still technically a tiny disadvantage over highly integrated MCUs.
A potential alternative would have been a directly memory-mapped NOR flash die, but that would have required more bond wires, more dedicated pads on the die, a port on the bus, and on top of that the memory die would have been more expensive too.
An older (and often impractical) alternative is to use a single die with both flash and SoC on, in the same process. This usually forces a larger-than-desired process node to match the flash technology, making the SoC take up more space. The result requires no extra bond wires or pads, but now you're really manufacturing a flash chip with an MCU attached.
A significant part of the problem is that a majority of people are unaware of just how simple what they consider "intelligence" really is. You don't need actual intelligence to replace the public-facing social role of a politician, or a talking head, or a reactive-only middle manager. You just need words strung together that fit a problem.
The disappearance of this blog post makes me think there's been a realization of some sort, perhaps that the post closely resembled accounts of "gang stalking", posted by individuals suffering from paranoid schizophrenia? I wonder if you realize in turn, how close you are to organized stalking yourself.
@kuschku is the developer of Quasseldroid, the pictured Android IRC client. Quasseldroid displays the screenshotted warning when joining channels on Leenode.
Considering john the ripper has a plugin that can churn out 50k c/s on a gpg key with a mid-tier GPU without specific optimizations, I'd guess a dedicated team of NSA researchers could get the cost for off-the-shelf hardware down to 5000 c/s/$ (based on a $100 GPU running 50k c/s + 10x speedup from engineering effort and specific optimizations), which makes the cost of the raw GPU hardware for a 1 trillion passphrase GPU cluster a smooth $200 million for a civilian assembling in his basement.
Wanna bet the NSA gets volume discounts from nVidia/AMD?
When I saw 1 trillion guesses per second I immediately wondered what algorithm was being referenced. My single GTX 780 hash performance varies wildly by algorithm. A few numbers:
1 trillion hashes/sec on a key stretching algorithm like bcrypt would be pretty horrific and might require quantum computing, while the same performance on MD5 might be achieved with <50k in hardware (very rough estimates).
I've heard rumors of storage technology that can store thousands of petabytes in a home appliance form factor. With that can kind of storage it would make sense to just start making salted rainbow tables. Even without fabled hardware, the Bluffdale NSA facility might have the capacity for it. I haven't even done napkin-based calculations yet to see if this is possible, so if anyone has some idea please speak up :-)
There are several symmetric encryption algorithms to choose from, the default is CAST5 (according to this[1] random mail post). This would only be used to encrypt the private key on disk.
Now I'm curious of the methods of decrypting data in transit. Does the NSA have the tools to break PKI based encryption at 1 trillion guesses/sec? I have some wild guesses, but if anyone knows I'd love to hear it.
>Wanna bet the NSA gets volume discounts from nVidia/AMD?
So you think they're actually using off-the-shelf GPUs for their password breaking? I would assume any operation with a budget like theirs would create their own ASIC chips specifically targeting the algorithms they need to run. We've seen this happen for Bitcoin hashing, so I'm sure the NSA is way ahead of them.
* Sun Microsystems put special instruction into their CPUs to aid in faster decryption efforts.
If I was designing such a system I'd stick a bunch of GOST, AES, SHA256, Blowfish ... brute forcing cores embedded in a small reconfigurable mesh. It would make a very effective multipurpose brute forcing device.
Antminer S2-b4 can do 2 trillion hashes a second and costs 1200 USD[0]. Imagine you are the NSA with tens of billions of dollars to spend on rigs, access to major fabs and you've been attacking crypto for the last 60 years.
Why FPGAs? I though they were much slower, and if you're building out a berjillion nodes you probably don't require the reconfigurability because the problem is known.
NSA has no budget limits with taxpayer dollars - they have a secret budget so the public can never review - what congressional secret hearing is going to turn down a request for more money by them?
1 trillion was TWO years ago. Assume they doubled that by now.
This is incorrect. The entirety of the NSA's budget has to fit "within" the existing government budget. And other departments do use money. So it is not unlimited. Defense and International security assistance is $643B this year [1] not a trillion dollars, and certainly not two trillion.
> If you think you can pinpoint secret project costs in budgets, go find the the X-37B expenditures as an example.
It is in the Air Force budget. It may or may not be a line item. If it is, it uses an unclassified code name. The revenue for the contractors works the same way; the exact revenues for the program are hidden in totals.
With the information they have, what is to stop them from "taxing" the global economy indirectly via the capital markets.
Seems like they could easily set up investment companies anywhere in the world, trading on all sorts of information and funnel those proceeds back to fund more NSA activities.
There's the classic movie plot of the person who devices a virus to skim fractions of a penny off many transactions, which in aggregate is significant. Now take an organization, whose mandate is to collect as much information as they can, has operated for 60 years, and has had the benefit of being attached to the country with the greatest global and political reach ever seen. The possibility to embed themselves in the economic fabric of every country and market in the world is well in the realm of possible.
It's not like funding humint and sigint activities from capitalistic activities is without precedent. Air America was one example [0]. That one mechanism only blew up and became know because it was beyond egregious. Too many average people had knowledge to keep such activities a secret.
Now imagine you have the largest surveillance apparatus in the World and it's the late 1980s. Do you not think that the NSA and CIA took some notice of the incredible amount of money that could be made on money in those days. Pretty much all the activities we have learned the NSA and CIA have been involved with are harder to justify (surveillance, torture, propping up bad governments, etc.) from the perspective of American ideology than making money from capitalistic activities based on information asymmetries.
These are the types of activities that we will only see a whistleblower for if they are finance focused (accountant/auditor/investor) and have broad access to information. Unfortunately, people with intimate awareness of the financial operations of the NSA are far more rare than a security and surveillance analyst like Edward Snowden (since that is the core business of the NSA)
At Enron, a massive multinational organization, the number of people who were truly aware of the shenanigans going only were a tiny fraction of a fraction of the employees. It's safe to assume that when ever any organization gets large enough they invest in financial professionals of great capacity and task them with being creative. Why would would we expect any different from the NSA or CIA?
It would be especially interesting if Laura Poitras and Glenn Greenwald were to connect with and involve freedom-loving financial forensic professionals who have the skills to comb through their trove and uncover questionable financial operations practices (assuming the archives that Snowden had access to were broad enough to include financial details, especially those that concern revenue and not just expenditures.
Even if cost ever became a problem, at some point the value of the information they're able to unveil would pay for itself. Think of the raw value of having advanced knowledge of confidential business and political dealings. What could that be worth?
Makes you wonder how many black ops they could fund through insider trading. Even without specific encryption breaking hardware the scope of NSA's programs would serve equally well to front run essentially any major market move.
How easy do you think it would be to scrape a little cream off the top of the HFT latte when/if you could see everyones source code and/or tap the ingress points on the exchanges?
Even precluding outright "cheats" like above, I can't imagine it'd be hard to beat wall street at it's own games. Prop firms like to buy satellite images of Walmart parking lots and count cars to extrapolate earnings, sounds like a technique the NSA would be in a position to improve upon.
