5,737 karma · joined June 19, 2013
https://www.newsweek.com/russia-highest-monthly-losses-ukrai...
They also loose equipment at a staggering rate. From the start of the invasion 2022 until end of July, they have lost more than 570 000 soldiers, 8000 tanks and 16000 artillery systems.
Oh and also 38 war ships and one submarine - against a country that didn't have a navy at the start of the invasion. Drones is ammunition and is used in air, land and sea.
We meat sacks don't stand a chance. We need to evolve.
But what the article describes, and what my experience as a VLSI designer is that as feature sizes decreases, being able to just be able to guarantee a specified lifetime given environmental limits (ambient and die temperature, voltage and current supply including stability etc) is a very hard problem. And is getting harder. Just the statistical variance in the wafer production process steps (amount of molecules deposited) will make migration harder to estimate and handle.
A few SiO2 molecules more or less in thickness of an oxide layer that is 10000 molecules thick is not very important. But when the layer is supposed to be 10 molecules thick, 12 or 8 molecules corresponds to a huge difference in leakage.
So stating a minimum lifetime is what you do. A lifetime where at least 99% (or so) of the device still works reliably. Not to increase sales, but to avoid getting class action suits, having to pay out too much in refund, replace devices etc.
You don't want to step on the toes of a zombie do you? (It's probably gross.)
If it is a MCU based on a simple ARM Cortex M0, M0+, M3 or RISC-V RV3I, then you could use an iCE40 or similar FPGA to provide a big acceleration by just using the DSPs and the big SPRAM.
Basically add the custom compute operations and space that doesn't exist in the MCU, operations that would take several, many instructions to do in SW. Also, just by offloading to the FPGA AI 'co-processor' frees up the MCU to do other things.
The kernel operations in the Tiny GPU project is actually really good examples of things you could efficiently implement in an iCE40UP FPGA device, resulting in substantial acceleration. And using EBRs (block RAM) and/or the SPRAM for block queues would make a nice interface to the MCU.
One could also implement a RISC-V core in the FPGA, thus having a single chip with a low latency interface to the AI accelerator. You could even implement the AI acceleator as a set of custom instructions. There are so many possible solutions!
An ice40UP-5K FPGA will set you back 9 EUR in single quantity.
This concept of course scales up to performance and cost levels you talk about. With many possible steps in between.
"Note: Update on April 18: Step 9 of the algorithm contains a bug, which I don’t know how to fix. See Section 3.5.9 (Page 37) for details. I sincerely thank Hongxun Wu and (independently) Thomas Vidick for finding the bug today. Now the claim of showing a polynomial time quantum algorithm for solving LWE with polynomial modulus-noise ratios does not hold. I leave the rest of the paper as it is (added a clarification of an operation in Step 8) as a hope that ideas like Complex Gaussian and windowed QFT may find other applications in quantum computation, or tackle LWE in other ways."
Companies (as in its employees esp managers, processes, culture) do things that are not at all efficient, correct, in the best interest of the company all the time. Not to break out into another discussion - but the resent push to force employees back to the office could be seen as one such example.
But it is also a source of entropy that is hard to manipulate (if we ignore the cameras, image generation etc). And it is a good source, initiator for discussions about entropy, random numbers and why we need them to secure things. This thread on HN shows that it is so. And for 'normal' people probably even more.
I'm sure they have a lot of non-geeky visitors that look at the installation, read the sign and start asking questions. And if it was me, starting to jump in front of the installation, waving my arms - to provide my bit of entropy to the pool.
Another benefit of a CSPRNG is vastly higher performance than most TRNGs can achieve. A TRNG often provide kbps birate. A CSPRNG can easily deliver Many MBps, even GBps.
Yes it is much more cool to look at than a zener diode, but it is still a source of physical entropy. And one could argue that a whole bunch of chaotic pendulums are more resilient to failure than a single diode.
Some would claim that CMVP, FIPS 140-3, Common Criteria stickers on a black box (HSM) is security theater.
Interesting that they go through all that work to protect against modern deobfuscation attacks and reversing - and then use RC4 and MD5. They may be ok in this context, but the choice is sort of odd.
It states: "I’ve plowed through the first wave of AVP (Apple Vision Pro) reviews, and it seems pretty obvious that, at the current price and form factor, it’s not gonna be a best-seller."
I'm not sure what constitutes a best-seller. But according to several news stories (which may all be based on the same anonymous rumor), Apple has sold 200.000 Apple Vision Pro headsets: https://www.macrumors.com/2024/01/29/apple-vision-pro-headse...
Does anybody know how many headsets the other VR vendors have sold?
``` Only the latest major release of macOS (currently macOS Sonoma) receives patches for all known security vulnerabilities.
The previous two releases receive some security updates, but not for all vulnerabilities known to Apple.
In 2021, Apple fixed a critical privilege escalation vulnerability in macOS Big Sur, but a fix remained unavailable for the previous release, macOS Catalina, for 234 days, until Apple was informed that the vulnerability was being used to infect the computers of people who visited Hong Kong pro-democracy websites. ```
The only other compiler for 6502 I've ever used (and had to modify) is the old cc65. The original author stop developing it (10+ years ago): https://cc65.github.io/
But there is some development ongoing: https://cc65.github.io/
Would be interesting to see comparison with the new one.
But customers are still counting every um2 chip area, gates/chip and ROM, code mem in bytes in 2024 to minimize per-device implementation cost. Automotive, a lot of consumer products and large scale SCADA systems are example of systems where these types of questions are at least evaluated during requirement development and system exploration, design. And yes, shaving of a few cents due to being able to use a MCU with 8 kByte RAM instead of 16 kByte can save a lot of money. And it is being done.
When I look and crypto algorithms (in a wide sense) I see more utility in having a primitive that provides AEAD, or can be reused n different way to do secure hashing (for integrity, kdf, random generation for example) as well as AEAD, than having lightweight block ciphers. Most lightweight block ciphers have small block sizes, short keys and a lot of rounds. Lyckily there are some nice algorithms that we can trust. And AES is very scalable.