The Side Channel

7:55 AM

@Patriot I'm sorry, but that makes no sense. Even if you come up with a totally novel noise source, it will at most cause brief interest. It might get a research paper or two out if you really push for it. But a shitstorm? No, we already have TRNGs that can generate randomness which is impossible to predict due to quantum mechanics, and you can build one for $5.

And you can use such a TRNG to generate an OTP, although there's really no need to.

@Patriot Cryptography doesn't fail at that because it doesn't intend to explain it. All Joe Schmoe needs to know is that the little green lock icon means they are safe. That's a UI issue.

10:07 AM

@Patriot Regarding "cutting the data trail", I'm not sure if this is what you had in mind, but perhaps you should look into the concept of red/black separation for EMSEC.

@forest You are on... every time zone.

I have a polyphasic sleep schedule.

does that actually work? I have a friend who has 6, 28 hour days, and I still don't know how he does it.

well, if you can be productive, better off eh?

It works if you don't have a job that requires you get up and go to the office to work for hours. In my case, I just have severe insomnia and the solution I've found is lots of little naps.

Since I work online, I don't need to be up at the same time every day.

Make sense. insomnia is what drove me to research over teaching. I didn't need to show up, just get the work done.

Hilariously, once I got on the research track, the insomnia stopped. :P

It was probably just student anxiety.

10:15 AM

I'm a consultant (so a contractor), so I don't need to worry about showing up on time. I am considering a PhD but I don't have enough time, so I just keep doing what I'm doing now.

I can get up 10 minutes before I'm supposed to be online, or five hours. It's nice.

I was recording all of my interactions with students because you never know who would blame you for something, so I guess that was stressful. Ph.D.s are sort of worthless. I only got one because there was a problem that I was very interested in, and it was the only way to get access to tools for the problem (IC fab). It was a waste of time as my skill set didn't really expand, only my tools.

7 years of lost revenue.

The kind of research I'm interested in is booming right now, and that's computational neuroscience.

But it's so different from infosec that it's hard to dedicate time to both subjects.

I mean, with infosec I can test pretty much anything I want on a $150 laptop. I don't need a team of researchers and expensive laboratory equipment, just a VM and toolchain.

That's what I actually when to work on. I read Carver's Book: en.wikipedia.org/wiki/Carver_Mead And then I found Hasler.

That's where I started doing asynchronous circuits because I wanted human/computer interfaces; however, the asynchronous work ended up being more useful as processors because if you can make a circuit that works at 1pA, you are pretty good at analog design.

Digital is a special case of analog, so one thing led to another.

Computational neuroscience is?

PYNN simulator is a pretty good way to simulate processing systems with simulated neurons, you then can target them to neural ICs to see if the power savings works.

10:23 AM

I think that's a different field since computational neuroscience is about modeling biological systems mathematically (e.g. taking cortical minicolumns and modeling them using attractor dynamics) and less about designing artificial neural nets. I forget what that field is called.

Neural nets are one of the most wonderful biologically-inspired set of algorithms for sure.

Machine learning is another thing that I think would be fascinating, but it's way over my head.

well, neuromorphic engineering is using the models to make processing blocks.

C. Elegans is well known, but sadly, we couldn't get it to work.

The worm simulator?

ieeexplore.ieee.org/abstract/document/6218734

OpenWorm or whatever?

Hm, that page isn't opening for me.

So, I made this IC, but I'm last because I pissed off my advisor. One sec.

10:27 AM

418 error lol

capocaccia.ethz.ch/capo/raw-attachment/wiki/2013/neurofpaa13/…

I started a hearing aid company that used neural nets to detect speech in noise, and I didn't include my advisor and they freaked.

They are super smart but suck at business.

damn

Is it really necessary to use an NN for that? I thought speech enhancement by augmenting certain frequency ranges to simulate our inner ear's natural sensitivity curves is generally sufficient.

Anyway, computational neuroscience is hot; however, modern computers take too much power. analog networks act like nerves and take very little power. It's necessary because you need to decide "what is speech". That takes tons of power in digital, and much less in analog.

Human can do things like audio processing with the power of a sandwich.

I suppose that's where computational neuroscience and neuromorphic engineering go hand in hand. The former to see how our own auditory system does it, and the latter to emulate it artificially without using an obscene amount of processing power.

Might I ask where you are geographically located?

10:33 AM

I don't like providing my location.

I probably can tell you where/who to lurk if you want to see what the state of the art is

no problem

So, USA west: one sec

However I can travel around the world with few restrictions, and I'm comfortable doing research in virtually any country. I'm not bound to any one location for long periods of time.

web.stanford.edu/group/brainsinsilicon/boahen.html

he's open lectures and his lab

oh nice

USA eastern: engineering.dartmouth.edu/people/faculty/rahul-sarpeshkar

Sarpeshkar is the smartest person that I've ever met, and that's saying a lot

Tobi Delbruk has a cool silicon retina that can balance a pencil: sensors.ini.uzh.ch/tobi.html

Those three are the most friendly

10:37 AM

I'll bookmark them! Sounds very interesting.

Or at least, they are to me. Kwabena and I both have the same Ghanian name as we were both born on Tuesday. He thought that was hilarious.

Good luck with you academic endeavors. neuralensemble.org/docs/PyNN is a good way to start things with software. you can push it into hardware later

I need to debate living conditions with a woodchuck. :/

I wish it was easier to break into. I've been so used to the ease of doing things entirely on a computer that it would require a complete paradigm shift for me to truly embrace academia.

10:53 AM

@EllaRose It's not unique to cryptography. In business, this is probably the most common issue of all. People can be absolutely sure that they have what it takes to penetrate a highly competitive industry and they just know that they have a product that people will buy. Then they create a kickstarter for $100k and make a grand total of $0 over a period of 5 years.

12:24 PM

@forest Don't worry, the consequences are way over everyone else's heads too and that doesn't stop them from throwing whatever math sticks on a wall of the MNIST data set.

Q: Verify partial message given hash of full message

2:01 PM

0

I need to verify partial downloads (retrieved sequentially), given a (preferably short) hash of the complete download. Is there a construct that can do that, even a probabilistic one?

Can anyone else figure out what this person is asking for? Somehow we are supposed to download a part of a file, but the file must not be broken into parts; somehow there is supposed to be a hash function that works on parts of a file, but the hash function must not be computed in parts and must not involve trees; in one minute we assume the hash is known good, in another minute we don't.

2:16 PM

(I don't seem to have a ‘move this discussion to chat’ button on the answer's comments, curious.)

Ella Rose

2:27 PM

I am pretty sure that moving comments to chat is a mod power

So I am guessing that they have some kind of file transfer mechanism, and that they want to check every X bits of the downloaded content. If it doesn't validate, then cancel the file transfer.

I have no idea how accurate/helpful this is, this is just my interpretation of the preview above

Patriot

3:10 PM

@ Forest I appreciate your comments. So, I did not know that a low-cost TRNG already exists, one designed to be very simple to use, along with an interface that anyone could figure out in two minutes, and one extremely difficult to subvert, which is really a major issue. I like the TRNG that can only be subverted via interdiction or surreptitious entry, and even then it might not be easy to mess up without the end-user knowing.

Patriot

3:22 PM

@ Squeamish Regarding the question above: I took it that the writer wants to know whether a message that has been broken up and is being delivered in sequence--part A, part B, part C, part D-- can be verified with a hash of the complete message (A+B+C+D) if we only have A, or A +B, or A+B+C...

A: Verify partial message given hash of full message

4:57 PM

0

Suppose you found an error-detecting code $f(H(m), m')$ (forget random oracles, second-preimage resistance, etc.) with the following properties: $|H(m)| = O(1)$, i.e. the checksum overhead is independent of the message length. $f(H(m_0 \mathbin\| m_1), m_0) = f(H(m_0 \mathbin\| m_1), m_0 \mathb...

Maybe this one is more satisfactory.