HNHacker News
TopNewBestAskShowJobs

KirillPanov

817 karma · joined December 30, 2020

Quantum Cyber 5G Blockchain AI of Things
submissionscomments
KirillPanov··on DMCA Excemptions to "Copyright Protection Systems" Being Reconsidered
and excretions.
KirillPanov··on Encrypted traffic interception on Hetzner and Linode targeting Jabber service
I feel like packet size was and continues to be a major obstacle for DNSSEC. Do you know why the DNSSEC/DANE world hasn't simply acknowledged this and switched to requiring ECC?

It is trivial to fit several compressed curve points (i.e. signatures) in a single packet, whereas you can't even fit two RSA signatures in a minimum-safe-to-assume DNS UDP reply packet after accounting for padding and ASN.1 overhead.

I get the feeling that there is some faction that really hates UDP and they sort of hijacked the DNSSEC situation to use as a lever to force people to allow DNS-over-TCP.

That seems to be backfiring, however, and DNSSEC has wound up taking a bullet for the UDP-haters.

Many very-large networks simply can't afford for their DNS traffic to be exposed to TCP's intrusion-detection malleability and slowloris (resource exhaustion) attacks. These networks appear to be simply ignoring the "thou must TCP thine DNS" edict. DNSSEC is not a good enough carrot for them. I think ditching RSA would have been a more pragmatic choice than ditching UDP or skipping DNSSEC.

KirillPanov··on Mitigating the Hetzner/Linode XMPP.ru MitM interception incident
Read it again:

> they would have to do their own client-side TLS handshake

By intercepting the STARTTLS port the attacker can merely decrypt -- rather than, as you wrote, decrypting and re-encrypting.

KirillPanov··on Mitigating the Hetzner/Linode XMPP.ru MitM interception incident
More likely the carrier upstream of them.
KirillPanov··on Mitigating the Hetzner/Linode XMPP.ru MitM interception incident
TLS client certificates.

Use them.

Stop using domains. Stop.

KirillPanov··on Mitigating the Hetzner/Linode XMPP.ru MitM interception incident
Easier to conceal the attack.

The MiTM attacker can pass through a command stream without STARTTLS. If they intercepted 5223 they would have to do their own client-side TLS handshake with the attacked server, which would look really obvious to anybody doing TLS fingerprinting on the server: all of a sudden, 100% of their clients have the exact same TLS fingerprint.

Stop outsourcing your PKI to ICANN, folks. Domains are not public keys.

KirillPanov··on Accounting prof testifies about FTX's misuse of customers' money
It wasn't a withdrawable loan; the attacker used pumped-up MobileCoin as collateral to go leveraged-long on some other shitcoin in a wash trade against another FTX account which he (presumably) also controlled. The attacker withdrew the profits from the second account; when the first account went into liquidation MobileCoin tanked and the collateral was worthless.

At least that's what Michael Lewis' book Going Infinite says happened.

FTX's mistake was that their liquidation engine valued collateral using something that could be manipulated. It's actually a very hard problem for an illiquid asset -- the correct valuation is the order book depth during the margin account's liquidation -- but that's always in the future so you can't know it exactly. And an attacker with two accounts can manipulate most of the things you might want to use as an approximation (like the last trade price or the order book depth right now).

The book claims that the bankrupcy trustee found the exploiter and the trustee CEO claimed they "had video of him entering and leaving his house" in (IIRC) Mauritius. However since the funds haven't been recovered the bankruptcy trustee could be bluffing or simply wrong about this. Or they can't extradite the guy. So not quite total silence, but also definitely not resolved.

BTW the book is really good, although it feels a bit rushed (obviously Lewis had a hard deadline for optimal publication date so this is totally forgiveable). There is a crapton of juicy stuff in there, like "Operation Warm Blanket". The number of powerful people who have egg on their face because of that book is absolutely unbelievable. No redactions. So many of them would have been able to quietly sneak out the back door with no public record if not for Lewis. Oh yeah and the mother of all easter eggs on the inside of the dust jacket.

KirillPanov··on Why the Culture Wins: An Appreciation of Iain M. Banks (2017)
... and yet my cerebrum still has not attacked my brainstem.

You're applying evolutionary reasoning to an algorithm (gradient descent) which does not involve fitness selection in any way, shape, or form. This is bogus.

KirillPanov··on Why the Culture Wins: An Appreciation of Iain M. Banks (2017)
No, I think it's a terrible metaphor.

The Minds have no reason to treat humans well. Humans have a powerful evolutionary motivation to treat their kin well.

KirillPanov··on Why the Culture Wins: An Appreciation of Iain M. Banks (2017)
> Why does this idea bother humans us so much?

I dunno; it doesn't bother me at all.

What bothers me is pretending that the Culture is humanity's. Or that humans are in any way relevant to the whole plotline. It's like if somebody rewrote all the great works of literature from the perspectives of the protagonists' dogs, and then threw away the originals, and pretended that this was somehow a useful perspective. That would bother me.

KirillPanov··on Why the Culture Wins: An Appreciation of Iain M. Banks (2017)
> Finally, I think Minds need humans... all the minds that had no humans with them (that I remember, at least) where the "unstable", eccentric or downright crazy.

Well okay; the humans aren't pets, they're emotional support animals. Not a huge improvement though.

KirillPanov··on Jobs working with the Rust programming language. No crypto
... but Anduril sure does.
KirillPanov··on Why the Culture Wins: An Appreciation of Iain M. Banks (2017)
What makes you think that the advent of AI will be any different from the advent of cerebrums?

When higher vertebrates developed cerebrums, those didn't replace the evoluntionarily-older cerebellum or midbrain. We humans still have those. From the perspective of a lizard, a human looks like a lizard with an awesome AI-like cerebrum strapped to it.

Importantly, the cerebrum does not have an independent self-preservation drive.

Frankly, I think "AIs with their own self-preservation drive and plans for the future" is implausible, although it is likely to be put forward as a deceptive explanation of what's going on by the first few humans who figure out how to strap on another layer of cerebrum. It's the easiest way to keep the rest of us from becoming resentful.

Come home, all you commodores, come home...

KirillPanov··on Why the Culture Wins: An Appreciation of Iain M. Banks (2017)
> and we just let the machines take over

... but then it's really the machines' Culture, isn't it? And we're just their pets.

I'm going to read Consider Phlebas based on the essay linked above, but I think it's a bit dishonest to describe this Culture as belonging to any of its inhabitants other than The Minds. It's The Minds' Culture, and everybody else are just their housecats. Kind of a disappointing future if there's any other option besides self-annihilation. Cozy, but disappointing.

KirillPanov··on Canon Nanoimprint lithography semiconductor manufacturing system (5nm)
> In addition, isn't the cost of a mask in it's design, not it's production?

They tend to be in the same ballpark.

There's a feedback effect: if the maskset gets cheaper you can afford to do shorter product cycles (think "iphone 15.5"), now you've got two masksets each with half the engineer-years devoted to it. This lets you be a bit bolder with the design since the cost of screwing up is lower. If masks get super-expensive then you only do one maskset and you throw more design-hours into making damn sure it's correct, and spend more time adding in "chicken bits" -- ways to turn off features that you screwed up, without ruining the whole chip -- and fallback options or insurance-policy type stuff. So the design cycle will lengthen or shorten if the engineering cost is several times larger or several times smaller than the mask cost. They sort of balance out.

Also the design cost is often impossible to define exactly. Blocks from previous designs get reused, so you can't really "charge" the effort it took to design them to just one chip. All the major design houses write a lot their own CAD tools too, and these tools have decade-or-longer lifetimes.

KirillPanov··on Mastercard Should Stop Selling Our Data
Exactly this.

The issuer can also sell that debt (daily rental up to car value) to a debt collector.

KirillPanov··on Why the Culture Wins: An Appreciation of Iain M. Banks (2017)
Christopher Rowe's The Voluntary State

https://nebulas.sfwa.org/nominated-work/the-voluntary-state/

https://www.lexal.net/scifi/scifiction/originals/originals_a...

Five years ago I would not have called it "realistic", but a lot has changed.

Come home, all you commodores, come home...

KirillPanov··on Canon Nanoimprint lithography semiconductor manufacturing system (5nm)
Image comparing the processes: https://global.canon/en/news/img/2017/p20170720b-02.jpg

Deposition is usually the same -- CVD.

Note that the stamp is applied to the resist:

> the new product does it by pressing a mask imprinted with the circuit pattern on the resist on the wafer like a stamp

KirillPanov··on Canon Nanoimprint lithography semiconductor manufacturing system (5nm)
Mask lifetime.

Last time I heard about this tech (~8 years ago, when it was nowhere near commercial availability) that was the obstacle. Since the mask comes into physical contact with each wafer there is unavoidable wear on the mask with each wafer (actually with each chip). So the mask life is a certain number of chips; then you have to make a new one.

KirillPanov··on Canon Nanoimprint lithography semiconductor manufacturing system (5nm)
How many impressions can you get out of each mask before it has to be discarded? Last time I heard about this tech, that was the limiting factor.

This uses e-beams to make the mask, right? Edit: here's their research paper (7 years old but note the identical model name FPA-1200NZ2C):

http://cnt.canon.com/wp-content/uploads/2016/08/Nanoimprint-...

Turns out they e-beam a "master mask" and then use that as an imprint to create "replica masks". Then the "replica masks" imprint the wafers. Sadly you still have to start with a crazy-expensive e-beam mask.

KirillPanov··on Mastercard Should Stop Selling Our Data
I've always wondered about this. You drive off with a EUR30k car. What do they do if you don't come back? Drain your bank account of 500 Euros? What then?

In the US, the rental agreements let them keep charging you the daily rental fee until the amount collected exceeds the value of the car.

KirillPanov··on American Big Tech Has Enslaved Us [video]
> massive money printing by central banks after the 2008 crisis provided the funding for this rapid expansion of cloud capital

This phenomenon is so massively underappreciated.

Now that interest rates are back to normal I am extremely interested in seeing how much of this phenomenon is permanent (i.e. path-dependent). I just hope that Big Tech hasn't become so important that it is bailout-eligible.

"It isn't until the tide goes out that you see who's been swimming naked" -- various successful investors.

KirillPanov··on Finding a Tech Job Is Still a Nightmare
Quality: the need for more of it than Wikipedia has.

Coherence: Wikipedia pages are small and loosely-coupled.

Stability: Wikipedia has no external API. Any page can be renamed and restructured at any time.

KirillPanov··on Amazon sellers say they made a good living – until Amazon figured it out
Because https://en.wikipedia.org/wiki/Sherman_Act
KirillPanov··on Amazon sellers say they made a good living – until Amazon figured it out
There is such an easy solution here: Amazon divests their logistics operation -- the warehouses, airplanes, and delivery trucks. They become a common carrier with a published price list; amazon.com and all the sellers on Amazon become customers of this logistics company. Probably a lot of ebay sellers too.
KirillPanov··on San Francisco is becoming a tech hub again, Y Combinator CEO says
I need to unload my real estate bags, Y Combinator CEO says
KirillPanov··on How LSP could have been better
"How" did the word "how" end up in the HN title for this?

Spin?

KirillPanov··on Curl/libcurl HIGH CVE-2023-38545 leaked early?
Wat.

So you can only be attacked if you're using a socks5 proxy, and even then you can only be attacked by your own proxy? Which rules out things like torsocks where you're running the proxy too.

Does this really merit all of last week's antics?

KirillPanov··on Extreme parkour with legged robots
> Ideally, the panacea to such nightmares is democracy.

Come on man, this is delusional. I want to believe it too but be realistic.

KirillPanov··on Stop EU Chat Control
Yes, quite a few in fact. That's why you'll see a lot of intentional mis-spellings like "virtual signaling".
← PreviousPage 2 of 27Next →