How Not to Redact a Warhead | Hacker News Reader

How Not to Redact a Warhead

blog.nuclearsecrecy.com

112 pointsbenbreen32 comments

Someone I went to uni with got a job at an engineering firm. One of his projects was unmounting nuclear reactors for decomissioning, which were sufficiently secret that he was told to treat them as a black box. He was given a procedure, including specific drill and cut locations which would safely detach the reactor, but no detail about what was inside the shell.

One day he was on holiday in France and went to a museum that had displays about nuclear technology. One of the information boards was an exploded view of the same reactor technologies he was working with.

Who knows if there were subtle differences, but it's fascinating how different decisions about redaction are made given knowledge that the information is public.

Its possible it was just the fact that a slightly different design was rated at a different secrecy level because of specific improvements.

For example, rockets and turbofans all use the same basic mechanics to function, the but differentiators are still under ITAR

RSA is still under ITAR too which just seems incredibly dumb.

Are you sure? Between the 1996 Wassenaar Arrangement and the 1992 transfer of cryptographic regulations from Defence to Commerce, I don't believe this is the case any longer.

Sometimes the secret is the fact of using a specific technology, not the technology itself.

Gotchas, tools, jigs, and tricks of design, manufacturing, maintenance, and support, as well as the very rare materials, are the most crucial bits to guard.

IIRC, the manufacturing of a low-yield, simple fission device based on an old design isn't complicated; it's the fissile materials that are the show-stoppers, hence nonproliferation of centrifuges, dual-use components, and yellowcake/ore.

Reminds me of a factory tour I went on where a big fancy robot was assembling car parts. I was told not to take photos — not because the robot was secretive (and in fact the same one was used in several factories), but because the computer screen beside it showed details about the specific configuration of that robot. There were hundreds of different settings that could be tweaked and that was the competitive advantage.

Makes sense. I toured the Dell factory in Round Rock, and it was the same deal. I think all businesses attempt to limit information disclosure as a standard practice, even if there aren't obvious trade secrets, because there maybe unrecognized intelligence in them.

Sometimes the secret is just bullshit, though.

https://www.cfr.org/blog/why-i-have-nothing-say-about-nsa-le...

> Emails from reporters started coming in last night. Could I comment on the leaked National Security Agency (NSA) report on Russian interference in the election?

> The short answer was no. The reason was simple: I couldn’t read it.

> As one of the 5.5 million Americans who hold a security clearance, viewing that document would violate my obligation to protect classified information.

Sometimes what you don't say reveals as much as what you do. This is the origin of the Glomar Response. By never providing information it is harder for third parties to tell when something is really secret or just public knowledge. This keeps the actual secret things that much more obscure. If I ask you ten questions and you answer four of them I learned something about all ten topics. If you refuse to give me useful information on all of them I learn nothing.

I suspect this solution makes sense in the short run. In the long-run it makes it so any organizational incompetence can be covered up with "it's classified". Over 80+ years the organization starts to struggle with basic reality.

I wouldn't be surprised if the CIA is filled with Byzantine Bureaucracy, fiefdoms, and departments that don't even know what they are supposed to do. In a kafka-esque twist I'd bet there are individuals who aren't even allowed to know their own job description due to some papered over incompetence.

This is common. I've worked at places where interns started rambling about classified stuff because they found articles on wikipedia and unknowingly put two and two together. The classification system is like any other extremely bureaucratic system: fairly arbitrary and frequently divorced from reality. That's not to say it's terrible and doesn't work, just occasionally struggles when dealing with the problem of determining what needs to be secret out of the set of all facts about the observable universe.

For those not fammiliar with classified materials, this sounds like a case of classification by compilation, where multiple pieces of information are individually unclassified, but are classified when put together.

For instance, you might have both the time and location of a planned event be unclassified, but have the combination of time and location be classified.

This can get very annoying when you have people working entirely on unclassified documents (possibly without ever having read the classification guidelines, since they never need access to the classified stuff), and they end up "leaking" classified info by compilation.

It also gets extremely silly when you have to say something like "I'm sorry, but I cannot copy table 7 from document A into document B without making it classified as it would leak information X".

In the end, as the article says, I don't think this really matters. We rarely care about classifying specific pieces of information; and when we do, it is usually a relatively clear line. For the most part, for a leak to be useful, it needs to contain a lot of separate pieces of information that come together to make a cohesive whole.

The complete design, construction, and operations manual for a nuclear bomb will probably help an aspiring nuclear power build it. However, a single page of said documents, even if said power could choose which page, just isn't going to be that useful.

I used to follow a Freedom of Information blog (I think it was this one [0]) and I remember a case where after a long drawn out battle they received a heavily redacted document... that it turned out had been publicly available for years.

[0] http://foi-privacy.blogspot.com/ gah no https!

>And if that’s the case… what’s the point of all of this secrecy, then?

Because any one thing isn't enough to reconstruct the full picture, but many pieces combined can provide the bigger picture.

And the problem is way more complex than "let's use a series of blurry pictures to reverse engineer a nuke"--there are more goals than this, like for instance figuring out the U.S. nuke arsenal capabilities, seeing if an adversary has figured out a technology we possess, etc etc

Come on now, this is basic OPSEC.

edit: Like for instance, look at the Snowden leaks. There were some stuff in the Snowden leaks that had very little to do with US public interest, but were accidentally leaked by stupid journos who don't know how to use a computer. And this did lead to actionable intelligence from our adversaries.

In your edit, you probably mean Assange, not Snowden.

No, I meant snowden.

I thought the classic in poor redactions was adding black boxes to pdf but not deleting the information underneath the black box.

Placebo redactions. Win-Win

Also, using a reversible algorithm for blurring.

On the topic of nuclear secrecy breaches, there was this recent case. It would aid someone trying to steal a weapon, which would help a lot in creating one, I imagine, though hard to pull off and probably impossible to do so in secret: US Soldiers Expose Nuclear Weapons Secrets Via Flashcard Apps

https://www.bellingcat.com/news/2021/05/28/us-soldiers-expos...

HN discussion (307 points, 5 days ago): https://news.ycombinator.com/item?id=27317946

This is a wild take, that this military intelligence isn't useful to opposing militaries, and that hydrogen bombs aren't a pretty damn dangerous thing to risk leaking. I'd much rather people be overcautious with respect to nukes than undercautious. Hasn't 2020 taught us to respect tail risks a bit better than that?

lol, society extrapolating. Nope.

I had the same reaction. Of course society is not going to take the lessons "learned" from a global pandemic and apply them to other fields. As soon as things go back to "normal", people will forget and repeat the same processes that got them into the situations that the pandemic "revealed".

I had the exact same thought. And why make it easy for an adversary? The game is about staying ahead, and part of that is making sure they don’t know exactly what you’re working on.

An under-appreciated aspect of most traditional engineering disciplines is how many unknowns there are from materials science, physics, quality control, and manufacturing precision.

You could know the shape, mechanics, yield, make-up, and components of the Fat-Man device while still being thousands of experiments and a multi-billion dollar industrial infrastructure away from being able to build one.

We got very lucky that it so happens that the fissible isotope of uranium is the rare one. Although I forget if there's a relationship with half life, so it might be inevitable?

Natural Uranium is not very radioactive. For example, it would take Uranium 4.5 billion years to release the same amount of radiation that Francium does in 22 minutes.

> a multi-billion dollar industrial infrastructure away from being able to build one.

This is true! And the biggest reason why despite AQ Kahn selling a "make your own fat man" kit only a few countries were successful with it.

However, additive manufacturing poses a big nuclear proliferation risk. Being able to manufacture accurate parts for cheap, without advanced export controlled equipment, poses serious issues.

As the other commenter said, showing people how to get eggs and flour is not the same as baking them a cake. The vast majority of the information is a by-product of treating our nation like a business: we now have shareholders to account for, and their collective vested interests are preventing the rest of the world from using this information to update reactors in France or advance experimentation at CERN.

Reminds me of my favourite winner of the underhanded c contest, back in 2008 http://notanumber.net/archives/54/underhanded-c-the-leaky-re...

Effectively this draws 'black boxes', but in a way that allows the developer to recover much information from the document if desired.

The trick is in the ppm format, where a 0, 00, and 000 are all valid 'black' values. The redaction works by turning all 'digits' into zeros in the specified area ... but obviously you can recover much information by turning those zeros back into full values. In the case of black-on-white text, this effectively recovers the text completely. Brilliant!

Reply on news.ycombinator.com