@FutureSecurity Some Schneier blog comments actually come from some very knowledgeable people, others are just crazy conspiracy theorists. You have to pay attention to the names, in which case a few valuable gems will pop out occasionally.

@FutureSecurity That's a real thing but I have no idea if it applies to GSM devices. I would tend to think not, although individual baseband processors might have their own backdoors^Wdebugging features.

@EllaRose No they do not.

@forest What am I looking at? (besides the most monstrous function header I have ever seen)

@EllaRose 62 arguments to a single function, an 80-something character function name, and indentation that would make even a GNU C style aficionado cry.

ah, so it was just horrible code

I thought it would be security bugs or something like that

Yes, horrible code that was accepted into Linux because people don't check commits.

Well I'm sure there are security bugs in there too. It's AMD code lol

that is pretty horrible. But, if it was the code for /dev/random, surely someone would at least glance over it?

Yeah. The random driver is actually pretty well written.

someone at AMD was paid to write that code?

Yep.

I should re-evaluate my doubts regarding my ability to be employed as a programmer....

Not great though. There's one proc interface it exposes that you can set but which is actually a no-op now. There's some inconsistent programming in regards to how entropy sources are collected, and there's at least one function that does nothing.

@forest They do apparently work with gsm and the command set has been expanded. Plus the usb hack article says you can even emulate touch screen input. That website lists keyboard entry too.

E.g. compare how bytes are mixed in here vs here

unsigned long time = random_get_entropy() ^ jiffies; actually kind of hurts.

@FutureSecurity Oh interesting.

GSM/UTMS/LTE/etc isn't something I'm super familiar with.

@FutureSecurity I'd be a bit surprised if they could be triggered over GSM though.

At least considering all law enforcement I've known has, at most, used GSM to triangulate people and, in a pinch, exploit the baseband. If it was as easy as sending a command to force the device to listen, then I'd think it'd be used more often.

atcommands.org/sec18-tian.pdf might be interesting.

Ooh. "Comprehensive". That's good. I don't know anything more about gsm than the average person.

Yeah it looks like AT commands can't be issued remotely through GSM.

Judging by the way this paper is worded (still reading though).

But I don't know jack about GSM either.

Wonderful... What are all these commands for? Some of them don't look like they'd be the job of the baseband.

I would guess it's something the OS sends to the baseband.

Not something sent by a cell tower to the baseband over GSM.

Which makes sense given that individual executables running on the device send AT commands to the baseband to control it, and that it can be done over USB.

That's what I'm thinking. But contacts list management is part of the command set. Let the "dumb" part of the phone do radio and the "smart" part do things like handle GUI.

That's what the original theory was.

Though it was probably more that the "dumb" part can implement proprietary GSM commands without needing the vendor to know more than they should.

Thankfully the baseband is pretty heavily isolated nowadays, though there are obviously still bugs (a recent one was with seL4 running on the baseband, which was only exploited because the hypervisor it ran on had a nasty bug, and seL4 is only proven correct if the underlying ISA is perfectly to spec).

@forest en.wikipedia.org/wiki/ECall#Privacy_concerns

> eCall was made mandatory in all new cars sold within the EU from April 2018.[1]

wow

e911 is similar, but at least it requires explicit activation.

I thought maybe the thing I barely remembered reading about might have something to do with European standards, emergency laws, or police backdoor. I'm sure wikipedians would say that section could be improved. It's not really informative or specific and the citations don't look good.

Apparently OnStar tracks your car whether you're subscribed to the service or not...

I hate "Well if you expect free services then you have to be okay with privacy violations." First, what people normally think of when they think free service is that it means it's ad supported. Ads do not require mass surveillance despite what Google says.

People are willing to let companies walk all over them.

At the expense of everyone else.

Second, companies that charge for their services collect just as much data. They have even more because they have your billing info to cross reference.

mhm

did you ever notice, if you copy a piece of data and share it with someone else, it's theft. If companies sell your information to each other, it's business!

^^^

The only kind of theft where the "victim" is not deprived of the original item.

It's like someone steals your care, but in the morning it's still there!

You know what I hear people say too much? "They have a right to make a profit" Americans will say that rather quickly but any other rights they'll be extremely hesitant about if not hostile to.

They believe people have the right to make a profit at another's expense.

We're so paranoid that working class people are out to get us. They will take from you. Their getting more means you get less. They are only deserving if they might my criteria.

It's repulsive, really, the way Americans twist the truth into making their victims look like the bad guys. Reprehensible, even.

I think in the post McCarthyism era we're having our humanity taken away from us with propaganda and taboo and misinformation...

Ever notice the rogue AI/robot trope parallels fear of slave revolts or worker rebellion?

Heh, I haven't noticed that, but it makes perfect sense.

A little ironic because "robot" comes from Slavic (Czech) "robota", meaning slave.

wow

I have a lot of difficulty thinking of popular media that portrays technological dystopia accurately. The Doctor Who reboot is the only example I know of.

I agree completely.

In addition to Doctor Who, the only good technodystopia portrayals are those that show machines doing only what they were designed to do, with the designers having bad (or at least malevolent) intentions, e.g. war machines/"kill bots".

The best example I think is the Japanese visual novel Planetarian.

Neither is American.

Oh, and of course the morphine-wielding robot dog in Fahrenheit 451.

And even Bradbury's works are distinctly opposing to American ideals.

At least modern ideals, where book burning is not looked upon as badly as it was.

Hell, I've seen "progressive American feminists" push for book burning!

As well as "conservative American Christians", who do the exact same.

It seems only libraries and information freedom advocates are against things like that today, and even libraries are starting to die now that platforms which give a company more control over how you use the things you buy (e.g. ebooks) are rising.

Don't believe that about feminists. YouTube is really messing up our culture. The low hanging fruit and incredibly attention grabbing stories are always the ones amplified, assuming the story is true. Some of the things people are most reactionary about come from far right people getting outraged over far right parody accounts, not realizing it wasn't meant to be taken seriously. It's a feedback loop that feeds on memes, click baitiness, confirmation bias, fame, and revenue.

Oh there are some feminists who are like that. I don't mean to accuse them all.

I was thinking of some 3rd wave feminist movement pushing to ban certain books from schools for their portrayal of women, and I am very much against banning books.

It's a very fractured movement, with people who have all sorts of beliefs.

Some of which are pro-censorship and anti-sexuality, others are the opposite.

I think third wave has become more of a dog whistle than accurate descriptor. But it is true that there are people very authoritarian, very gender essentialists, usually some type of self described radical feminist.

I think the third wave just doesn't have a defined, unified goal.

I'm more critical of that wave than the previous ones.

I'm an egalitarian feminist, not a gender feminist. I think that sums my views up.

Eh. It's a word about as useful as "millenial"

I.e. I see meritocracy as a positive thing that promotes equality.

Whereas many gender feminists think meritocracy is bad (which horrifies me).

Merit is something most often that manifests retroactively. It just so happens that the people who are in power and the people you like have merit and people you dislike have found their own way of cheating.

Merit is not something that someone in power judges.

In software development for example, merit is part of your compiler.

The compiler doesn't care about gender, ethnicity, religion.

It's a common misconception that meritocracy relies on merit from judgement.

It's specifically about intrinsic (objective) merit to avoid that issue.

Is it? Are the best employees the ones managers assert are best or the ones that have to clean up for other people?

Managers in modern corporations do not operate on only meritocracy.

Which is one reason why there's so much corruption in business.

If corruption tries to hide behind meritocracy, it's not meritocracy.

Any more than a Nazi can hide behind the theory of evolution for eugenics.

Same for feminists and Scottsmen.

Sure. My point is only that attacking the idea of meritocracy repulses me.

That's where I see a divide in feminism. Egalitarian feminists are pro-meritocracy, whereas gender feminists are anti-meritocracy. As an egalitarian feminist, I support it.

It's one reason I strongly dislike so many of the modern code of conducts, since they almost all stem from one specific one that is a self-described and vehemently anti-meritocratic document that asserts that "cultural contributes are more important than technical merit", rather than supporting the kind of meritocracy that ensures that someone will be hired because they are good for the job, regardless of gender.

The thing about "quota" filling. You're hiring just a few people. You automatically feel that someone from that group is less qualified. You only hire the best applicant.

In that way, I see embracing meritocracy as the most important step to knocking down barriers for minorities or at-risk groups. It asserts that the only thing that is important is your abilities, and nothing else matters. And, unlike what many detractors try to claim, it also focuses on the fact that people who are disadvantaged and do not have the necessary skills can be taught. After all, merit is also about potential.

@FutureSecurity And that's a good thing, assuming it's done in such a way that subconcious biases do not affect the outcome (e.g. blind reviews).

And naturally having programs to train minorities or other at-risk groups is important, since it allows them to gain the skills that can make them the best hire.

But then people are upset that scholarships and funding isn't meritocratic.

It depends on whether or not the scholarships/funding is evidence-based.

You can easily have an outreach program without a scholarship.

Meritocracy is about giving everyone equal opportunities and then selecting the best. It's vitally important to make sure that you aren't being preferential to any one group, or the other groups will fall behind, and the system breaks down.

E.g. make education in the subject equally available to men and women, whites, blacks, and Asians, the wealthy and the poor (yes, that's a classism issue), and ensure that they all have the same quality of education.

Unfortunately, we're not doing that right now when so many people refrain from supporting giving money to underprivileged communities (e.g. "ghettos").

Because of the poor education in many poorer areas, these people are unable to acquire the knowledge they need, which is a perfect example of inequality.

If humans were fact machines, meritocracy would work smoothly. But I'm not a machine. Other people aren't machines. And even machines develop biases. We need to look at our own decisions and compensate for our own biases. Our biases are always greater than we believe them to be. "Merit" has baggage. There are a lot of feelings involved and no non-subjective metrics. Usually we gravitate towards deciding who deserves or doesn't deserve something based on past performance or aggregate metrics.

There are always ways to measure merit objectively.

It's done regularly in peer review in academia, for example.

Now it's true that meritocracy is very hard in certain fields, e.g. politics, but in others like software development, it works amazingly and promotes equality.

It is unfortunate that some subjects have an unclear definition of merit, and as such meritocracy does not work for them yet. Those are not the ones I'm talking about.

Meritocracy implemented in the real world is looking at the past. We want to fill positions with people that will (future tense) maximize our well being. (And sometimes teams of diverse people perform better than teams of elites. Especially sheltered and arrogant elites whose metrics may have been boosted by nepotism. )

But that's not the case. In academia, it works amazingly.

Same with software dev and hard sciences.

I mean the lack of meritocracy is, well, bias and discrimination.

Either you hire someone based on merit, or based on something else.

And if that something else is not relevant to the job description, it's discrimination.

> And sometimes teams of diverse people perform better than teams of elites.

If it's better, than diversity is merit!

Look at the Olympics. People look at US athletes, learn that one person is gay, and then assume that that person didn't get the position based on merits. And qualifying for the olympics is probably the closest thing we have to true meritocracy.

People's homophobic assumptions are irrelevant to actual merit.

One in ten people are gay. Of course there will be some gay athletes.

The most important thing about meritocracy is that merit is defined as whatever works. Sure, for an "old white guy contest", your merit comes from being old and white. For software development, merit comes from your skills with the programming language, your ability to get along with others (in collaborative work environments), your ability to work under stress, etc, and ethnicity or gender no longer matters.

I gotta go. Will be back in a bit. Fun convo. :p

There is no cosmic referee we can consult. We are limited knowledge, emotional, biased, irrational beings. We ought to look at things like one-in-ten frequencies, plug our numbers into a binomial distribution inverse CDF and see if we need to compensate for our own issues. ;) Merit defined as creating the best outcomes of course creates best outcomes. So of course I don't disagree with the principle.

I think I'm satisfied with one last comment: Meritocracy is something we should strive for. We haven't achieved it. Merit as we think of it is muddled with deservedness. Our metrics of merit are flawed. We should target that goal by finding ways to eliminate bias. (And training a neural net doesn't work.) We undervalue potential in our judgement of merit, and there isn't anything that people won't whine reverse discrimination about, even and especially true meritocracy.

Steve Gibson! And Donald.

@FutureSecurity Yeah, I agree. The whole time I only meant that we should strive for it, not that we should assume our measures of merit are flawless. It's the exact same with utilitarianism. We will always need to improve our understanding.

Also there's no such thing as "reverse" discrimination, just discrimination.

Also Steve Gibson is an insane snakeoil salesman. He lacks merit. :P

@forest don't you think your answer here should include some kind of disclaimer? It currently reads like you're recommending that they actually use two-level E0 for whatever their project is.

I saw a reference to Rivest's spritz algorithm. What a useless thing.

I got fairly angry the first time the paper because he notes that, although the algorithm is slow, he notes that he hand't optimized his implementation and someone else could do better.

When you have enough reputation...

I wonder what it was like between now and him first gaining reputation.

We're in an era in which we have good cipher algorithms. If you propose a new algorithm it better be more efficient in some way than existing algorithms. You're supposed to bake in efficient implementation during the algorithm design process. Not bumble around in hopes that you find an efficient implementation after the fact.

Did we have good implementations for RSA as soon as it is published?

I also lost respect for him because he didn't realize a software implementation cannot be optimized. The array swaps are the bottleneck in spritz and rc4, and the remaining operations are all single cycle. That's obvious and he had rc4 implementations to use for reference.

I don't know. Probably not, since asymmetric crypto is complicated.

I see now, you are talking about the obvious problem on the speed up.

Yes. He should have at least known that whatever performance problems that applied to RC4 would also apply to Spritz. I think RC4 takes upwards of 11 cycles per byte on PCs.

@EllaRose is there a paper of your trapdoor?

@kelalaka No, not really. I mean, there might be somewhere on the hard drive of my old computer, but it was not a "paper" in the sense of something you can find on eprint.

The writeup for the underhanded crypto contest is the most convenient copy I can think of

I can explain it here if you like

with a prelude stating that all of the variants that I tried (trust me, there were a lot) were broken

though it required LLL to do so (though different parameterizations might yield better results)

@kelalaka Shall I continue?

Not yet. Later. Let me watch the video fully so I can ask better question instead of stupid ones :)

well, the video about the UCC from DEFCON shows a variant of the design that was adapted to work as a backdoor for a (broken) key agreement scheme

but the core ideas are still in there

@kelalaka the writeup from UCC: github.com/erose1337/crypto/blob/master/underhandedcrypto2018/…

(I notice they still haven't updated their website)

@EllaRose thanks

a note about the security reduction in the writeup: It applies to RSA with similar parameters, which is not the same as RSA in general

@EllaRose Good point. I'll do that, thanks.

@FutureSecurity I wouldn't be surprised if Spritz was just a fun pet project he had that someone convinced him to take far too seriously.

@EllaRose Regarding that question, I think that's a severe case of an XY problem. OP is afraid that he might have non-repeating IVs due to using an embedded device and wants to protect from that, and also wants to avoid padding attacks and so does not want to use CBC. Apparently he also thinks that it's fine to hardcode the keys in the device... So his question is just misguided. I think he should just use GCM.

I agree completely

I was hoping an answer would talk about that

SIV mode might be relevant too

or SIV-GCM

Unfortunately my answer was after I found out it was an XY problem, and I don't think I would be able to answer his unspoken question directly.

Especially since it's a hardware deal now (how to get randomness on embedded).

I really don't get why people don't just use known-good implementations...

I mean even RC4-drop3072 is going to be better than what he could come up with.

thinking

OK I'm going to stop reading through his comments.