* Posts by Christian Berger

4850 publicly visible posts • joined 9 Mar 2007

Don't panic, but Linux's Systemd can be pwned via an evil DNS query

Christian Berger

Re: what's fascinating is how the SystemD fanboys react

Well there's a deeper problem with SystemD in the eyes of the people who don't like it. It basically boils down to values like simplicity, and being able to choose what software they run on. Yes, previous init systems were not very simple either, but at least that was based on shell scripts. If people had learned about actual SysV init, they'd probably want that.

The Pro-SystemD arguments always seem to turn around "being modern", even though most of their ideas have already been implemented in Windows NT.

It seems like an example of that phase most learning programmers go through wanting to re-invent the world in a more complex way. More experienced programmers have learned to live with that feeling.

Christian Berger

what's fascinating is how the SystemD fanboys react

I mean apparently there's a common missconception that that resolver runs as PID 1. I haven't checked that, but let's see how the fanboys react to this:

As far as I've read the discussion only one pointed out that this is an incorrect assumption.

All the rest just blamed people to not understand systemd, and therefore be idiots without giving any indication at what assumption was wrong.

Please, if you want symphaty for your cause (not an easy task) please at least try to be on a level moral playing field with those who criticise you. Don't call them idiots, respond to their critisism in a logical way. Otherwise you'll end up looking like idiots. If you have arguments that are not easily argued away, use them. If you don't have them, maybe you should reconsider your opinion.

Christian Berger

Better than SysV init? How?

"P.S. systemd is the greatest achievement of humankind. Well, it works. Usually. Better than SysV init anyway."

SysV init is based on a single text file with service definition. It supports automatic restarts and runlevels, so it's trivial, for example, to change over from the main HTTP server to one showing a "maintainance page". Just have 2 runlevels, one with the main server one with the other one. (or 2 different configuration files) Then switch the runlevels, and daemons will automatically be started and stopped.

It's simple and elegant and gets the job done.

Rackspace goes TITSUP in global outage outrage

Christian Berger

"What gets me - why the hell is the ticket system behind the same thing, making it inaccessible if it screws up?"

Why wouldn't you? That way you won't distracted by people whining that nothing works.

The bloke behind Star Fox is building a blockchain based casino. No, really

Christian Berger

Cool, a new market for miners

Since miners essentially control what happens with the block chain, and mining is controlled by a few miners (the main failure of Bitcoin), they could easily offer "cheating as a service". Just make them sign a different reality.

NHS WannaCrypt postmortem: Outbreak blamed on lack of accountability

Christian Berger

But they had Sophos

And Sophos advertised with "The NHS is totally protected with Sophos".

I mean they claimed to be able to do something against malware, yet they failed badly. They funneled away money that could have been used for actual security. Shouldn't they get, at least, part of the blame?

And no, other companies in that field aren't any better. Calling their products snake oil would give snake oil a bad name.

Robots will enable a sustainable grey economy

Christian Berger

How can a car-based future be long term sustainable?

I mean seriously, you either run out of crude oil or lithium, and before that global warming will have run off.

Cars are good for certain situations, but shipping around a 100kg person in a 2000kg car every day just to go to work or to buy stuff is idiotic. That's a problem decent public transport can solve cheaper and better.

Of course the car industry has a big problem with this, and since its very powerful (at least in Germany) no measures to reduce car ownership will be taken.

Tavis Ormandy to Microsoft: Have another Windows Defender vuln

Christian Berger

Stop trying to disprove Turing

There is no way you can find out if code is bad or good. Even if your sandbox is perfectly well written, malware will be able to detect it and behave. Neither does static analysis work as directly proven by Turing. Signature based virus scanners are something security students learn how to circumvent in their early semesters.

So please don't design your system in ways that assume you can somehow detect malware by running it through some software. You simply can't.

So uninstall Acrobat Reader and Office as well as Flash if you haven't done so already.

AES-256 keys sniffed in seconds using €200 of kit a few inches away

Christian Berger

Now lets put that in perspective...

FoxIT is one of those "security" companies working for the Dutch agencies.

This works via the magnetic field near to the device, so it's very limited in reach. It's hard to shield as you need ferromagnetic shielding for this, but it also won't reach very far, anyhow. So in any case, you need that device under your control. You can do loads of stuff in that scenario.

So, if you combine that, the obvious use for this is the following:

You have an "encrypted" mobile phone of your "suspect". Instead of having to ask them for the PIN, you can now simply sniff the key... and you don't even need to disassemble the device. All you need to do is apply a coil to a model specific position at the phone, then wait a minute in which you can also get the IMSI via an IMSI catcher. All of that works quickly enough to get the encryption keys during a normal "random search".

Heaps of Windows 10 internal builds, private source code leak online

Christian Berger

It'll probably be rather irrelevant...

Remember that time when the Windows 2000 source code leaked? Appart from some jokes about how much profanity it contained and through which lengths Microsoft was going to ensure compatibility with broken software, nothing actually happened.

IoT coverage for 95% of UK by 2019? We can't even do 4G, Sigfox

Christian Berger

Re: Mixing IoT with cellular networks

You'll notice that you're not the target market for this technology. It's aimed at industrial operators. Security cameras would be impossible to run over those networks as you don't have the bandwidth to send a picture per day.

Christian Berger

Mixing IoT with cellular networks

IoT has nothing to do with cellular networks. It's just a buzzword.

However the hype has attracted companies to build specialized "IoT" networks which can be rolled out much more easily as they have receive only base stations. Essentially you need no license for that, so most of the effort is actually coding some bizarre cloud platform which does get the data from the base stations and provide them to the users... the same thing could be done via E-Mail.

Those dedicated networks typically work one way, so the device sends data to a central receiver. Typically they can send a few dozends bits. A typical protocol is LoRa explained here:

https://media.ccc.de/v/33c3-7945-decoding_the_lora_phy

(or in that PoC||GTFO article)

BTW it being a send-only protocol solves lots of security problems. Privacy usually isn't an issue with such applications as they usually monitor things like pipe pressures or weather clock are still running.

FCC: LEO ISPs A-OK

Christian Berger

My guess is that they'll follow the Iridium model...

... create a company which sends up the satellites before it gets bankrupt... then you create a new company which busy up the assets from the first one.

Otherwise it's probably impossible to finance this, particularly since such a company is bound to make losses for the first decade or longer.

Cisco's 'encrypted traffic fingerprinting' turned into a product

Christian Berger

Re: It's an old idea

"Although I've got to say, this is an argument against pretty much any malware detection technology. If the author has tested their malware against your defensive product, they will surely have been able to find a way to circumvent it."

Yes, and this is why security experts (outside of malware detection firms) call such products snake oil. Now add to this, that those programs typically are rather complex, run with high privileges and try to unpack every obscure format you may not even have software to use them otherwise. Right now, for example, it's likely that you can take over computers with that RAR decoder bug that's been found recently... even if that computer doesn't have an unpacker that supports RAR.

Christian Berger

Re: It's an old idea

Yes, but since this is a packaged product, you can test it in your laboratory for as long as you want to. I'd guess that there is virtually no contrast between "normal" data and "malevolent" data, so those systems will probably spit out far to many false positives to be of any use.

Christian Berger

It's an old idea

There were already papers on finding out what place one looks on google maps, based on the size of the tiles (which were received via https).

It's not really suitable for finding malware, as it's trivial for malware to simply randomize its traffic. So instead of transmitting a file all at once, you send it in chunks of varying length, or mimic normal browser behaviour.

Researcher calls the fuzz on OpenVPN, uncovers crashy vulns

Christian Berger

Re: It shows that there is one feature missing

Yeah, but then I'd have used far more of the code, and those bugs would have been relevant.

TLS simply isn't a very good protocol security wise as it's to complex to be implemented sufficiently error free, and it has the outdated security model of CAs.

Christian Berger

It shows that there is one feature missing

There's a comparatively high number of OpenVPN installations that are just 1:1 connecting 2 computers to each other. Currently those typically use a shared key which has the advantage of shutting out most of the bugs mentioned here, but has the disadvantage of being able to decrypt all the data, once the key has leaked.

If there was a feature that would get around that, by using the shared key only for authentication, but using forward secrecy to negotiate its own key, we'd gain a lot more security for little more code.

In the week Uber blew up, Netflix restates 'No brilliant jerks' policy

Christian Berger

Well they already cut themselves off from a big portion of the "market"

For many developers, particularly in the higher class, Netflixs insistence on DRM is a no-go. It's just no a very ethical company.

They could at least drop the DRM for their own productions.

Debian 9 feels like home with security upgrades and a flaming vulpine warming your toes

Christian Berger

Re: " then other people criticise it, and instead of replying to the critique in a sensible way"

a) We live in a vendor dominiated economy. If no vendor offers open systems, you cannot buy them. Most high profile ARM devices are already locked down, and Microsoft has already made some steps in that direction, shutting down (still rather obscure) features like connected standby. Windows 8 on ARM even forced secure boot. https://www.heise.de/newsticker/meldung/Microsoft-erzwingt-auf-Windows-8-ARM-Geraeten-UEFI-Secure-Boot-1413109.html

Being able to turn off Secure Boot is now an optional feature:

https://www.heise.de/newsticker/meldung/Windows-10-Neue-Geraete-nur-mit-UEFI-Secure-Boot-und-TPM-2582371.html

b) I need to have the secret keys on the same system, since in the full concept of "Trusted Computing" no code should ever run that is not signed. Therefore every little 3 line program I write as a bash script would have to be signed. Since the whole hypothetical advantage of "Trusted Computing" relies on there being a complete chain of trust to be unbroken, I have to sign every command. Otherwise the whole tower of babel breaks down.

c) Yes, I know what malware can do at the boot level, but that's largely irrelevant as you can reach the same relevant goals even in unprivileged userspace. It doesn't matter if you have control over the kernel without the user knowing, as you can access all user files via its user privileges. It doesn't matter where you enter the system, once you are inside, you have typically reached your goal.

d) BTW nobody talks about "running pirated copies" of whatever. I don't see why you bring up that topic.

However, at least you had some arguments, and I applaud you for that.

Christian Berger

Re: Stop claiming that secure boot is evil just because you can't undestand it.

"Most issues around secure boot are ideological, not technical, and only because Microsoft is involved."

And that's virtually the _only_ reply you get when you actually make sensible points against Secure Boot and "Trusted Computing".

This is unfortunately a common problem right now. People invent something, then other people criticise it, and instead of replying to the critique in a sensible way providing arguments, they claim that the critique is just ideological and therefore bad. That's the typical strategy one applies when they are out of arguments. At this point the other side is usually giving up, which sounds like winning to the first side.

And yes, in a way it is an ideological argument. I want to have control over the computer I bought. I paid for it, so I make the rules on my computer. You may disagree, but I see this as the base to being able to trust my computer, which I see as an extension to my brain. (just like pen and paper when your're solving equations) In this ideology Secure Boot does not bring any advantage. Even if I can add my own keys to the system, I'll have to have the private keys on that system so I can sign the code I want to execute.

Christian Berger

Re: Stop claiming that secure boot is a security advantage...

"It is about having an uninterrupted chain of verified components starting at the BIOS/Firmware/Bootloader and all the way to the executable proven by cryptography."

Yes, but what does it prove? It only proves that there is a signature which was signed by a CA. It doesn't prove that it's in any way secure, it doesn't prove that the CA private key was not stolen or the CA was forced/tricked into signing malware. Signed malware is not just a theoretical problem, there has been some in the wild.

"Secure Boot" tries to outsource security to a 3rd party which somehow is secure. It gives you a false sense of security without providing any actual security.

Christian Berger

Stop claiming that secure boot is a security advantage...

...it's not, it's only a feature to protect business models. Nobody attacks systems via the boot process any more. (yes there were bootsector viri way back in MS-DOS)

If you have access to the physical device there are so many things you can do so much simpler, with the simplest being a modified laptop of the same model which you swap with the original one. It only needs to show a password promt, transmit that password to your computer and display "incorrect password." Then you can use that password to unlock the actual computer, get all the files, and bring it back, claiming that you mixed them up on mistake. After all companies typically have standardized on laptop models and they all look alike. For most larger companies that should work, as few employees would report that. After all it would mean admitting that they made an error.

Google, Mozilla both say they sped up the web today. One by blocking ads. One with ads

Christian Berger

Re: Why don't we ditch HTML+CSS+JS and just deliver websites as PNGs/JPEGs?

Well you could still delive the bare text as an addition and it still wouldn't be a large as our current mess. And obviously you'd transmit the window size and preference for font size to the server.

BTW most of the problems you mention already happen because of bad web design.

Christian Berger

Re: Why don't we ditch HTML+CSS+JS and just deliver websites as PNGs/JPEGs?

Well there is a crucial difference to Opera Mini. For it you have a centralized service. Here normal websites would do the job for you.

Christian Berger

Re: Paradox. Everyone hates ads. Everyone wants stuff for "free".

Ads have gone from simple "animated GIFs" to von Neuman complete bundles of Flash and/or Javascript. Before you had to display ads, which was rather harmless, now you need to execute them, with all the security implications that brings.

It's now a serious security problem not to have an ad blocker.

Christian Berger

Why don't we ditch HTML+CSS+JS and just deliver websites as PNGs/JPEGs?

I mean for many websites that would mean _much_ smaller pages and ads that are impossible to block.

For the user it would mean much smaller pages, and that ads no longer run Javascript on your computer.´

You could ad "tags" to your png file for tags and perhaps even video and audio stored in separate "HTML" objects. You could even ad "links" to images bordering your current image so you can scroll.

Those websites would be rendered on the server for your display resolution. Webdevelopers would get pixel precise control over how it would look like.

And on top of it all, it would greatly reduce the complexity of the web.

Samsung's 'Magician' for SSDs can let crims run evil code

Christian Berger

Re: I wonder how security would be...

Well, Samsung has 2 advantages with this:

a) They develop their own chips, so they have a head start when it comes to working with them. They can already write and test the firmware for early prototypes or even for simulations of those chips.

b) They market themselves as a premium company, so they would even have a bit more time for such things than the competition. Well tested Firmware would be a big advantage justifying the price premium.

Christian Berger

I wonder how security would be...

... if we'd all just avoid the obvious problems. I mean updating firmware shouldn't be so common you need a GUI application for that. Then if you download something of the internet, you should at least use TLS certificate pinning, or sign the firmware itself. (however do not have mandatory firmware signing for local updates, as that would prevent people from patching the firmware themselves)

'OK, everyone. Stop typing, this software is DONE,' said no one ever

Christian Berger

It's caused by multiple points

a) We now have lots of beginners writing code for GNU/Linux who would otherwise have written shareware software for Windows which nobody would ever have wanted. The userspace is virtually overrun by those people. See systemd, much of FreeDesktop, and PulseAudio. There's even a term for it, Lennartware.

b) Some software projects have grown in size so much that they need companies to support them. One way to get an income in software is to sell maintainance contracts. That gives you a strong motivation to build your software in a way so you need to contract them. One example for that is Asterisk. You can run it without a contract, but the documentation is less than what you're used to. Most of it is some independent Wiki.

Report estimates cost of disruption to GPS in UK would be £1bn per day

Christian Berger

It depends

"An "atomic" clock is maybe $2,000. I've not checked lately. Could be less. "

It depends on what you want. A rubidium clock is comparatively cheap, but they aren't quite as precise as caesium clocks. The later still costs 50k new and regularly requires changing the tube inside of it. That's to expensive to put in every cell-tower, but not a problem for a TV station which already pays 100k for a camera and another 100k for the lens.

"One GHz terrestrial system I know has an oven based "clock reference" only in the main mast and then the clients lock to a pilot carrier for stability, to save £100 per unit."

That's also standard for mobile phones, otherwise you couldn't get the carrier precise enough.

Frying all the satellites is unlikely, BTW, as half of them are on the "night" side of earth. The failures I've found so far were only temporary.

Christian Berger

That's why the EU has started Galileo...

...which prompted the Russians and the Chineese to do the same, so now we're going to have 3 politically independent systems. (There's an agreement for the US to be able to shut down Galileo if they wish to, and currently it doesn't seem like it's ever going to work, so I'm not counting it)

Building combined receivers is simple and that's why many phones have them. However some critical applications (timing mostly) may only have single standard receivers. Those would need to be upgraded with an additional multi standard timing receiver. For applications like DVB-T or DAB SFN synchronisation it would still take a day or more for the oscillators to drift enough in holdover mode for it to become a problem.

Perhaps it's time for someone to build an NTP-disciplined oscillator. :) (NTP has lots of short time yitter, but obviously no long term drift, so it might work)

That's random: OpenBSD adds more kernel security

Christian Berger

Well, you should always read documentation first

The documentation seems to be rather decent, and you should read it first, before asking your questions.

The reason is the same as with tier 1 call centres. Most people in IT have no f*cking clue what they are doing. They could get some clue by reading the documentation, and apparently they put lots of effort into the documentation.

So they provide a way to get your question answered, but you choose the most annoying way to do it they provide for only serious questions not answered by the documentation.

Christian Berger

Re: A bad copy of Windows

"It could be worse. Think how awful it would be if GNU/Linux devs turned Linux into a good copy of Windows."

Well Microsoft has a head start on it, as they are working hard on abandoning many of their problems. Virtually no modern software uses OLE. VBA, once an essential feature, is now seen as an evil. Software which needs logs typically writes them by itself, ignoring the Windows logging system. Essentially many Windows developers are now old enough to understand why the nifty features they have heared of in the 1990s and 2000s are utter shit.

Christian Berger

It's amazing where you can go when you have only completent people

I mean OpenBSD just ads sensible security feature after sensible security feature, while the GNU/Linux community is overrun by the Freedesktop/Pulseaudio/SystemD people which try to turn the Linux userspace into a bad copy of Windows.

Amazon.com just became a 90,000-seat Azure case study

Christian Berger

Back in the olden days...

Windows 2000 included a RDP server, so you could just get a small cluster of Windows servers and have your users log into them. Even a modest server could easily support 30 users working at the same time... so I don't quite see why one would go to a cloud service for this.

FOIA documents show the Kafkaesque state of US mass surveillance

Christian Berger

Of course there's a comparatively easy fix

Truly Free Software, which does not rely on a company or organisation to work on it. You cannot hand a court order to a loosely connected bunch of software developers, or in fact to a product that's long been finished.

We could start by taking GPG and simplifying it to a point where security critical bugs are improbable, then we'd have some unchanged piece of software which couldn't be backdoored.

Worried about election hacking? There's a technology fix – Helios

Christian Berger

Re: "Because you can"

"Whine as you like about paper ballots being slow to count "

Actually in Germany we have paper ballots, they are counted by volunteers. Polling stations close at 18:00, and most polling places finish counting at about 18:30-19:00. In time for the 20:00 news there's already a "preliminary official end result".

Christian Berger

No, elections don't work this way

The big problem is that only very few people will understand that protocol. An average person will not be able to check an election.

It doesn't matter how easy it is to fake an election, as long as people, without any kind of special training or abilities, are able to understand that system enough to check for all possible kinds of fraud. With a pen and paper system that is easy. At the start of the election you look into the ballot box to check that it's empty. Then you check that everyone entering the polling station will be crossed out from the list, and only throw one ballot into the box. Then at the end, you count all the ballots and check if they are the same as crossed out names on the list. If everything matches up, you can be sure nobody did ballot stuffing. Then you manually count them (or watch the people doing it) and make sure those numbers match up with the others.

Then you compare the numbers to the ones published on the election website or in the newspaper.

Seriously all of those steps can be done by someone who doesn't have any kind of special knowledge. Any anybody can come up with those points when they only think hard enough.

You can only have trust in a system you can understand, and elections are all about trust.

EPYC leak! No, it's better than celeb noodz: AMD's forthcoming server CPU

Christian Berger

Well ARM would need a single hardware platform

Essentially an "ARM-PC" where you can boot any install media without having to specifically port your operating system to your SoC.

Essentially they'd need to have some basic peripherals (like serial ports or SATA/SAS) interfaces defined and always on the same ports. Then they'd need to ditch UEFI for something much simpler (i.e. Open Firmware), to satisfy the demand for boot systems that are sufficiently bug free to not cause major problems.

The niche for ARM would be cheap "single purpose" servers. Kinda like the blade systems of the past, but _much_ less expensive. Perhaps systems that load an OS image into RAM and then run a single service on it. ARM is good at scaling down the power of a system that just sits around idly.

Samsung releases 49-inch desktop monitor with 32:9 aspect ratio

Christian Berger

Casual TV and movie watching

If you want to watch something, while you are doing something else, it's great to have a screen with many more pixels than the video has.

Christian Berger

Re: weighs 45 pounds

" as it was so lopsided with all the weight in the screen."

What's why you always carry CRT monitors with the screen towards you, much easier that way. There is no significant danger as the front of CRTs is the most sturdy piece about them.

Christian Berger

It's fascinating to see 3 kinds of commenters

One, seemingly working in full-screen, who can barely understand why one would want such a big monitor.

The other one, probably working with a windowing system, who want to get as many windows as possible onto their screen.

There's another group of people who use vector fonts, which complain about that display only having 80 dpi, instead of the 200-300 dpi they want to have.

Christian Berger

Re: Only 1080 high?

I've tried that, it works rather meh, as the screens get to narrow. Also bad web designers tend to make websites which fill the screen with junk, so you cannot make a narrow window without horizontal scrolling.

Christian Berger

It's a typical marketing decision

Companies see that 2 "Full-HD" monitors next to each other are popular, therefore they bring out one that combines the same number of pixels.

Instead they should look at the underlying problem, and realize that people would also like to have more pixels in height, because they can already get more pixels in width by placing 2 monitors next to each other.

So the smart thing would be to bring out a 40 inch or larger "4k" monitor.

Brit hacker admits he siphoned info from US military satellite network

Christian Berger

The details would be interresting

Was this an insecure website? Were the classes of bugs already widely known? If so, why didn't they hire people who know what they were doing when handling sensitive data.

Small carriers aren't showing up to IPv6 standards chats, consultant warns

Christian Berger

Re: There are fundamental technical reasons for it

Yeah, so you're saying that I should use G-Mail instead of my own mailserver?

And seriously, with UPnP all those crappy devices you don't want to have on the net are _already_ on the net, despite of NAT, they simply open port forwarding.

We have lots of problems already which stem from NAT and which aren't addressed by advances in IPv4. Look at mobile devices. People cannot even transfer a file from computer A to computer B because they are likely in 2 different NATs. That's why we even have data-slurping companies like Instagram or WhatsApp. Nobody would even think about using that if there was E2E communications, you'd just use the network like you use the telephone network.

"Smart" devices often have to go through a central computer so you can talk to them through the net. If that vendor decides to shut down that service, you can no longer use that feature.

IPv4 simply separates the net into 2 classes of users. Those with public IP addresses, and those without. Increasingly people, particularly in poorer countries, end up behind multiple layers of NAT.

It's 2017 and someone's probably still using WINS naming. If so, stop

Christian Berger

Re: "the Internet would be using WINS naming exclusively, and that DNS would be totally gone"

"MS love to create "standards." If enough of them take off they continue to force people to stay on MS platforms. "Everyone will be using this in X years. You'd be a fool not to start coding for it" blah blah."

Well unfortunately that's something virtually all inexperienced programmers do. That's why we have FreeDesktop, PulseAudio or SystemD, or if you think it further, all those new half-baked features coming to new poppular language versions.

Programmers need to learn to restrict themselves to solve problems as simple as possible. Software architecture isn't about using every language feature or every new fancy methodology, it's about choosing the right tools for the job, to make the problem as simple as possible so it can be solved as simple and accurately as possible.

Buggy devices and lazy operators make VoLTE a security nightmare

Christian Berger

Nothing new

Mobile telephone number portability in Germany is still exchanged via X.something mail, not even standard Internet E-Mail.

And actually a mailing list is much saner than the fixed line number portability scheme in Germany which goes through intermediaries which charge lots of money for handling a comparatively small databases.

Christian Berger

It's not CCC 32...

... it's 32C3, and here's the URL of the talk:

https://media.ccc.de/v/32c3-7502-dissecting_volte