Schneier on Security

Clive Robinson • November 17, 2019 12:02 AM

@ RJ Brown,

Bonus question: What does ubuntu mean?

Yes “officially” it is a word from the south of Africa from around Swaziland, and Zimbabwe, which comes from the Nguni Bantu language. Unfortunatly the word has many meanings that do not readily translate into English. Many asume the word is a philosophical doctrine or approach to life that emphasizes social unity and generosity of spirit, especially at the end of Apartheid.

Some also say it means “is correct behavior” with respect to your fellow humans. Which is ironic because of the telemetry that Mark shuttleworth has added to spy on the users, which many find to not be even remotely “correct behavior”…

Clive Robinson • November 18, 2019 7:38 AM

@ Bruce and the usuall susprcts,

https://www.reuters.com/article/us-interpol-encryption-exclusive/exclusive-interpol-plans-to-condemn-encryption-spread-citing-predators-sources-say-idUSKBN1XR0S7

It appears that the FBI has told Interpol to bring out a statement favourable to thr FBI view point, without it going through the normal expected voting process,

At the [Interpol] group’s conference in Lyon, France on Friday, an Interpol official said a version of the resolution introduced by the U.S. Federal Bureau of Investigation would be released without a formal vote by representatives of the roughly 60 countries in attendance, the sources said.

Unfortunately it’s the same old “think of the children” corner case designed like that of a “Health and safety” argument, to stop people raising objections to it as it would make them look sympethetic / evil etc.

the larger [Interpol] group will cite difficulties in catching child sexual predators as grounds for companies opening up user communications to authorities wielding court warrants.

Whilst it might be called “exceptional access” today, if history shows us anything, if it is given the access will be “mundane” tommorow and “required” the day after.

There are aproximately 5billion people who access the Internet quite a few via the likes of mobile phones or mobile computers, most of which are “walled gardens” where the user has no choice but to run applications that the likes of Google or the hardware manufacturers force on them via the faux “user safety” mantra. Which as we know has been often used for malware or worse.

It’s not just many individual users who will be put at risk by this, it actually has the potential to destroy many economic sectors, and as it’s the likes of the US and other Western First World Nations that are most dependent on strong security that will be hit the hardest by job losses etc in such sectors.

If I was designing a realy harmful “Cyber-weapon” to destroy countries economically, then it would be this sort of access I would be looking to exploit…

Clive Robinson • November 19, 2019 1:14 PM

@ Anders,

Did you spot the “night fighter” high VHF / low UHF radar that was in some of the photographs?

The displays are not just crypto kit, some are even more interesting such as radio navigation 😉

Clive Robinson • November 21, 2019 1:39 PM

@ Bruce and the usual suspects,

On a twitter post, it’s been suggested that the XKCD “horse,battery…” passphrase system at 44bits is in effect “toast”,

https://twitter.com/taoeffect/status/1197576478298263553

Clive Robinson • November 23, 2019 6:25 AM

@ SpaceLifeForm,

My thinking is eight words with at least a nine digit number.

In part it depends on your dictionary size you select yoir words from. But 2^10 or 1024 words would be not unusual. So assum Hex digits gives you 2^4 bits and with eight words you get,

2^84 = 2^16 x (2^10)^8

At 84 bit’s equivalent it sounds kind of just about acceptable.

But…,

Unfortunately, hard for the user to remember.

Yup, watch the entropy tumble when users get their “diceware word list” and rearange the word order to make a more memorable pass phrase. They might also get a couple of lists to see if one is more to there liking.

You lose bits fast doing things like that.

As a rough rule of thumb in a human generated passphrase you say 2^4 for the first letter then 2^2 for the next couple of letters and then 2^1.4 from then out…

So 2^8 for first three letters of a word rather than 26^3 ~= 2^14.1 for three random letters. Then each letter there after in a sentance giving around 2^1.4 –rather than the random ~=2^4.7– means to get 84bit equivalent is a natural language sentance with ~57 letters in it rather than 18 for a random letter string.

Again the entropy drops like crazy if the sentance is well known from a, “public body of work” like a poem, quote or from a book or song.

58 letters being the length of the first part of the above sentance, to the comma, is something that can be fairly easily seen, is NOT something the average human is going to be able to easily remember. Each attempt to make it more memorable (or less turgid 😉 is going to knock bits of the entropy count like “pins flying in a bowling alley on competition night”.

Thus a combination of techniques needs to be used. Adding random numbers takes it up around ~2^3 for each which is a quick gain[1]. Likewise random or semi random capitalisation of letters adds bits of entropy. But there is also deliberate “misspellings” even if it’s using “y” instead of “i” or “Q” instead of “u” racks up the brut force trys quite quickly, taking the 2^1.4 back up towards 2^4.7 of each random letter.

But the honest answer to the problem is,

Something you know authentication

Is broken when it comes to humans and machine adversaries.

It’s why I’ve thought about other systems for a number of years. The first part of which is to “rethink the why, the how and the what”.

Passwords come from an age of mechanical teletypes on the end of serial lines and later dial up lines and even VDU’s to some mainframe in an air conditioned room you like as not had no idea where it was. We just don’t roll that way any more. The serial lines have either gone with “local devices” or been replaced with semi-secure TLS over a radio or wired network connection if more than arms length away to fully remote.

Local devices are where we can apply a second layer of authentication “something we have” which is why a lot of people have gone down the “Password manager on a phone” route. The problem is such devices are “hackable” and “grabable” and worse as is seen with LEO behaviour “something you are” is nolonger an authentication factor under your control.

So are there other authentication factors to replace it. The simple answer is yes, we have “location” and “time” we can consider.

The problem with them is that they are not just “restricting in use” they quickly become “obvious in use” if used regularly. Thus we need to add in another asspect which is a “hierarchy of authentication”. That is we have authentication based on their frequency of use and usage.

Thus a swipe of a finger on the screen is fine for making 911 calls but not answering a phone or making a call or text. For that it would be a swipe of the finger against the fingerprint reader, but not for seeing previous texts or listening to previous recordings. For that the entry of a short PIN would be added. The use of timeouts as well with longer PINs for doing other day to day activities. But you also need a “deadman’s switch” which could be anything like swiping your finger in the wrong direction or not far enough. At which point the phone is returned to timed out fully locked for say 30 seconds with a count down timer, if you don’t enter the long pin then it goes into security locked. This is where being at a “time” and “place” as well as using a different long PIN/PASSWORD comes into play.

In essence you build up security layers that need more entropy, but where ever possible it’s of the “something you know” variety, which currently LEO’s etc can not grab[2], which is also the reason for timeouts bumping up security. They can be set so you only have to resist duress for a short time thus the XKCD “$5 wrench” attack nolonger works.

[1] It’s less than you would think because way to many people think “733t speak is n34t”

[2] Various people believe that in the near future “brain wave analysis” techniques will with AI, be able to “read projected thoughts”, sufficiently well to enable people to control machines like wheelchairs. Obviously that would also enable control of a “phantom keyboard” to be used as an input device. Whist this may be possible in the near term and some work suggests it is, it still leaves the rather vexed question of “reading peoples minds”. Some people have postulated that fast imaging of a brain combined with AI technology will alow private thoughts or information to be read out of a persons mind, thus rendering “something you know” available to LEO’s. Whilst personally I’m skeptical, the laws of physics does not prevent it…

Clive Robinson • December 10, 2019 8:03 AM

@ Wael, Curious, MarkH,

And Something on prime numbers as a bonus.

Would appear to be a short snippet of the longer and much more interesting.

https://m.youtube.com/watch?v=hYxBH1YY9z4