Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Show HN: Turtl – A secure, encrypted Evernote alternative

turtl.it

Post reply on HN
Print view View original

51–60 of 100 posts

Re: Show HN: Turtl – A secure, encrypted Evernote alternative

#51

Post by sarciszewski »

mrmondo wrote: post #50
sarciszewski wrote: post #49

https://github.com/turtl/js/blob/2ca59900d71284795e278e75585... Nothing says secure like PBKDF2-SHA1 with 50 rounds. https://github.com/turtl/js/blob/2ca59900d71284795e278e75585... ...or timing attacks on MAC validation. (Yeah, you switched to GCM, but a downgrade attack could potentially be used to find a valid MAC for a chosen ciphertext without the attacker knowing the key...)

"When the standard was written in 2000, the recommended minimum number of iterations was 1000, but the parameter is intended to be increased over time as CPU speeds increase. As of 2005 a Kerberos standard recommended 4096 iterations,[2] Apple iOS 3 used 2000, iOS 4 used 10000,[3] while in 2011 LastPass used 5000 iterations for JavaScript clients and 100000 iterations for server-side hashing" source: https://en.wikip…

Nope, PBKDF1 was a typo.

Re: Show HN: Turtl – A secure, encrypted Evernote alternative

#52

Post by logicallee »

mrmondo wrote: post #50
sarciszewski wrote: post #49

https://github.com/turtl/js/blob/2ca59900d71284795e278e75585... Nothing says secure like PBKDF2-SHA1 with 50 rounds. https://github.com/turtl/js/blob/2ca59900d71284795e278e75585... ...or timing attacks on MAC validation. (Yeah, you switched to GCM, but a downgrade attack could potentially be used to find a valid MAC for a chosen ciphertext without the attacker knowing the key...)

"When the standard was written in 2000, the recommended minimum number of iterations was 1000, but the parameter is intended to be increased over time as CPU speeds increase. As of 2005 a Kerberos standard recommended 4096 iterations,[2] Apple iOS 3 used 2000, iOS 4 used 10000,[3] while in 2011 LastPass used 5000 iterations for JavaScript clients and 100000 iterations for server-side hashing" source: https://en.wikip…

how can anyone take anything seriously that requires 1000 iterations. if iterations had an exponential function then the difference between 1000 iterations and 10,000 is ridiculous, like the difference between 16 bit encryption (no such thing, this would be a joke) and 160 bit encryption.

But if it's not exponential, then adding thousands of iterations doesn't do much. Maybe instead of 1 day to break it, now it takes 1000 days, or instead of spending $4,000 on CPU time now you need to spend $4M to crack it. Big deal - it's just as broken.

So can anyone explain the math here? How can they seriously suggest linearly increasing the number of iterations over time?

Re: Show HN: Turtl – A secure, encrypted Evernote alternative

#53

Post by mrmondo »

marknutter wrote: post #46
mrmondo wrote: post #28

Earlier quoted context omitted.

Not from what I've seen, after hosting a bunch of NodeJS based web-apps I can tell you it doesn't perform well on the server-side or on the desktop. It's a decent rapid prototyping language / tool - but it's taking what should (or perhaps shouldn't) be in a web browser, wrapping it up and hosting it somewhere that the designers of products often don't know much about the landscape.

For the majority of users, it performs well enough. Step out from behind your own lens. For instance, the number of times I've heard someone complain about Slack's performance (that isn't a developer) is precisely zero .

I think my lens is actually quite large, I have 15 years of production experience hosting, building, deploying applications and build pipelines, currently I'm working across 20~ end user facing products in their various application life cycles, backed by hundreds of services running across many hundreds of instances. They scale from applications that turn over only 3-5RPM up to 8000 RPM throughout the day. We host and deploy Python, PHP, Perl, Ruby, JS, Erlang, Go and Java apps. I can say without a doubt that the JS (in our case NodeJS) apps are by far the worst to work with, not just in regards to performance but also in regards to community quality.

*Edit: clarification

Re: Show HN: Turtl – A secure, encrypted Evernote alternative

#54

Post by sarciszewski »

logicallee wrote: post #52
mrmondo wrote: post #50

Earlier quoted context omitted.

"When the standard was written in 2000, the recommended minimum number of iterations was 1000, but the parameter is intended to be increased over time as CPU speeds increase. As of 2005 a Kerberos standard recommended 4096 iterations,[2] Apple iOS 3 used 2000, iOS 4 used 10000,[3] while in 2011 LastPass used 5000 iterations for JavaScript clients and 100000 iterations for server-side hashing" source: https://en.wikip…

how can anyone take anything seriously that requires 1000 iterations. if iterations had an exponential function then the difference between 1000 iterations and 10,000 is ridiculous, like the difference between 16 bit encryption (no such thing, this would be a joke) and 160 bit encryption. But if it's not exponential, then adding thousands of iterations doesn't do much. Maybe instead of 1 day to break it, now it takes…

Attackers can use special hardware (ASICs, for example) to perform a lot of low-memory but CPU-intensive calculations quickly. Modern KDFs emphasize a property called memory-hardenss: It should be reasonably fast on consumer devices, require a reasonable amount of memory, and trading off memory usage should require an absurd CPU slowdown.

PBKDF2 doesn't have this property, so if you're forced to use it, the standard recommended iteration count is 86000.

50 is a joke.

Re: Show HN: Turtl – A secure, encrypted Evernote alternative

#55

Post by neoeldex »

biokoda wrote: post #20
triplesec wrote: post #17

Earlier quoted context omitted.

TL;DR for parent comment: "It's not as mature as Evernote, and it has a bigger footprint." Therefore your comment is unfair: if provides a less complex evernote function that many could use. Neither of these is unfixable over time, as the product develops, and some may be a function of the increased security implementation. None of these criticisms stops the app being usable, as far as I can see.

> Neither of these is unfixable over time JS on desktop is never going to perform well.

Howso?

Just like the new macbook pro's, which are thermalthrottling as soon as you look at pictures of cats. You could say they never going to perform well, but then you could always throw it in a cold bath.

Re: Show HN: Turtl – A secure, encrypted Evernote alternative

#56

Post by logicallee »

sarciszewski wrote: post #54
logicallee wrote: post #52

Earlier quoted context omitted.

how can anyone take anything seriously that requires 1000 iterations. if iterations had an exponential function then the difference between 1000 iterations and 10,000 is ridiculous, like the difference between 16 bit encryption (no such thing, this would be a joke) and 160 bit encryption. But if it's not exponential, then adding thousands of iterations doesn't do much. Maybe instead of 1 day to break it, now it takes…

Attackers can use special hardware (ASICs, for example) to perform a lot of low-memory but CPU-intensive calculations quickly. Modern KDFs emphasize a property called memory-hardenss: It should be reasonably fast on consumer devices, require a reasonable amount of memory, and trading off memory usage should require an absurd CPU slowdown. PBKDF2 doesn't have this property, so if you're forced to use it, the standard…

so it's not really exponential - if ridiculous amounts of memory somehow became very cheap and available, then it would break it? this is quite different from the exponential properties most encryption has. I would expect "ridiculous amounts of memory" to mean "more bytes than the number of atoms on Earth", that sort of thing. It sounds like rather than these kinds of theoretical limits, they chose much more practical limits - which seems a lot more dangerous and less future-proof, but I guess I'm not an expert.

Re: Show HN: Turtl – A secure, encrypted Evernote alternative

#57

Post by biokoda »

neoeldex wrote: post #55
biokoda wrote: post #20

Earlier quoted context omitted.

> Neither of these is unfixable over time JS on desktop is never going to perform well.

Howso? Just like the new macbook pro's, which are thermalthrottling as soon as you look at pictures of cats. You could say they never going to perform well, but then you could always throw it in a cold bath.

Because it implies running on top of a VM and much worse rendering data through HTML/css.

Re: Show HN: Turtl – A secure, encrypted Evernote alternative

#58

Post by biokoda »

LeonidasXIV wrote: post #31
biokoda wrote: post #20

Earlier quoted context omitted.

> Neither of these is unfixable over time JS on desktop is never going to perform well.

Then get rid of JS. I doubt Turtl is married to JS, it just happens to be the greatest common denominator.

A complete client side code rewrite to get rid of something is pretty much married to it.

Re: Show HN: Turtl – A secure, encrypted Evernote alternative

#59

Post by NoGravitas »

orthecreedence wrote: post #29
colordrops wrote: post #3

I saw the "pricing" link at the top and was about to shit all over this for a privacy oriented app being closed source, until I scrolled to the very bottom and saw the innocuous open source note. May want to emphasize that more.

This app has been a balance between advertising the fact that it's open source and just presenting itself as a "privacy app" without shoving it in people's faces. Maybe it should be a bit more prominent (and thanks for reading to the bottom)! To be clear, the pricing section only applies if you connect Turtl to our hosted service. If you run your own server, the only pricing that applies is the price you pay for the…

I actually love this business model: web/api server and clients are all Free Software, and the developer or company behind it runs the "canonical" hosted service. I hope it works out for you.

Re: Show HN: Turtl – A secure, encrypted Evernote alternative

#60

Post by phlyingpenguin »

polm23 wrote: post #24

Is something wrong with the CSS on the site? Everything looks kind of overly large in my browser (Desktop Linux Chrome).

Just to add in a screenshot, here's what it looks like on my copy of Chrome on Linux: http://i.imgur.com/6YPbAg6.png
Post reply on HN