How embarassing for telegram if what he says is true.
This excerpt was taken from the google translate version of the article.
Crypto is hard to get right, that doesn't mean anyone shouldn't even bother trying.
Likewise, building a hot new secure messaging app with existing well-analyzed, battle-tested cryptographic schemes is generally going to be welcomed.
If you try to do both at once, you're building your application on shaky, untested cryptographical foundations. Cryptographers would similarly probably warn you not to base your application on a new cipher someone else announced at a cryptography conference last week - give it a bit of time for others to analyze it and spot any flaws it might have before you entrust anything sensitive to it.
I could go and try to play around with real time systems for flying planes, it still wouldn't make sense for me to sell you the "safest" plane in the world without going through the proper steps to get such system certified beforehand.
Note: This is not so much in response to thom's comment, but rather the criticism of jimmytucson's comment (which actually has some substance).
They overlooked some security weakness and didn't see it. They didn't do it on purpose like the NSA. In their eyes, the protocol was perfectly secure and most of it is of course.
The only way to avoid this self judgment bias is to use review by other people.
That there is exactly the problem. There is no reason to believe the protocol is secure, and when looking at crypto you start from the assumption that it's "maybe broken", not "perfectly secure". Assuming that some new crypto is secure is hubris.
Pavel previously set up contests with monetary rewards about half the same value for developing a mobile VK app for iOS, Android and WP.
It's not a gamble—it is really an expensive way to find holes. What I don't understand is why they don't hire a crypto consultant instead.
Telegram's response was basically, "LOL. We know better cuz' we got binders full of mathematicians and I'm not listening unless you win our rigged contest designed purposefully to instill a false sense of security in our customers."
Cryptographic software isn't developed the same way as ordinary software. A normal program is launched with tons of bugs and gets fixed over the course of years. Custom crypto solutions should not be delivered to customers in a similar state without explicitly telling them that it hasn't been proven secure.
Because it is dangerous to everyone when non-crypto guys call themselves crypto guys.
Failure is celebrated -- if you make an effort, and try, and you fail, and you learn, and you share what you learned, people care.
But when you (Telegram, not you, XorNot) have a bad idea, and people smarter than you patiently explain why its broken, and you try to buffalo and bluster and bullshit past them, and your product's entire purpose is to provide a security, then your product is worse than a buggy tool; it's worse than not using the product at all. It's the modern-day equivalent of patent-medicine snake oil, and it hurts people.
What you have here is homeopathy.