2

The Wikipedia article on TLS article has a tremendously long list of attacks against SSL. A table on that article shows a survey from September 2015. If you look at that survey, it shows that nearly two thirds of a selection of the most popular sites online are insecure.

I assume actually implementing these attacks is tremendously difficult and complex, but it still seems a little odd that ~66% of websites have known security shortfalls and 90% are susceptible to the BEAST attack, known about since 2011, if used with an old browser.

This leads me to basically having zero faith in SSL. Is this overly alarmist? Does that survey somehow paint a picture worse than it really is? Does TLS/SSL seem like a really broken concept if it's repeatedly circumvented from a tremendous number of angles?

Improve this question
3
  • It might depend a lot on the ciphers used - many of them are insecure/cracked, but most companies don't care. Check howsmyssl.com for more details. I have locked all insecure ciphers in my browser (Chrome supports that), and for some banks, I cannot connect anymore.
    – Aganju
    Dec 28 '15 at 3:33
  • A bit of subtlety: Part of the reason sites are still susceptible to BEAST is that they maintain backwards compatibility for older clients (that require TLS 1.0). If you are using a modern browser you should be using TLS 1.2 when connecting to a server (barring a downgrade attack). So even though 66% percent of servers are vulnerable, connecting to them via TLS 1.2 eliminates any risk (regarding BEAST) on your end.
    – puzzlepalace
    Dec 28 '15 at 7:04
  • All that survey says to me is that web admins don't know how to configure TLS properly. Also, to my understanding, it's fairly well accepted that the standard point of BEAST attack mitigation is the user's browser.
    – d0nut
    Dec 30 '15 at 16:46
2

Is this overly alarmist?

To a point, yes, it is alarmist. Certainly better to be concerned and care about securing communications though.

Does that survey somehow paint a picture worse than it really is?

You may not be taking in enough understanding about the types of attacks or requirements to launch one, causing the slight jump to mistrust. Such as you make no mention to understand that this survey reflects site owner's configurations and not representative of the TLS/SSL packages current or updated state such as the latest from openssl.

Does TLS/SSL seem like a really broken concept if it's repeatedly circumvented from a tremendous number of angles?

No, a symbol of a strong system is to be able to find and fix mistakes. To have a plan to patch or mitigate with supported upgrade paths and a large community assisting with expertise and monetary means. Take your jump at beast attack numbers, you could easily brag about those heartbleed numbers being low and have opposing confidence. Also openssl and other implementation packages are not to blame here, this survey paints the picture of what are the site owner's configurations not the state of SSL/TLS package capabilities that are well maintained and updated.

Improve this answer
3
  • 1
    I'm not sure I agree having mistakes and fixing them is a signal of a strong system. It's a signal of a system that is fallible but fixable. It means it was the best of available options, not that it's the best option possible. Compare SSL to something like one-time pads - something that works today and is guaranteed to work tomorrow if you follow extremely simply guidelines (use real randomness, don't re-use pads, don't reveal your pad). It seems like there has to be better options that don't rely on servers/browsers being configured perfectly.
    – Vayeate
    Dec 28 '15 at 4:22
  • And obviously one-time pads aren't really comparable to the complexity that is HTTP, but my point is more that it's possible to have a system that is relatively fool-proof and not susceptible to relatively small mistakes. There seem to be a lot of papers out there regarding anonymous delivery protocols, and it seems like the majority of these would solve the issues SSL has (though at some cost of latency/complexity/etc). When it comes to highly sensitive data though, it seems like that cost would be worth it.
    – Vayeate
    Dec 28 '15 at 4:27
  • 2
    Perhaps you don't really understand what SSL provides @Vayeate. I suspect that none of the systems even attempts to handle the variety and complexity that SSL does. And if they do, they undoubtedly suck at it because they haven't been hammered on by hackers and researchers alike for 2 decades.
    – Neil Smithline
    Dec 28 '15 at 4:45

Not the answer you're looking for? Browse other questions tagged or ask your own question.