[Busec] WhatsApp default settings vulnerability

Ran Canetti canetti at tau.ac.il
Mon Jan 16 18:58:28 EST 2017


yes, if the OS is completely malicious and you dont have secure enclaves 
then all bets are off, but an application can still try to protect 
itself from incompetent but not malicious OS, even without enclaves, to 
reduce the trusted code base. A secure messaging system that prides 
itself on being end-to-end security must take these issues into 
consideration...

and yes the metadata issue is another elephant in the room...

Ran

On 1/15/2017 2:54 AM, Ari Trachtenberg wrote:
> Not sure how you can protect from the OS without heavy-duty crypto or 
> some trusted computing base (often an attack surface in its own right).
> The OS, for example, can completely replace the app, at its discretion.
>
>> On Jan 14, 2017, at 8:28 PM, Ran Canetti <canetti at tau.ac.il 
>> <mailto:canetti at tau.ac.il>> wrote:
>>
>> Question: Is anyone aware of a study of the level of protection that 
>> Whatsapp/Signal  gives from the OS itself, or from other applications 
>> on the phone? If anything, this attack surface appears to me much 
>> more scary than these re-encryption buglets... presumably the 
>> Signal/Whatsapp application keeps a lot of sensitive information -  
>> public keys, secret keys, buffered messages, etc - both on RAM and on 
>> secondary storage. I once tried to look at what Whatsapp say about 
>> this in their documentation but didnt find much.
>> Ran
>>
>>
>> On 1/13/2017 5:32 PM, Sarah Scheffler wrote:
>>> I mean, calling it a vulnerability definitely makes it sound worse 
>>> than it is, but I also think that a lot of people basically assume 
>>> that as long as they're using WhatsApp, nothing they send will be 
>>> read by anyone other than who they're sending it to.  I think 
>>> calling this a vulnerability in the news is actually good, as it 
>>> brings public awareness of the issue, and now people know whether or 
>>> not they want to check the box, or look at other settings. Perhaps 
>>> my email could have been named with less hype, but to be honest this 
>>> /is/ a vulnerability as far as most users' usage is concerned, and I 
>>> think it's fine to treat it as such.  At the very least, this will 
>>> hopefully make people think "hey, there are things that are not 
>>> automatically solved by me using WhatsApp."  Which is obvious to 
>>> people used to thinking about cryptography, but not to the average 
>>> person, who's basically been showered with advice that WhatsApp will 
>>> solve all of their privacy problems.
>>>
>>> Also, I think a much better thing would have been for WhatsApp to 
>>> start with Signal's behavior, with a little blurb that says "if you 
>>> don't want to see these messages anymore, check this box."  I think 
>>> opting out, in general, is better than opting in.  That way, if 
>>> people are going to click through, they can check the box and it's 
>>> the same end result.  And if they're not going to click through, 
>>> then we helped some people have a little more security at the cost 
>>> of verifying a key change once every month or so (or whatever the 
>>> rate of their friends getting new phones is).
>>>
>>> But it's fair, causing a panic about a not-really-vulnerability is 
>>> only going to make it worse when a /real/ vulnerability comes along. 
>>> So I don't know.  Information is difficult.
>>>
>>> Cheers!
>>> Sarah
>>>
>>> PS: If anyone wants to participate in the MIT Mystery Hunt this 
>>> weekend and doesn't have a team, I have a team of people from Harvey 
>>> Mudd College and we're always looking for new team members; send me 
>>> an email if you want into our slack room.
>>>
>>> On Fri, Jan 13, 2017 at 5:00 PM Mayank Varia <varia at bu.edu 
>>> <mailto:varia at bu.edu>> wrote:
>>>
>>>     Hi Sarah,
>>>
>>>     I think Signal is overhyped sometimes, but calling this a
>>>     "vulnerability" or a "backdoor" seems way overblown to me. It's
>>>     important that Signal/WhatsApp supports key migration somehow,
>>>     since keys can change for many innocuous reasons, such as simply
>>>     un/reinstalling the program on your phone or recovering your
>>>     entire phone state from a backup snapshot (which, at least in my
>>>     case, didn't save my old keys). For a long time Signal also made
>>>     notifications of key changes unobtrusive by default; I had to
>>>     enable the warning messages manually on my phone.
>>>
>>>     Basically, nothing about this post seems like news to me; it's a
>>>     conscious decision by the developers of a security software to
>>>     provide the best security/usability tradeoff to their customers
>>>     as they can. Compare to the alternative. If the billion(ish)
>>>     WhatsApp users received one of those "security warning" messages
>>>     every time any single one of their friends migrated to a new
>>>     key, I'm pretty sure people would be overburdened by these
>>>     messages and would quickly learn to ignore them and simply click
>>>     through. I don't see any benefit to this strategy at all. Signal
>>>     itself only seems to be able to handle a "warn by default"
>>>     mechanism because its user base is currently smaller and more
>>>     tech-savvy/paranoid than WhatsApp's.
>>>
>>>     FYI, Open Whisper Systems' official response is here:
>>>     https://whispersystems.org/blog/there-is-no-whatsapp-backdoor/.
>>>     I agree with the criticism that the Guardian never bothered to
>>>     ask the experts they interviewed about the (so-called)
>>>     vulnerability, but rather the unrelated and completely-leading
>>>     question "are backdoors in crypto bad?" That's all that the
>>>     quotes in the Guardian article seem to indicate, as I read it.
>>>
>>>     Mayank
>>>
>>>     P.S. for a shameless plug: if you want to learn more details
>>>     about the Signal messaging protocol, take my applied crypto
>>>     course at BU this semester (CS 591 V1).
>>>
>>>
>>>     On Fri, Jan 13, 2017 at 4:42 PM Sarah Scheffler <sscheff at bu.edu
>>>     <mailto:sscheff at bu.edu>> wrote:
>>>
>>>         This might be old news for some of you, but it was news to
>>>         me.
>>>         <https://www.theguardian.com/technology/2017/jan/13/whatsapp-backdoor-allows-snooping-on-encrypted-messages>
>>>
>>>
>>>         TL;DR: If you use Signal, you're good.  If you use WhatsApp,
>>>         you should set the setting where it tells you if the
>>>         recipient's key was changed while they were offline, and
>>>         also be aware that messages sent to people who are offline
>>>         may be re-encrypted under a different (!) key and sent
>>>         without your intervention.  Or switch to Signal.
>>>
>>>         Basically if you send a message in WhatsApp to someone who
>>>         is offline, WhatsApp can replace the public key of the
>>>         person to whom you're sending with a new one, and the
>>>         messages you sent will be automatically re-encrypted and
>>>         sent under the new key.  Only after they are successfully
>>>         transmitted are you told that this key change happened, and
>>>         even then only if you check a little (non-default) box that
>>>         says so.  It was explained a little more sanely and with
>>>         more pictures by the finder, Tobias Boelter from Berkeley:
>>>         https://tobi.rocks/2016/04/whats-app-retransmission-vulnerability/
>>>
>>>         Apparently Facebook knows about this and isn't planning on
>>>         changing anything.  The finder of this vulnerability says
>>>         <https://tobi.rocks/2017/01/what-is-facebook-going-to-do-a-suggestion/> he's
>>>         pretty sure it was a bug, but also that they should claim
>>>         that it wasn't and that they just made a poor design choice,
>>>         and change it.
>>>
>>>         Cheers!
>>>         Sarah
>>>
>>>         _______________________________________________
>>>         Busec mailing list
>>>         Busec at cs.bu.edu <mailto:Busec at cs.bu.edu>
>>>         http://cs-mailman.bu.edu/mailman/listinfo/busec
>>>
>>>
>>>
>>> _______________________________________________
>>> Busec mailing list
>>> Busec at cs.bu.edu
>>> http://cs-mailman.bu.edu/mailman/listinfo/busec
>>
>> _______________________________________________
>> Busec mailing list
>> Busec at cs.bu.edu <mailto:Busec at cs.bu.edu>
>> http://cs-mailman.bu.edu/mailman/listinfo/busec
>
> ---
> Prof. Ari Trachtenberg  ECE, Boston University
> trachten at bu.edu <mailto:trachten at bu.edu> http://people.bu.edu/trachten
>
>
>
> _______________________________________________
> Busec mailing list
> Busec at cs.bu.edu
> http://cs-mailman.bu.edu/mailman/listinfo/busec

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://cs-mailman.bu.edu/pipermail/busec/attachments/20170117/f2a04149/attachment-0001.html>


More information about the Busec mailing list