Re:MEGAChat now includes end-to-end encryption
by hannob on Saturday January 23, 2016 @07:25AM
(#10958149)
The solution is called reproducible builds. I'm aware f-droid is working on these, but most of the work is currently happening more in the Linux distribution space (Debian is leading that cause). I hope in the future we will be able to say "if it doesn't have a reproducible build process it's not to be considered trustworthy".
(Also: I think this question is mostly independent from the disagreements between moxie and f-droid - they were regarding other issues.)
Re:MEGAChat now includes end-to-end encryption
by buu700 on Saturday January 23, 2016 @08:45AM
(#10958363)
Reproducible builds are on our very-near-future roadmap at Cyph, which will be important to mitigate a hypothetical scenario in which we somehow lose our minds and deploy code that differs from the GitHub repo.
We control the whole packaging/signing framework (WebSign), so there isn't any need for us to wait on a third party to roll this out.
Re:MEGAChat now includes end-to-end encryption
by alkonaut on Saturday January 23, 2016 @08:25AM
(#10958301)
Aren't builds typically reproducible? What makes builds non-reproducible? I know the Debian folks are very enthusiastic about reproducible builds, and for a huge system I can see that things can vary -- but what tends to be varying in the case of individual components? If I build a library from the same revision of a source tree I expect the exact bits to come out, if the same compiler version was used? When does it not?
Is it a problem with C/C++ in particular, where toolchains are complex? Or would non-reproducible also be a problem e.g for a Java library?
Googling reproducible builds mainly gets you Debian info, not a general description of the issue
Re:MEGAChat now includes end-to-end encryption
by wila on Saturday January 23, 2016 @09:09AM
(#10958431)
No builds are not easy to reproduce bit by bit as there are much more factors as the actual source.
For example:
- compiler version used
- date / timestamps at the time of compile
- build id's
A good article on this can be found here:
https://blogs.kde.org/2013/06/19/really-source-code-software
Re:MEGAChat now includes end-to-end encryption
by sandworm101 on Saturday January 23, 2016 @08:30AM
(#10958318)
A build is not reproducible where it incorporates both open and closed code. A company may "open source" some important stuff (encryption algorithms) but keep enough code closed that the final build cannot be duplicated from the "open" code. It is therefore difficult to judge whether the open code has actually been adopted properly, or whether any closed code hasn't created new bugs.
Don't get this wrong please: this is still probably THE best solution right now, which simply as well has its security limitations, right?
Is there any work ongoing on a kind of "verification process" for Apps like these? How can an end-user tell if the binaries running on their device are untempered with?
/sorry for hijacking ;)