RE: [Algorithms] Message signature in token ring

[Tony C. <to...@mi...>]

>It doesn't seem to make much sense to use built-in OS encryption rather
>than a freely available or commercial 3rd party one with source, =
because
>
>1) You can't proof or verify code you have only in binary form,


No offense, but given that you are not a security expert, you probably =
couldn't prove/verify code you saw in source form. Neither could I. I =
mean, you could give it a general skim, but that's not really the same =
thing.

>2) You can't port or ensure interoperability between platforms if you =
have
>to rely on the encryption native to each platform,
=20
There are implementations of standard algorithms, they should be =
compatible.

>3) Um, how many times a year do they find gaping security holes in
>Windows?  Why would we assume that an OS cryptography layer is =
magically
>all that great?
=20
They find security holes in all OSes. I will note that none of them, to =
my knowledge, have involved the Crypto API, probably because it's one of =
the most carefully reviewed pieces of code in the system.
=20
Your comments about Windows in general also miss an important point - =
Windows has millions of lines of code, and is used by millions of people =
every day, including a very large number of malicious attackers. Games =
have considerably less code, a much smaller user base, and a smaller =
absolute number of malicious attackers. Yet games frequently have bugs. =
It's quite likely that the bug rate in games per line-of-code or =
per-user is higher than in Windows. And yet you're talking about writing =
brand new code in a notoriously tricky area. If you think it's going to =
be less error prone than the heavily used and reviewed code in the OS, =
then you are dreaming.
=20
In any event, my argument was not particularly about Windows (although =
you seem to want to turn it into one). My point was a caution against =
rolling your own. If it makes you feel warm and fluffy to have the =
source code, then use some thoroughly peer-reviewed open source library. =
What you should not do is learn about this stuff for the first time by =
reading a couple of books, download some student's homework assignment =
implementation from the web, hack it about a bit, ask your buddy in the =
next-door office to take a quick look at it and then sit back and assume =
that your testers will find all the bugs.
=20
In short: DO NOT ROLL YOUR OWN UNLESS YOU HAVE NO OPTION
=20
Tony Cox - Development Lead, Hockey
Microsoft Game Studios - Sports