Http: added HTTP header with random length to complicate BREACH attack

[fabik]

Here's a nice demonstration of BREACH attack: http://resources.infosecinstitute.com/the-breach-attack/

[JanTvrdik]

Corresponding Symfony issue for reference symfony/symfony#8682

[Majkl578]

Why is this removed?

[fabik]

Because it does not only apply to IE and pages with error status code, but to all browsers and all status codes.

[enumag]

Majkl means someone could want disable the breach protection but enable the IE fix. I don't think it's necessary to have both though. According to the comment above it only applies to IE6, I think we can drop support for such old version.

[fabik]

But it's not possible to turn off the protection, because the variable Response::$preventBreachAttack is private. It just prevents sending of the invisible garbage more than once.

[enumag]

Sorry I missed that.

[Majkl578]

Not really @enumag. I understand IE, I was asking about status codes only. :) But yes, explicit disabling might be good in some cases.

[spaze]

Just for the record, this approach might kill some HTTP caches on the way, for the two guys out there still caring about them. Is also listed as a second least effective mitigation (out of 7), just sayin'

Also, even the guys themselves say that "Unfortunately, we are unaware of a clean, effective, practical solution to the problem.", so making this a default might not be as great idea as it might seem.

My 2 cents.

[dg]

This is of course not intended for „prevent BREACH attack“ just for „complicate“. Subject is incorrect. And I don't like it much, better should be send HTTP header with random length.

[spaze]

That's correct, length hiding makes the attack just take longer:

While this measure does make the attack take longer, it does so only slightly.
The countermeasure requires the attacker to issue more requests, and measure the
sizes of more responses, but not enough to make the attack infeasible.

the BREACH paper

[fabik]

I rewrited it, so it just sends a header of a random length and on HTTPS only. That should be good, shouldn't it?

[dg]

Better is str_repeat. And length should not be random for the same requests.

[pascal-hofmann]

With HTTP compression HTTP headers are not compressed, only the HTTP body is. So adding header data does not prevent BREACH. This PR is nonsense and should be closed.