Audit Cleared Google Privacy Practices Despite Security Flaw | Hacker News Reader

Audit Cleared Google Privacy Practices Despite Security Flaw

thehill.com

35 pointsinfodocket19 comments

Open articleView on HN

Ernst & Young also give a positive audit to Startssl/Wosign even though they had problems at the time.

For that reason Mozilla "no longer accept audits carried out by Ernst & Young Hong Kong." [0].

[0]: https://blog.mozilla.org/security/2016/10/24/distrusting-new...

To be fair, I have never seen an external security audit do anything useful.

Either they bend to the folks paying the bills, or they flag a zillion, generally known to engineers, issues.

If you really want to make you security audit useless, hand the auditors C code. They will stare at you as though you just handed them the head of a small child.

I have personally been a part of security audits that have found critical vulnerabilities in some of the best-known tech companies on the globe.

Just because there exists a world of auditors who function as rubber stamps does not mean we all do.

I've been part to numerous audits during my career and there's always an initial findings report with the opportunity to explain away findings such that they never show up on the final report.

If that's the case then you haven't been working with good auditors.

That said, the opportunity to do this (scrub reports) is a driving decision behind many companies' choice to work with more permissive firms. If you want a rubber stamp, you can certainly get it, but if you want a true partner who will find and exploit issues you can get that too.

> if you want a true partner who will find and exploit issues you can get that too.

Could you name such a company? This is a perfect opportunity to get to know a good auditor, actually I can't imagine how would one look for a good auditor otherwise.

This is a witch-hunt by Murdoch, I'm guessing. 500,000 people's information with the potential to have leaked, and yet Facebook and Equifax fucking skate on DEMONSTRABLY having leaked well over half of America's personal information to domestic and foreign interests alike.

This is sour grapes of the worst kind.

I mean, what is it at this point, 2 days of bad press? All Murdoch properties I'm aware of covered Facebook and Equifax for a lot longer than that, and certainly didn't pretend that what they did was okay.

...Can you show proof that Murdoch properties didn't cover Facebook and Equifax breaches?

(hint: you can't, because they covered them extensively)

Agree. When did Murdoch start with his vendetta towards Google?

It's only a 'potential' because they failed to keep any auditable logs of the behavior of unvetted third parties, AND failed to vet them in any meaningful fashion.

The amount and type of data exposed isn't as bad but this is pretty similar to the Cambridge Analytica scandal, except in this case Google literally didn't even notice. Facebook at least noticed.

This story doesn't make any sense. An FTC-mandated E&Y "Privacy Audit" isn't a software security assessment and would never have been expected to either (a) uncover subtle application security flaws or (b) metabolize internal reports of subtle, fixed flaws. Nor, for that matter, is the internal discovery of a software security flaw a "breach": routine internal discovery of vulnerabilities is what good security teams do.

The reporting on this G+ story has been startlingly bad. There's apparently some juice to the narrative that Google is running headlong into a collision with the Trump administration, and so reporters seem to be starting with a public policy conclusion and working their way back. Somebody needs to start informing these people how professional software security works.

In any audit situation --- PCI, HITECH, SOC-2, you name it --- there is neither a norm nor a duty to notify auditors of internally discovered security problems. Nor would any such norm be productive: it would warp internal incentives, both to discover and to properly handle vulnerabilities.

If the auditors discover a security problem, that problem gets documented, just like it (typically) does with external reporters (researchers, bug bounty claimants). When that happens, you have an indication that your internal software security controls failed. But when your team finds a bug in its own code? That's a sign that the team is doing things right.

All of your characterization here is correct, but does it really make sense to classify it as an internally discovered security problem instead of a potential-or-real-breach when it was present for years, trivially exploitable by third parties, and is impossible to vet? There's no way to know whether or not it was used because there was no logging in place for third-party use of user data, and apparently no audits (manual or automated) were ever performed of these APIs. Google more or less admitted that all they had to go on here was an examination of the third parties who potentially had access to the data, because they have no way to know what they actually did with it and apparently can't contact them to find out.

The reporting is definitely exaggerating the significance of the data that was exposed and making it sound like it's known that the data was collected by third parties, but it's absolutely the case that the ball was dropped here and it's inappropriate to pretend nothing happened. There's simply no way to know, which is itself a breach of user trust.

It's kind of a no-win situation here given the way the media reports on these matters, but attempting to keep it quiet was incredibly foolish - everything remotely bad that happens at Google is going to get leaked by internal troublemakers now, and legal/pr should know that and should have gotten ahead of this.

There is nothing remarkable about finding a vulnerability that had been latent in a codebase for years. It is not clear to me what "ball" you believe has been "dropped" here. If it's that the first N groups of security testers that looked at this code missed the vulnerability that their N+1th found, sorry, no: that happens all the time.

If the vulnerability is 'any one of our API users could have accessed <large set of private data> over the past 3 years' that's a lot more than 'we found an obscure latent security issue'. Especially if you can't prove it didn't happen by examining logs (which they can't, because they didn't keep logs) or auditing your API users (which they couldn't).

It doesn't feel like splitting hairs to separate this from things like 'we found an obscure xss exploit internally which is now fixed'. It is explicitly a Possible Breach, not just a Bug. If you're arguing that Possible Breaches don't require disclosure, well, that's a choice to make, but it worked out poorly in this case and it will probably keep working out poorly because everything G does is under a microscope until the far right gets bored with them.

There’s a strong argument against disclosure of internally discovered bugs (security or otherwise) that have no evidence of user impact. Internal efforts to find and fix bugs improve actual quality/safety but, if indiscriminately announced, harm perceived quality/safety. We have seen repeatedly how bad security reporting of actual weaknesses can drive users to more risky alternatives or behaviors, lowering actual security and privacy. It’s just too easy to say that all major security defects should be announced or that no major defects should ever exist.

There are about 430 developers that could have exploited this vulnerability. There was no evidence in the available two weeks of logs of it being exploited. And the number of users and types of data available make Google+ a low value target. The decision to not announce was reasonable, common, legal, and moral.

It sets a poor standard to suggest they should have acted different because they’re now under fire from people with far darker motives than reporting truth or advocating for consumer privacy. Those exploiting the story for their own gain should be shamed rather than capitulated to.

The vulnerability that was discovered here was less significant than an "obscure XSS exploit", not that it matters at all to the analysis.

An xss bug exposes even more data than this vuln. Why would that be less severe?

> but does it really make sense to classify it as an internally discovered security problem instead of a potential-or-real-breach

Literally all security bugs can be described this way. "Company closes security bug" is not really news, nor should it be. How many XSS bugs have been closed at your place of business? Those are literally all potential breaches. Should each one of those get published in the news? Or does this requirement only exist for the major companies?

I'd wager that FAANG companies are closing thousands of similar bugs annually. The only reason this one is treated differently is the chosen narrative by the journalists.

Reply on news.ycombinator.com