[mu53]

I think the scenario that security through obscurity fails is when the end user is reliant on guarantees that don't exist.

For example Intel's Management Engine, it was obscured very well. It wasn't found for years. Eventually people did find it, and you can't help but wonder how long it took for bad actors with deep pockets to find it. Its this obscured cubby hole in your CPU, but if someone could exploit it, it would be really difficult to find out because of intel's secrecy on top of the feature.

[_yb2s]

It seems like people are really talking about different things with obscurity. Some are referring to badly designed weak systems, where secrecy and marketing hype is used to attempt to conceal the flaws. Others, like my comment above, are talking about systems carefully engineered to have no predictable or identifiable attack surfaces- things like OpenBSDs memory allocation randomization, or the ancient method of simply hiding physical valuable things well and never mentioning them to anyone. I’ve found when it is impossible for an external bad actor to even tell what OS and services my server is running- or in some cases to even positively confirm that it really exists- they can’t really even begin to form a plan to compromise it.

[nkmskdmfodf]

> where secrecy and marketing hype is used to attempt to conceal the flaws.

That's literally the practical basis of security through obscurity.

> Others, like my comment above, are talking about systems carefully engineered to have no predictable or identifiable attack surfaces- things like OpenBSDs memory allocation randomization,

That's exactly the opposite of 'security through obscurity' - you're literally talking about a completely open security mitigation.

> I’ve found when it is impossible for an external bad actor to even tell what OS and services my server is running- or in some cases to even positively confirm that it really exists- they can’t really even begin to form a plan to compromise it.

If one of your mitigations is 'make the server inaccessible via public internet', for example - that is not security through obscurity - it's a mitigation which can be publicly disclosed and remain effective for the attack vectors it protects against. I don't think you quite understand what 'security through obscurity[0]' means. 'Security through obscurity' in this case would be you running a closed third-party firewall on this sever (or some other closed software, like macos for example) which has 100 different backdoors in it - the exact oppposite of actual security.

[0] https://en.wikipedia.org/wiki/Security_through_obscurity

[_yb2s]

You're mis-representing my examples by shifting the context, and quoting a wikipedia page that literally gives the same examples to two of the main ones I mentioned at the very top of the article as key examples of security through obscurity: "Examples of this practice include disguising sensitive information within commonplace items, like a piece of paper in a book, or altering digital footprints, such as spoofing a web browser's version number"

If you're not understanding how memory allocation randomization is security through obscurity- you are not understanding what the concept entails at the core. It does share a common method with, e.g. using a closed 3rd party firewall: in both cases direct flaws exist that could be overcome with methods other than brute force, yet identifying and specifying them enough to actually exploit is non-trivial.

The flaw in your firewall example is not using obscurity itself, but: (1) not also using traditional methods of hardening on top of it - obscurity should be an extra layer not an only layer, and (2) it's probably not really very obscure, e.g. if an external person could infer what software you are using by interacting remotely, and then obtain their own commercial copy to investigate for flaws.

[nkmskdmfodf]

> You're mis-representing my examples by shifting the context,

Specific example of where I did this?

> literally gives the same examples to two of the main ones I mentioned at the very top of the article as key examples of security through obscurity: "Examples of this practice include disguising sensitive information within commonplace items, like a piece of paper in a book, or altering digital footprints, such as spoofing a web browser's version number"

I mean, I don't disagree that what you said about changing port numbers, for example, is security through obscurity. My point is that this is not any kind of defense from a capable and motivated attacker. Other examples like the OpenBSD mitigation you mentioned are very obviously not security through obscurity though.

> If you're not understanding how memory allocation randomization is security through obscurity- you are not understanding what the concept entails at the core.

No, you still don't understand what 'security through obscurity' means. If I use an open asymmetric key algorithm - the fact that I can't guess a private key does not make it 'security through obscurity' it's the obscuring of the actual crypto algorithm that would make it 'security through obscurity'. Completely open security mitigations like the one you mentioned have nothing to do with security through obscurity.

> The flaw in your firewall example is not using obscurity itself, but: (1) not also using traditional methods of hardening on top of it

Sooo... you think adding more obscurity on top of a closed, insecure piece of software is going to make it secure?

> if an external person could infer what software you are using by interacting remotely,

There are soooo many ways for a capable and motivated attacker to figure out what software you're running. Trying to obscure that fact is not any kind of security mitigation whatsoever. Especially when you're dealing with completely closed software/hardware - all of your attempts at concealment are mostly moot - you have no idea what kind of signatures/signals that closed system exposes, you have no idea what backdoors exist, you have no idea what kind of vulnerable dependencies it has that expose their own signatures and have their own backdoors. Your suggestion is really laughable.

> not also using traditional methods of hardening on top of it

What 'traditional methods' do you use to 'harden' closed software/hardware? You literally have no idea what security holes and backdoors exist.

> if an external person could infer what software you are using by interacting remotely, and then obtain their own commercial copy to investigate for flaws.

Uhh yeah, now you're literally bringing up one of the most common arguments for why security through obscurity is bullshit. During WW1/WW2 security through obscurity was common in crypto - they relied on hiding their crypto algos instead of designing ones that would be secure even when publicly known. What happened is enough messages, crypto machines, etc were recovered by the other side to reverse these obscured algos and break them - since then crypro has pretty much entirely moved away from security through obscurity.

[_yb2s]

You are operating on a false dichotomy that the current best practices of cryptographic security, code auditing, etc. are somehow mutually exclusive with obscurity, and then arguing against obscurity by arguing for other good practices. They are absolutely complementary, and implementing a real world secure system will layer both- one starts with a mathematically secure heavily publicly audited system, and adds obscurity in their real world deployment of it.

If there are advantages to a closed source system, it is not in situations where the source is closed to you and contains bugs, but when closed to the attacker. If you have the resources and ability to, for example, develop your own internally used but externally unknown, but still heavily audited and cryptographically secure system, is going to be better than an open source tool.

[nkmskdmfodf]

> They are absolutely complementary, and implementing a real world secure system will layer both- one starts with a mathematically secure heavily publicly audited system, and adds obscurity in their real world deployment of it.

Ok, let's start with a 'mathematically secure heavily public audited system' - let's take ECDSA, for example - how will you use obscurity to improve security?

> If you have the resources and ability to, for example, develop your own internally used but externally unknown, but still heavily audited and cryptographically secure system, is going to be better than an open source tool.

Literally all of the evidence we have throughout the history of the planet says you're 100% wrong.