[eridius]

The URLs contain a long random string in them, no? This makes them unguessable. Public-by-default seems to enable easy sharing of documents. I would hope that corporate plans would impose stricter privacy controls, but for personal use, being able to share a document just by sharing its URL seems very convenient. Though I suppose defaulting the permissions such that anyone with the URL can edit the document seems maybe too permissive.

[mevile]

This is security by obscurity, it's still technically possible to access the documents. Like if you accidentally revealed the URL somehow. You shouldn't have to worry about things like that.

[jph]

Surprisingly, it turns out that using security based on a URL with a random string is /not/ security by obscurity.

The security pattern of a URL with a random string is security-equivalent to a pubic username and a random string password, and also equivalent to the security pattern of a bearer token: so long as the URL is shared only with authorized users, it's the same security hardness.

The pattern can tune the security by using more randomness, such as more characters. There are implementation areas to consider, such choosing a random number generator with high quality randomness like /dev/urandom. There are some access control areas to consider, such as all the people having the same bearer token, which means there's no way to do finer-grained permissions per-user or per-role or per-attribute. There are some user interface areas to consider, such as if a user/agent doesn't treat the URL as secret, because it shows content rather than masking characters such as "*".

For comparison, "security by obscurity" means there's a weakness in how the security is built, such that if you saw the source code, or the physical insides of a lock, then you would understand more about how to crack the security.

In URL pattern, an example of security by obscurity would be if the URL string was not actually random but instead was simply incrementing, or was based on a reversible function of the time or username, etc. If you read the source code, you would discover that there's guessable sequence or guessable trick, and thus become much more likely to break the security.

Edit: I strongly favor higher security, and fine-grained access control, and multi-factor authentication, and UI/UX masking, etc. This post is just to look at security by obscurity.

[jblow]

Browsers do not treat URLs as secure. If you just go to the page and happen to be live-streaming on Twitch or whatever, anyone can access the document because the information is printed visibly on the screen. This makes it starkly different from a password.

[eridius]

And if you type your password onto a keyboard and happen to be live-streaming your physical self on Twitch or whatever from an angle where people can see your hands, they know your password too.

[mevile]

I did not know this distinction, thank you, it makes sense!

[thedz]

Whatever the semantics of security by obsucrity, it's still not ideal that an unintentional screengrab or screen share could reveal the fully editable path of a doc

(i much prefer the google docs method of being able to limit to specific orgs or people)

[AmericanChopper]

It’s obscurity in the same way a password or a bearer token is obscurity. Which is not what anybody would reasonably call security by obscurity, as your definition includes any use of a secret.