This reasoning is really flimsy. Are you suggesting a that a normal user will do better at securing their home grown system than a team of trained security engineers?

While it is true that individual end users probably comprise a large range of security expertise and practice it's also likely that fewer users would be affected per targeted attack.

But it would be much easier to run untargeted, large scale, automated attacks which would affect many more people. Just look at e.g. the world of self-hosted Wordpress installs. Somewhat competent people get hacked by scripts all the time because they can't keep up with patching their servers to respond to every new threat (even assuming they learn about said threats in time). Normal people don't stand a chance.

(Speaking of which, I just got script-hacked like this few days ago, and I think it's finally time to dump Wordpress and migrate to a static blog...)

Security assessment is all about risk + impact. The ability to compromise one thing and get access to lots of stuff is more severe than the near certainty of a single Windows XP compromise with no data.

I'd say that if it wasn't for Facebook, a lot of communication would take place over safer channels. If you doubt whether Facebook is insecure, click the "include 'only me' activity" in your activity log and imagine how much damage would happen if that leaked.

You and I remember very different pre-centralization Internets. Before Google, decentralized hand-administered email services were probably the single most popular way to bust into someone else's servers.

I'm not saying that we should go back to those times, I'm saying that we should be aiming for decentralization in the long run. It's not like the security paradigms we had back in the time are tied to decentralization. They're mostly tied to ignorance.

Decentralization is great for many things but has little effect on security. Decentralized channels are not inherently safer.

Aside on self-hosting: This reminds me a lot of the self-driving car accident debate. Self-driving cars could be thousands of times more reliable than humans, but unless they are mathematically perfect, some people feel safer driving themselves because they feel they are in control. Nevermind that Google/Facebook etc have millions of man hours invested into security.

I don't think it's so much that decentralized systems are technically safer, but that the people who own and control the centralized ones have incentives to keep them unsafe (at least from themselves) whereas if each person runs its own server, the incentives are properly aligned for security.

In essence, it's an economic argument, not a technical one.

>whereas if each person runs its own server, the incentives are properly aligned for security.

I don't see how this is true at all. Most people don't care/know about security. See Wordpress. Many companies don't take security seriously, there are many open mongodb pointing to the outside world - HackingTeam, a team chockfull of blackhat hackers, was partially done in by sloppy authentication and passwordless databases.

Today, when people run their own server, it doesn't seem the incentives are properly aligned at all - and given that even blackhat shops don't even take the time to secure their own systems properly, the economic argument falls flat.

Decentralization in the context of social interaction services does not work. As long as the concept of a social network exist, people will connect to similar platforms.

It's been tried, again, and again, and again, and it fails for the same reason all such projects fail - the only people who care about this architecture stuff are those who are geeky enough to be running their own decentralized services.

Heading off the inevitable response: Email is not a social interaction service, it is a message passing service that happens to have social uses.

What is a social interaction service, ans why can't it be federated or peer to peer?

Because not-techy people don't want to deal with the hassle of setting up and maintaining those things.

Don't put all your eggs in one basket.

...especially if it's not your basket.

With that logic you might as well stop using the internet altogether.

It's more like we shouldn't trust a third party closed source software for a task that is supposed to be done in a secure way.