There are literally hundreds to thousands. Many of them are horded by governments, APTs, and pen testers. I personally abused a 10 year old CVE for pen tests that was known to be used by non US government entities for a zero click code execution on opening a word doc.

Then there are things that are vulnerabilities but cannot be fixed as they are intensic to how Windows functions. Some can be hardened from the defaults but break compatibility and some cannot be fixed without a complete rewrite of how Windows and AD work. Disa stigs will give you defaults that can be hardened. Requirements for all domain users to see all GPOs, users, groups in order for AD to work is an example of something that cannot be fixed without a complete rewrite. That means an in privileged user can get a list of all users, all domain administrator, names of all computers on the domain, etc. As an attacker, that is invaluable.

Short answer, that list is to big and changes constantly. None that would be comprehensive, but disa stigs is a good place to start.

If you think anything on the Internet can ever be forgotten… Your going to have a bad time. Passwords, one of the most protected data types, are compiled from beaches into huge databases so that hackers can use them to try to log into website. There are literally dozens of not hundreds of those password databases on the public Internet to be downloaded, not to mention private or dark web collections. If passwords are not safe, what makes you think publicly available social media would be any different?

Even if somehow the whole federation agreed to purge all post every year, things like the Internet archive and Google cache of pages would retain the data.

For best practice, my personal recommendation would be to not have any service public facing besides a VPN that requires MFA. segment self hosted services into separate VLANs based on how sensitive the content is. Disallow all traffic between VLANs unless required and only allow based on port number, specific resources needed. Don’t forgot to disable outgoing Internet access unless required. Devices like Chinese made video cameras should never have an Internet connection.

My network looks something like: home vlan, work vlan, Netflix \ hulu streaming devices, cctv, wireless work, wireless home, wireless guest, iot, servers, network management. Would be way overkill for vast majority of people, but I would be hypocritical not to considering what I do and I do have a different threat profile than most.

Another thought: self hosted through VPN with MFA and nothing public facing is probably safer than cloud as long as you have cold backups.

It was meant more as a joke than criticizing hosting your own services. I personally have a VPN with MFA, and services that I host for my self that are segmented to a paranoid level (home camera system on own vlan, restricted from being able to reach any other vlan or the Internet, etc) with a deny all and explicit allows on per host and traffic type. The amount of work that went into building the network is probably overkill, and it is still susceptible to nation state and supply chain compromise but hopefully whoever gets in will curse me if they try to move around the network.

Realistically, every added service and host is added attack surface and chances for misconfiguration \ supply chain attack, but being alive is a risk too…

I’m guessing system admins and dev op is over represented here so some of our home networks may be targeted as a path into a corporate environment, but I’m guessing the chances are low. Sadly even the most secure networks are not an impossible target. The attackers are well ahead of defenders of networks. Attackers need exceptions, while defenders need everything perfect. Much harder to accomplish.

As an offensive security worker… I can’t help but read people listing out their attack surface 😂