I've always wondered: if you take the cause of each CVE, look at underlying cause of that & repeat, how many essential root problems would you be left with?
"Can't do math" (off-by-1 errors etc)
"Sanitize your inputs" (escape sequences, parsing etc)
"What are you referencing?" (null pointers, use-after-free etc)
and error handling. Plain error codes bubbling up a call stack, or bypassing that somehow which may violate program structure.
Compared to faffing around with certificates, Wireguard or if you're lazy, Tailscale, is a breeze to setup.
Generate the config, one per client, setup the client to only route specific RFC1918 networks and disable on well known wifi networks, and that's it. It literally works for years without any maintenance.
Tailscale is easier (presumably), but also puts a corporate entity between you and your data.
Tailscale is ridiculously easy, and headscale is a thing. It doesn't support everything corporate Tailscale does, but it covers the basics of "get me into my network from anywhere". You can run it on a $1/month VPS that has a public IPv4 address.
I don't work with computers professionally, at all. It still only took me about 2 hours to get the VPS, get Debian going on it, set up headscale, and connect to it from home. Do I still use corporate Tailscale? Yes, it has some features that aren't beautifully supported on headscale, and so for the moment I'm fine with that. But I do have my backup plan in place if I need it.
Personally I’d rather let Tailscale manage securing the public-facing service than use Headscale. Tailnet Lock seems like a good compromise. And as much as I’d like to use it, I don’t trust Tailscale SSH.
How many mTLS configs are anything but all-or-nothing access? It is so common for an app to slap mTLS on the endpoints and make the rule "anyone with a valid cert is admin on this API; anyone else gets 401." This is a step back towards a superuser/regular-user model.
Now if you're going to use mTLS as the basis for mature authn/authz, that's a different story. But I never see anyone do that.
VPNs these days are far easier to configure than mTLS (which often need proxies on both server and client side to handle the TLS terminations and needing to worry more with certificate lifecycle).
mTLS is a better option for distributed systems (service to service mesh networks for attestation of workload identities) instead of for human users.
I want to make sure every picture I take will be backed up to the NAS.
Do I turn on the VPN when I take a picture and turn it off when finished? Every time?
What if I need multiple VPN's? Living in a country that has a bad geoip reputation, it's already a struggle of switching on and off for individual websites to work.
Strongly agree, Tailscale / WG is the way go. Unless something needs to be available to randos on the Internet, in which case mTLS is still a huge headache.
mTLS is the one that requires a client certificate. right? It can be quite difficult to get protocols other than HTTPS to behave well with that, mostly because the application has to worry about it and many don't bother.
WireGuard is also very much like SSH private keys.
A basic config to set up a VPN that only adds access to services on the VPN itself is very short and simple, both for WireGuard server and for WireGuard clients.
You can run the WireGuard server from your home computer. For example on your NAS machine itself or some other machine you have at home. And open the port for WireGuard to the outside.
Also some routers can run WireGuard server on the router. For example Mikrotik routers have WireGuard server built in and ready to configure [0]. Routers running OpenWrt can too [1].
Or you can rent a small VPS and use that as your WireGuard server.
And then, any which where you run the WireGuard server, run the WireGuard client on your laptops, iOS and Android devices, and connect client apps using VPN internal IP addresses. And your NAS etc too, if the WireGuard server isn’t running on the NAS machine itself.
It really is quite quick to set up with server and all clients across devices.
The only real hurdle is that on iOS you have to turn the VPN back on manually after reboot and which is sometimes easily forgotten if you only use your self hosted services occasionally. But you quickly realize oh right I have to open the WireGuard app and turn on the VPN again whenever you try to use one of your self hosted services and the client isn’t connecting.
I think if you want mTLS to have any significance, the application itself has to support it, otherwise you're just doing a VPN again and there's no good reason to use it over Wireguard (especially because an mTLS server will respond to port scans and a Wireguard server will not)
You should design your home server with two layers of security. The actual server and apps should be secured in the same way you would a public application, and then wireguard on top. That way an exploit needs to compromise both layers at the same time. And with auto updates enabled the chance of that happening is low.
I agree. However, I also want to share photo albums with family and friends with Photo Station and/or Immich. I may also want to run Jellyfin so the same family and friends can watch my collection of... 'Linux ISOs'. Do I simply pinhole those TCP ports through my router and "hope for the best," or is there a better solution?
Run them in containers, put them behind a reverse proxy that does TLS, point fail2ban at the proxy logs. Use strong passwords. Take backups. Life is full of risk, learn to live with it.
Wait out the current wave of madness. In a year or so we will have the most secure software ever seen. We also might start to see servers with Memory Tagging to block exploits not yet found.
The oldest one I have still serving data at home is an 1813+ running DSM 7.1.1-42962 Update 9. It's inside my network and just does SMB1 and AFP for my retro computer collection and serves as a dumping ground for security camera footage. Really curious when Syno's going to finally cut this thing off from updates since it's been years since 7.2 came out and it didn't get it.
I just recently took a 415+ off the internet (it had Photo Station and Apache only exposed to the outside). I did the "resistor mod" on its JTAG connector to save its Atom from literally frying itself and it's also still working fine. As usual, the hardware far outlives the software.
I know this is not relevant to the topic at hand but tangentially related:
Does anyone know of any UGREEN UGOS alternatives? I don't mind the terminal but friends prefer a web GUI and for some mundane stuff I too have started to (begrudgingly) prefer it.
Since we’re off topic already - I do have the dxp6800 but I swapped the OS after first boot. Might be worth it to get the hardware but invest in say HexOS or similar
What did you swap it for? I don't like the direction Synology is going, so my 1621+ is going to be my last one from them (but it's easily got another decade before I think about replacing it), and I'd like to be able to recommend something that's about as easy as Synology and DSM for the technical-but-not-that-technical friends and family.
Kind of wild to say there is a critical security vulnerability but the update isn’t out yet. Honestly would have preferred they didn’t tell me until it was patched
I've always wondered: if you take the cause of each CVE, look at underlying cause of that & repeat, how many essential root problems would you be left with?
"Can't do math" (off-by-1 errors etc)
"Sanitize your inputs" (escape sequences, parsing etc)
"What are you referencing?" (null pointers, use-after-free etc)
and error handling. Plain error codes bubbling up a call stack, or bypassing that somehow which may violate program structure.
How many of these fundamental issues are there?
Boil it down enough and you get four:
1. Failure to differentiate data from instructions
2. Business logic flaws
3. Insecure operating environment
4. Incorrect use of cryptography
“In the beginning the Universe was created. This has made a lot of people very angry and been widely regarded as a bad move.”
You just can't have a NAS exposed to the internet. Put it behind a VPN or you will have it hacked.
VPNs are bad from a UX perspective.
I think the answer is mTLS.
Unfortunately, Synology apps don't support it.
Compared to faffing around with certificates, Wireguard or if you're lazy, Tailscale, is a breeze to setup.
Generate the config, one per client, setup the client to only route specific RFC1918 networks and disable on well known wifi networks, and that's it. It literally works for years without any maintenance.
Tailscale is easier (presumably), but also puts a corporate entity between you and your data.
Tailscale is ridiculously easy, and headscale is a thing. It doesn't support everything corporate Tailscale does, but it covers the basics of "get me into my network from anywhere". You can run it on a $1/month VPS that has a public IPv4 address.
I don't work with computers professionally, at all. It still only took me about 2 hours to get the VPS, get Debian going on it, set up headscale, and connect to it from home. Do I still use corporate Tailscale? Yes, it has some features that aren't beautifully supported on headscale, and so for the moment I'm fine with that. But I do have my backup plan in place if I need it.
Personally I’d rather let Tailscale manage securing the public-facing service than use Headscale. Tailnet Lock seems like a good compromise. And as much as I’d like to use it, I don’t trust Tailscale SSH.
How many mTLS configs are anything but all-or-nothing access? It is so common for an app to slap mTLS on the endpoints and make the rule "anyone with a valid cert is admin on this API; anyone else gets 401." This is a step back towards a superuser/regular-user model.
Now if you're going to use mTLS as the basis for mature authn/authz, that's a different story. But I never see anyone do that.
Applications are vulnerable for exploits.
mTLS prevents connecting to the application without valid cert.
Sure the reverse proxy handling the mtls can be exploited, but so can any VPN server as well.
VPNs these days are far easier to configure than mTLS (which often need proxies on both server and client side to handle the TLS terminations and needing to worry more with certificate lifecycle).
mTLS is a better option for distributed systems (service to service mesh networks for attestation of workload identities) instead of for human users.
How are they bad? You turn it on and it works. Don’t need to touch it. Don’t need to deal with anything but pasting key or scanning qr code
I want to make sure every picture I take will be backed up to the NAS.
Do I turn on the VPN when I take a picture and turn it off when finished? Every time?
What if I need multiple VPN's? Living in a country that has a bad geoip reputation, it's already a struggle of switching on and off for individual websites to work.
You set up Tailscale as a split tunnel so requests to your own server go through the vpn while everything else does not.
You can do split tunneling and only forward certain ips (or apps) through the vpn for your first problem.
With increased battery consumption, yes.
How does a rando tv connect to something behind a vpn like Emby or jf?
Your TV is probably on your home network. If not you could use one of those cheap travel routers to connect the hotel tv to your local
tailscale / wireguard on your router are completely transparent
Strongly agree, Tailscale / WG is the way go. Unless something needs to be available to randos on the Internet, in which case mTLS is still a huge headache.
mTLS is the one that requires a client certificate. right? It can be quite difficult to get protocols other than HTTPS to behave well with that, mostly because the application has to worry about it and many don't bother.
Yes, that's correct. It's like SSH private keys, but for HTTP
WireGuard is also very much like SSH private keys.
A basic config to set up a VPN that only adds access to services on the VPN itself is very short and simple, both for WireGuard server and for WireGuard clients.
You can run the WireGuard server from your home computer. For example on your NAS machine itself or some other machine you have at home. And open the port for WireGuard to the outside.
Also some routers can run WireGuard server on the router. For example Mikrotik routers have WireGuard server built in and ready to configure [0]. Routers running OpenWrt can too [1].
Or you can rent a small VPS and use that as your WireGuard server.
And then, any which where you run the WireGuard server, run the WireGuard client on your laptops, iOS and Android devices, and connect client apps using VPN internal IP addresses. And your NAS etc too, if the WireGuard server isn’t running on the NAS machine itself.
It really is quite quick to set up with server and all clients across devices.
The only real hurdle is that on iOS you have to turn the VPN back on manually after reboot and which is sometimes easily forgotten if you only use your self hosted services occasionally. But you quickly realize oh right I have to open the WireGuard app and turn on the VPN again whenever you try to use one of your self hosted services and the client isn’t connecting.
[0]: https://manual.mikrotik.com/docs/virtual-private-networks/wi...
[1]: https://openwrt.org/docs/guide-user/services/vpn/wireguard/b...
I think if you want mTLS to have any significance, the application itself has to support it, otherwise you're just doing a VPN again and there's no good reason to use it over Wireguard (especially because an mTLS server will respond to port scans and a Wireguard server will not)
I wouldn't expose anything from a home network except for a single Wireguard port.
Honestly even Wireguard at this point makes me nervous, seeing what happened to Nginx
You should design your home server with two layers of security. The actual server and apps should be secured in the same way you would a public application, and then wireguard on top. That way an exploit needs to compromise both layers at the same time. And with auto updates enabled the chance of that happening is low.
WireGuard locked behind a secure port knock firewall service.
A secure port knock firewall service which is only enabled for a limited time when a TXT record is dead dropped on a time-based subdomain
I agree. However, I also want to share photo albums with family and friends with Photo Station and/or Immich. I may also want to run Jellyfin so the same family and friends can watch my collection of... 'Linux ISOs'. Do I simply pinhole those TCP ports through my router and "hope for the best," or is there a better solution?
Tailscale is the easiest way to do this. It is a single point of security failure, but it has a good track record so far.
The obvious is that it’s either or this year. You can’t have both.
Run them in containers, put them behind a reverse proxy that does TLS, point fail2ban at the proxy logs. Use strong passwords. Take backups. Life is full of risk, learn to live with it.
Yes, share photos with family by sending them ones you feel like sharing over messaging app of your choice.
Let the family members have their own Linux ISO collections or let them pay streaming scammers.
I am not a big family guy as you can notice. But in my defense no one really cares about my photo albums or ISO files I like.
Wait out the current wave of madness. In a year or so we will have the most secure software ever seen. We also might start to see servers with Memory Tagging to block exploits not yet found.
Tied to this weakness:
So, they were building shell commands from some kind of user input in the OAuth API?These smaller consumer devices always have the worst security practices.
Are surprised? Direct interpolation leading to command injection has been in the owasp ten for over a decade now.
Prepare and parameterize statements you guys.
Kinda surprised that was happening on an oAuth API though
Mitigation for now is to go to Package Center -> Installed -> OAuth Service and stop it.
The oldest one I have still serving data at home is an 1813+ running DSM 7.1.1-42962 Update 9. It's inside my network and just does SMB1 and AFP for my retro computer collection and serves as a dumping ground for security camera footage. Really curious when Syno's going to finally cut this thing off from updates since it's been years since 7.2 came out and it didn't get it.
I just recently took a 415+ off the internet (it had Photo Station and Apache only exposed to the outside). I did the "resistor mod" on its JTAG connector to save its Atom from literally frying itself and it's also still working fine. As usual, the hardware far outlives the software.
I never even moved to 7, but I seem to recall a security update not super long ago.
I know this is not relevant to the topic at hand but tangentially related:
Does anyone know of any UGREEN UGOS alternatives? I don't mind the terminal but friends prefer a web GUI and for some mundane stuff I too have started to (begrudgingly) prefer it.
Since we’re off topic already - I do have the dxp6800 but I swapped the OS after first boot. Might be worth it to get the hardware but invest in say HexOS or similar
What did you swap it for? I don't like the direction Synology is going, so my 1621+ is going to be my last one from them (but it's easily got another decade before I think about replacing it), and I'd like to be able to recommend something that's about as easy as Synology and DSM for the technical-but-not-that-technical friends and family.
Kind of wild to say there is a critical security vulnerability but the update isn’t out yet. Honestly would have preferred they didn’t tell me until it was patched
Yay finally something I can be worried about! Was starting to feel left out by the recent security issues in PHP/Java/Ruby etc.
Feels like the have a new security update every week lately…
Which is probably too slow for the number of bad security practices in these things.
i'm glad claude helped me install debian on my DS918 and DS920 a few weeks ago
Can you give us some détails?
can we like ban friday from releasing patches? it is called read only for a reason
How do you see this working?
simple, you push for release on thursday or wait till monday
Except the security patches are for vulnerabilities that people have discovered.
What's your plan for avoiding having attackers find vulnerabilities on Friday, or exploiting them on Saturday/Sunday?
As soon as you successfully ban exploitation from happening on the weekends.
Don't hackers have worker unions?
no? schedule for release on thursday or monday
If you find exploitation in the wild that's not your decision anymore. It doesn't apply every time of course, but sometimes it does.