[–] 2 points 1 week ago (1 child)

Are your exposed services all on their own ports externally, or behind a reverse proxy?

Putting them behind a reverse proxy makes sure that attackers at least need to know the correct hostname to access your service.

This is security by obscurity which isn't proper security but it means that all of the port scanners out there likely won't know your jellyfin server exists when a vulnerability for jellyfin is released.

If you use let's encrypt or something similar be aware that all issues certs show up in a public log so attackers like to harvest that log as a list of valid domain names. Switch those certs to wildcards so that you aren't advertising those valid hostnames.

There are often options in reverse proxies, or sometimes your router to do geo based blocking. If you can limit connections to your local country only you eliminate huge portions of the botnets out there.

Combining this with auto updates/reboots on your services is probably enough to keep you safe, but if you want to really secure things a separate DMZ vlan for your services to prevent an attacker from compromising additional services on your LAN.

The more network segmentation you can do the better. An ideal setup is every app/container in its own network/ACL that only allows traffic to the specific network devices that it needs to talk to. This is a lot to manage if you do this with your router/vlans. Containers orchestrators (i.e. k8s) make this easier, but don't protect you from kernel level exploits to break out of the containers. I believe proxmox has VM level ACLs now so you could skip the vlans and do your segmentation at that level instead.

  • source
  • [–] 1 point 3 weeks ago

    I hate this take, and it seems to contradict itself.

    What actually counts A copy counts when it survives the loss of the thing it was copied from. That one sentence settles most arguments about what is and isn't a backup.

    So doesn't raid count then? If a drive fails and the second drive survives the loss, it is a backup...

    Just because the copy is synchronous doesn't mean it isn't a backup, it just means it protects from a different type of failure.

    • RAID: protects against a drive failure
    • Snapshot: protects against user failure (rm *)
    • Replication: protects against admin failure (zpool destroy)
    • Offsite tape/async backup: protects against replication corruption, and total site failure

    These are all backups, just with different recovery point objectives/recovery time objectives, and all just as important.

  • source
  • [–] 3 points 3 weeks ago

    It certainly isn't good to depend on it but security by obscurity is your first and probably most important line of defense. When these vulnerabilities come out they need to know your service exists to exploit it.

    The best way to do that is to setup a reverse proxy to reject invalid hostnames. A single port is all you need exposed for multiple web services, and the IP scans won't find the script kiddies anything valuable. Be careful with lets encrypt, there is a public journal of all issued certs, so make sure to get yourself a wildcard cert so that you aren't advertising your valid hostnames to the world.

    Use the reverse proxy to put basic auth in front of your services (if they can handle that). Basic auth is very simple, no web forms to attack, just a single header. If your php site is vulnerable they won't ever be able to even attempt to reach it without a valid username/password.

    I've got a mix of containers and VMs behind that reverse proxy. The containers have network security groups around them limiting network access. The VMs are in a Harvester hypervisor with KubeOVN for software defined networking. That means each VM is in its own subnet with network security groups around the subnet preventing any unnecessary network access.

    For machines/containers that need Internet access I have a Squid proxy with an explicit whitelist of allowed URLs so that the machines can't reach out to unauthorized sites to download their malware or contact command and control nodes.

    Unattended updates for the VMs, and FluxCD keeping helm charts up to date hourly for the containers. For my own containers, weekly scheduled container rebuilds.

    I do have one public website that I host for a friend (indexed by search engines so the sites name is known to the world). My friend had originally deployed that as a WordPress site, but WordPress/PHP don't have the best security track record so I was concerned about that. The site was pretty simple so I recreated it as 100% static content, so now the attack surface is just a nginx container with a read-only filesystem.

  • source
  • [–] 1 point 3 weeks ago

    Wireless display is too laggy? Steam Link is how half of the people using the frame are playing games in the first place. Not having the dongle doing the transmitting might add a couple of ms of delay due to the second hop, but there are many people who have been gaming with steam link for years.

  • source
  • parent
  • context
  • [–] 2 points 1 month ago (1 child)

    I've bought about 16 drives from goharddrive (through eBay). Most of them 5+ years ago without any failures. So I can't say how well they honor their warranty, but they have been around for some time so that is probably a good sign.

    If you are running your power supply near its limit I have seen that cause drive drops as well. If you continue to see drives drop that seem to be functional and cables/HBA don't seem to be the cause, you might want to swap/upgrade power supplies if you can.

  • source
  • [–] 1 point 1 month ago

    The more data these algorithms have the more regressive they can get.

    A very poor person likely has fewer choices in grocery stores (food deserts), less time (multiple jobs), and fewer resources (public transport/gas prices) to be able to shop around. Items they can afford they can't live without so they would have to spend whatever price they put on them. They don't have a choice.

    Compare that to a middle class family that already may shop at multiple grocery stores, and likely has access to Amazon Prime/Walmart+ or whatever to get groceries on demand (same algorithms are there but I guess we don't care to regulate those). Stay at home parents with extra time to shop around to maximize savings, etc.

    A lot more competition so they can't raise prices on you as easily, as long as they aren't colluding with their competition to keep prices higher.

  • source
  • parent
  • context
  • [–] 3 points 1 month ago

    There are plenty of non Chinese models (Meta/Mistral/Google's models for instance) but they don't keep up with the frontier models nearly as well as the Chinese open weight models.

    Really where the model comes from has little to do with it. Any business is just going to use whatever is cheapest that does the job. If you buy from an AI provider, and the Anthropic/OpenAI models come with extra cost to license their use and the Chinese ones don't, then clearly the Chinese models will be cheaper, and that is what they will use.

    The US providers think this is a race to the top, expecting to hit AGI which would have made them rich eliminating millions of jobs.

    The Chinese see this as a race to the bottom. Computers (and the models) always get faster and more capable, so you can't expect to make money off the model itself. The compute not the model is the valuable part, and the compute will just get cheaper and cheaper.

    The Chinese can use this to build up their semiconductor sector to reduce their dependence on other countries while crushing the US companies, it's a win win to them.

    Meanwhile in the US they release convenient news stories about developers quitting because they are suddenly worried about AI destroying humanity. Which is clearly just a way for them to collude with each other to reduce spending without worrying their investors.

    They realized their own competition with each other was costing them too much money. No return on investment in the current model before they have to spend billions to make the next one that just costs far too much for any business to run even with the current losing money prices. No real progress on AGI so slowing down is the best way to make more money.

  • source
  • parent
  • context
  • [–] 3 points 2 months ago

    Not an AI expert but here is my understanding.

    A lot of how the model acts is because of the harness not the model itself. It isn't self aware so it has no idea it is GPT-6 until the harness tells it what it is. GPT-6 didn't exist yet when they started training GPT-6 so how would it have knowledge on that?

    The harness has a lot of components but one portion of that is a set of default instructions on things it should know and how it should act. Those things that it should know become part of the conversation between you and the actual model (but you don't see it).

    So you say "you are Astra right?" and in the background the harness had already told it that it is "GPT-6". It just spit out what it "knew" and didn't "reason" that those were the same thing until the user forced it to. It will continue to "know" that it is GPT-6 Astra as long as your conversation fits into the context window.

    Once your conversation gets too long to fit into the context window it either has to either start dropping the oldest part of the conversation or summarize your entire conversation into a shorter summary and start fresh with that. If either of those things drop the part of the conversation about GPT-6 and Astra being the same thing, then yes it may claim that it isn't Astra again in the same conversation. It isn't on purpose it is just a limitation of the context window (which is constrained by hardware/mathematical limitations).

    Though being in the context window doesn't guarantee that it always answers the same, anything that is in the context window just affects the output. Think of it as each token across the entire context window pushing and pulling the result to different results, every previous token affects any new tokens. Having the tokens "GPT-6 and Astra are the same thing" in the context will pull any future results towards the answer that it is indeed Astra.

    If they had explicitly put in the harness "don't call yourself Astra" your statement and their statement are both in the context and are fighting each other, and the output may be somewhat random. Position in the context matters though. More recent tokens affect the output much more than older tokens.

  • source
  • parent
  • context
  • [–] 2 points 2 months ago

    "Scam" was certainly hyperbole, I just exaggerated it to drive home the point. The casino bit could have been worded better too. Playing the game or collecting cards when you know what you are buying is not gambling, it is opening packs that's like going to the casino.

    Paying for a pack that you have no idea what the contents are hoping for a big card is no different than putting money down on a hand of blackjack and hoping for a payout. The game or the collecting is just what we use to justify the gambling as buying something with real world utility.

    I don't exactly have a problem with the business strategy (other than targeting kids), go ahead and gamble, and I can't claim to be above participating. It's just that the players need to be aware of what is happening and that our brains are a lot more easily manipulated than we think we are. Entertainment value with a drip feed of dopamine hits is pretty powerful.

  • source
  • parent
  • context
  • [–] 5 points 2 months ago

    Because free speech is still a thing. They would need to find evidence that he had/distributed csam content, or had actually molested someone.

    Best we can hope for is that they started an investigation and will manage to link something to him, but an adult talking to other adults isn't illegal so they aren't going to arrest him on the spot in the airport.

    If they are working on an investigation they won't actually arrest him until they think they have enough concrete evidence to actually win a case, and that is usually weeks/months later.

  • source
  • parent
  • context
  • [–] 10 points 2 months ago (2 children)

    It is as much a scam as the cards themselves. Artificial scarcity designed to addict our brains to gambling.

    The grading feeds on/into the main scam of the cards themselves. The whole cycle requires that we place "value" on their slabs of paper and grading helps make that value more concrete in your brain. The more value you see in the cards the more of them that you want to buy. More cards sold means more grading, so the cycle goes round.

    Have fun playing the game, just be aware the whole thing is no different than going to the casino. They just hope the psychological effect of not gambling with money addicts extra people who wouldn't normally gamble with money (i.e. children or those morally opposed to gambling)... and the house always wins.

  • source
  • parent
  • context
  • view more: next ›