5 ms·

Be warned that IPFS is not anonymous, and the IP addresses of the nodes serving the content are quite easily obtainable. And since the FBI already got involved

Be warned that IPFS is not anonymous, and the IP addresses of the nodes serving the content are quite easily obtainable. And since the FBI already got involved and arrested people, this is quite the risky endeavor.

Tepix 4y ago
If you run an IPFS server, you don't know what's on it, do you?
- dale_glass 4y ago
  You do. IPFS isn't Freenet, you're not contributing to some global pool of amorphous storage. It's just like running a webserver, but with content-based addressing, the possibility of having multiple machines provide the same content and much more convenient mirroring than over HTTP. If you start an IPFS node, it'll remain empty until you upload something, or pin (copy) somebody else's content. I think your node might also cache requests you make to it for some time, but that's also something that's a result of you asking for it.
- danwee 4y ago
  So, if the FBI comes to arrest me can I just get away with that? "Sorry guys, I do have an IPFS server but I had no idea it was distributing illegal content to the world. Can I go now?"
  - sdze 4y ago
    Who wants to deal with that? In the best case they confiscate all tech stuff from my home (including my employer’s laptop).
  - dale_glass 4y ago
    Some systems have tried that approach, yes! In Freenet you contribute to a global pool of storage, and end up storing and transmitting whatever the network happens to need. Freenet is intentionally obscure about it. The archive is encrypted, there are no logs, there's no way to list what your node contains, and storage is probabilistic making it hard to determine if something is there because you asked for it, or it passed through your node. Tor also tries something similar, and has premade explanations to give to authorities if they ask about your exit node. But those are specifically built to try to work with that case. Stuff happens behind your back that you have zero involvement in besides providing a piece of infrastructure for the network. In IPFS on the other hand you make a personal, intentional choice to mirror or upload something.
    - schoen 4y ago
      I want to slightly clarify the case of Tor: while many people historically thought that running an exit node on your own network (that you also actively used) might create ambiguity between your own activity and strangers' activity, this is not really considered correct or advisable. One reason is that a fairly naive local adversary monitoring your connection at only one location can still tell whether particular outbound clearnet connections do or don't correlate with inbound Tor traffic. That allows distinguishing activity by a local user from activity by a Tor user that's proxied through the exit node. Tor nodes themselves also don't host any data.
kkielhofner 4y ago
Another warning: IPFS is extremely resource intensive (from CPU to bandwidth) . I’m pretty disappointed they didn’t mention this in the article. If this gets any traction it will eat as much CPU and bandwidth you give it. Running with Docker and limiting CPU is necessary along with a bandwidth limiter/policer/shaper. Having run IPFS for a long while it also ends up in strange states and stops responding to requests. I’ve taken to using a monitoring system to GET a known CID and restart the IPFS container if it fails. My bandwidth and hardware have 100% uptime but IPFS needs restart at least a couple of times a day on average (looking at logs). In short: this isn’t practical.
- lijogdfljk 4y ago
  There's a new project to make IPFS very performant (within limitations of the protocol, ofc). They've got some really great goals, hope they succeed. TBH i love IPFS, and i wish i could get behind it and help.. but ever since they starting making Crypto i have moved away from it. I have no way of knowing if they're working on IPFS or if it's merely a means to a Coin.
  - smoldesu 4y ago
    I agree. IPFS saw the fork in the road and decided to go down the unsuccessful path - there's still time to fix the bugs that matter, but they've wasted a lot of time on infrastructure that will never materialize.
- Sin2x 4y ago
  You can just renice/set Low priority to the process for the CPU part, no need for Docker. Bandwidth limiting functionality should be built-in, not sure if it actually is.
  - kkielhofner 4y ago
    nice/renice isn’t what it used to be[0]. Bandwidth limiting is not built in[1]. [0] - https://stackoverflow.com/questions/10342470/process-niceness-priority-setting-has-no-effect-on-linux/40957947#40957947 https://stackoverflow.com/questions/10342470/process-nicenes... [1] - https://github.com/ipfs/go-ipfs/issues/3065 https://github.com/ipfs/go-ipfs/issues/3065
welshwelsh 4y ago
Filecoin might be a good solution for this. Basically pay other people to host it for you
- gnarbarian 4y ago
  filecoin is built on ipfs.
makeworld 4y ago
This is why the guide suggests to use a VPN.

Tepix 4y ago

If you run an IPFS server, you don't know what's on it, do you?

dale_glass 4y ago
You do. IPFS isn't Freenet, you're not contributing to some global pool of amorphous storage. It's just like running a webserver, but with content-based addressing, the possibility of having multiple machines provide the same content and much more convenient mirroring than over HTTP. If you start an IPFS node, it'll remain empty until you upload something, or pin (copy) somebody else's content. I think your node might also cache requests you make to it for some time, but that's also something that's a result of you asking for it.
danwee 4y ago
So, if the FBI comes to arrest me can I just get away with that? "Sorry guys, I do have an IPFS server but I had no idea it was distributing illegal content to the world. Can I go now?"
- sdze 4y ago
  Who wants to deal with that? In the best case they confiscate all tech stuff from my home (including my employer’s laptop).
- dale_glass 4y ago
  Some systems have tried that approach, yes! In Freenet you contribute to a global pool of storage, and end up storing and transmitting whatever the network happens to need. Freenet is intentionally obscure about it. The archive is encrypted, there are no logs, there's no way to list what your node contains, and storage is probabilistic making it hard to determine if something is there because you asked for it, or it passed through your node. Tor also tries something similar, and has premade explanations to give to authorities if they ask about your exit node. But those are specifically built to try to work with that case. Stuff happens behind your back that you have zero involvement in besides providing a piece of infrastructure for the network. In IPFS on the other hand you make a personal, intentional choice to mirror or upload something.
  - schoen 4y ago
    I want to slightly clarify the case of Tor: while many people historically thought that running an exit node on your own network (that you also actively used) might create ambiguity between your own activity and strangers' activity, this is not really considered correct or advisable. One reason is that a fairly naive local adversary monitoring your connection at only one location can still tell whether particular outbound clearnet connections do or don't correlate with inbound Tor traffic. That allows distinguishing activity by a local user from activity by a Tor user that's proxied through the exit node. Tor nodes themselves also don't host any data.

kkielhofner 4y ago

Another warning: IPFS is extremely resource intensive (from CPU to bandwidth) . I’m pretty disappointed they didn’t mention this in the article. If this gets any traction it will eat as much CPU and bandwidth you give it. Running with Docker and limiting CPU is necessary along with a bandwidth limiter/policer/shaper. Having run IPFS for a long while it also ends up in strange states and stops responding to requests. I’ve taken to using a monitoring system to GET a known CID and restart the IPFS container if it fails. My bandwidth and hardware have 100% uptime but IPFS needs restart at least a couple of times a day on average (looking at logs). In short: this isn’t practical.

lijogdfljk 4y ago
There's a new project to make IPFS very performant (within limitations of the protocol, ofc). They've got some really great goals, hope they succeed. TBH i love IPFS, and i wish i could get behind it and help.. but ever since they starting making Crypto i have moved away from it. I have no way of knowing if they're working on IPFS or if it's merely a means to a Coin.
- smoldesu 4y ago
  I agree. IPFS saw the fork in the road and decided to go down the unsuccessful path - there's still time to fix the bugs that matter, but they've wasted a lot of time on infrastructure that will never materialize.
Sin2x 4y ago
You can just renice/set Low priority to the process for the CPU part, no need for Docker. Bandwidth limiting functionality should be built-in, not sure if it actually is.
- kkielhofner 4y ago
  nice/renice isn’t what it used to be[0]. Bandwidth limiting is not built in[1]. [0] - https://stackoverflow.com/questions/10342470/process-niceness-priority-setting-has-no-effect-on-linux/40957947#40957947 https://stackoverflow.com/questions/10342470/process-nicenes... [1] - https://github.com/ipfs/go-ipfs/issues/3065 https://github.com/ipfs/go-ipfs/issues/3065

welshwelsh 4y ago

Filecoin might be a good solution for this. Basically pay other people to host it for you

gnarbarian 4y ago
filecoin is built on ipfs.

makeworld 4y ago

This is why the guide suggests to use a VPN.