[–]▶ No.1070270>>1070273 >>1070600 [Watch Thread][Show All Posts]
Today I woke up thinking about doing a new network. I thought such thing could be fun even if only I use it to connect my computers.
The most distinctive feature would be the asynchronous packets. A node can send a packet to disconnected node, and this packet would be stored (for a limited time) by intermediate nodes until it could be delivered. Of course, the packets would be end-to-end encrypted to prevent the routing nodes from reading the content.
Now, I'm in doubt about a point. Should the network be anonymous? I think this could be achieved by identifying the destination node with a code that only this node would recognize (think of something like an RSA signature) and each packet must be distributed and stored only its claimed. But them, how can a DoS be mitigated if you cannot identify the source? Of course, you can know the where the packets came from (the IP address if the network is on top of the Internet), but a clever attack would route the packets through different nodes in the network.
So, in short, opinions? Suggestions for reading? I will be checking the FreeNet because the cache capability may be something similar to what I describe as asynchronous packets.
____________________________
Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.
▶ No.1070273>>1070276 >>1070277
>>1070270 (OP)
>end-to-end encrypted except the receiver doesn't exist when the connection is created
Shit's broke.
Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.
▶ No.1070276
>>1070273
>the receiver doesn't exist when the connection is created
The receiver exists, they just aren't connected. It's like SMTP: you still know their email address, even when they're offline
>t. not op
Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.
▶ No.1070277
>>1070273
>Diffie-Hellman key exchange
Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.
▶ No.1070279
Bitmessage has a "duplicate and store among peers/aggregators" feature, I think.
Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.
▶ No.1070600
>>1070270 (OP)
What you're thinking of is called a "store-and-forward" network.
>I think this could be achieved by identifying the destination node with a code that only this node would recognize
That's a good idea. You could encrypt the ID + some junk using the receiver's public key, but any time you make the ID field store not an uniquely mapped identifier but a value that maps many-to-one to an ID, you're increasing the size of the field compared to a bare ID which maps one-to-one.
And in my opinion the only plausible way for a decentralized network to function is that each message has the approximate physical location of the receiver stored within. If you scramble the receiver's ID, you can't rely on progressively closer to the target nodes to know with more accuracy where to route the packets, so you have to include a more accurate physical location within the packet, or spread out the routing target area which adds overload to the network.
Maybe you could make it so the closer to the target nodes have the ability to partially decode the receiver's ID in a way that adds more accuracy to the target's location data but doesn't reveal his identity outright, although I have no idea how you would do this. Maybe you could layer the encryptions (or just append the encrypted blocks of target data to one another) such that there are multiple keys which are distributed inversely proportional to the amount of info they reveal about the target.
Check out >>1062020 and >>1063733 if you haven't already
Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.
▶ No.1070619>>1071968 >>1072580
Create a network that mental cripples and pedophiles could use, and only mental cripples and pedophiles would use it.
The mental cripples and pedophiles will demand you do all the development, too.
I'm edgier than you. I'm developing a network with non-refutability.
Because it keeps all the cripples and pedophiles out.
Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.
▶ No.1071824>>1071891 >>1071965
A suggestion for OP: Have a cryptographic PK as the identities, and split the network into groups according to the PK hash. This means you can know the general topological vincinity of a recipient, even if he's not in the network. It also makes establishing paths much simpler and faster.
>DoS
Add either a proof of work or other proof of invested resource to all messages, then. And make it so that the network adjusts the minimum difficulty based on load levels. What this does is increase the cost of creating messages, so the network doesn't get clogged easily.
Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.
▶ No.1071891>>1072172
>>1071824
That doesn't reflect the actual physical topology of the network. If you mean overlaying it over the internet, sure, whatever, all the nodes have direct connection anyway.
Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.
▶ No.1071965>>1072172
>>1071824
>Add either a proof of work or other proof of invested resource to all messages, then. And make it so that the network adjusts the minimum difficulty based on load levels. What this does is increase the cost of creating messages, so the network doesn't get clogged easily.
Smart! Will use this in my distributed systems course! But then you need some (additional) consensus mechanism to agree on the load level though.
Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.
▶ No.1071968>>1071987 >>1072032
>>1070619
Pædos are useful idiots to anonymous network, the canary in the mine. If they get caught due to a technical failure, then you know the whole system may be compromised.
If you manage to build a good anonymous network, pædos'll use it you like it or not.
Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.
▶ No.1071987>>1072032
>>1071968
Hotwheels dubbed this 'The Pedo Problem'
Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.
▶ No.1072032
>>1071968
>>1071987
Excuse me but the term is ephebophile.
Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.
▶ No.1072172>>1072178
>>1071891
>All nodes have a direct connection
anyway.
Not if you have millions of nodes. You need nodes to act as relays, so that you can send a message to a PK without knowing the IP of the recipient.
>>1071965
You don't necessarily need that. You can make it a local consensus alrogithm, which is only an approximation, but should be much faster.
Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.
▶ No.1072178>>1072235 >>1072580
>>1072172
>Not if you have millions of nodes.
Yes they do lol. You not wanting to use it is another thing entirely.
>You need nodes to act as relays, so that you can send a message to a PK without knowing the IP of the recipient.
Oh, so you're talking about onion routing. Yeah, I was talking about mesh metworking separate from the Internet, another thing entirely. Onion routing is pretty much a solved problem, go look at Tor, i2p, Freenet, IPFS, DHS for peer exchange, etc. for ideas. But Tor is already pretty much as good as it gets except for a few tweaks here and there, and yeah they do it with public keys for the hidden services.
But I don't think that's what OP had in mind, given he talks about store and forward kinda stuff.
Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.
▶ No.1072235>>1072580
>>1072178
I think if OP explained what the network would be used for / what requirements it needs to meet, then we could give more specific help.
Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.
▶ No.1072580>>1072635
>>1072235
OP here. First, thanks for the discussion and suggestions.
This would be hobby project, so there aren't any big or overly specific goals. One idea I have in mind is to connect devices I have physical access (i.e.: desktop computer, laptop, phone, work computer) in a private network and use the asynchronous features to keep files, messages or a database of events synchronized. Some devices may connect through the internet and others through LAN (or even directly coping files like an Sneakernet). Of course, for this anonymization wouldn't be so important, that and the question about DoS attacks was out of interest in the topic. But, it could be an interesting feature in case some nodes function from devices I don't own and control completely.
Another idea I have related to this use case is that of "asynchronous streams". Some server node could host a stream that would be just a sequence of permanently stored packets that can be retrieved starting from any point. This could be useful for file (o message) synchronization, so a node would host (o broadcast and let other host) periodic information about the changes and subscribed nodes read the stream continuing from where then left last time.
Of course, I'm aware something like this was in use with NNTP/UUCP before the Internet and permanent connections became widely available.
For now this are just ideas, but I think I can learn something at least and maybe release some code if I advance enough.
>>1072178
About mesh networks, smartphones connecting with each other directly through NFC, Bluetooth or WiFi in densely populated areas could make an alternative network, if enough devices participate.
All the examples of onion routing are interesting study cases, though not all of them have the asynchronous feature (I think IPFS and DHS with bitorrent does) and I'am not sure they are suitable for deployment as private low-footprint networks.
>>1070619
What do you mean by "non-refutability"?
Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.
▶ No.1072635>>1072705
>>1072580
Not him, but it probably means a network where if somebody accuses you of requesting, sending, or receiving certain data -like cp- you can produce tangible proof signed by other nodes on the network that you never requested, sent, or received that data, and if you did and somebody accuses you of it and you can't produce evidence, it becomes very hard to refute the accusation (as opposed to completely anonymous networks where everybody can refute accusations with about the same level of credibility). Thus non-refutability, being unable to refute something.
Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.
▶ No.1072705
>>1072635
No, it's the other way around. You can present proof that a certain packet originated from a certain node. It's a proof of knowledge, proofs of ignorance are impossible (in most/all nontrivial cases).
Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.