Has networking using multiple “unrelated” connections to share an encrypted data stream been researched?

Question

I'm not exactly an expert, so I'm hoping you'll give me the benefit of the doubt and try very hard to make sense of this...

I've been thinking about a simple measure, which I'll call Multiple Networking, that may enable a computationally secure network. The idea is that we use multiple connection points that don't give information about each other; in other words, knowing information about one network connection won't reveal information about another network connection.

The essential framework behind this form of networking is that multiple connection points are used in some ways like a single connection. Communication is done between groups of connection points. The key element of this idea is that a third party can't gain access to all of the connection points. Then the information shared between two parties can't be accessed by a brute force computational attack, since the third party (attacker) doesn't have access to all of the information shared.

I'm wondering if techniques like this already exist. This is to help me determine my main question, "Is this a worthwhile research paper?".

A CLARIFICATION

Perhaps this will help explain what I'm thinking of. If you were to look at the message from inside a network, if would look like a message from A to B, a message from C to D, and a message from E to F. ONLY OUTSIDE OF THE NETWORK ARE THE MESSAGES COMBINED TOGETHER. So it looks like a group of messages, but it is really a single message. That way a third party has to figure out what messages are related in order to attack, eavesdrop, etc. The idea is that this can't be done from inside the network.

ADDENDUM I'd like to thank those below for trying to understand this. I've thought about this idea some more, and this is a lot like a message that is split apart and sent through multiple Tors. I guess the point I'm wondering about, though, is if there is a way to break up the message outside of the network. It seems to be a fatal flaw to have the message pieces traceable to a single source, even if they're routed through multiple Tor networks. This is still a work in progress, but thanks again for your patience.

Answer 1

The network technique of establishing single logical connections over multiple links is called inverse multiplexing. Channel bonding and VCAT are specific examples of inverse multiplexing. It's been use for a long time for various reasons.

Generally, I've seen inverse multiplexing used as a performance or operational enhancement, not as a security tool.

Answer 2

I don't think your proposal will be effective at improving security. Your approach is basically using "spread spectrum" style techniques to split a message up into multiple shares, and then send each share over a separate link. This idea has been considered before, but it doesn't offer many benefits.

Here are some of the limitations:

• In practice, the security of information systems tends to be driven primarily by the security of applications, people, and processes -- the network is secondary. Therefore, no matter how you change the network, it is only going to make a modest effect on overall security.

• Your scheme helps only with eavesdropping. But if eavesdropping is a threat, there is a better solution: encrypt the data. Encryption is more secure, more robust, and easier to deploy. And, encryption doesn't require a massive change to the network infrastructure, which is, let's face it, totally unrealistic in practice.

• Your scheme assumes there are multiple independent network links between A and B, and that an eavesdropper won't have control over all of them. But in practice that is not a valid assumption. Let's look at the number-one most common scenario today where eavesdropping is a realistic threat: open wireless networks. In that situation, A is a wireless access point, B is a laptop connected to the wireless network, and there is only one link connecting A and B: namely, the wireless network. Therefore, your proposed techniques are inapplicable in exactly the situations where defense is most needed.

In short, your scheme is (for the most part) solving the wrong problem, and is (in a broad sense) inferior to other alternatives that are already available.