multi-ISP metadata fragmentation layered with Tor
I’m not an expert in networking, Tor, or privacy research. I’m just an amateur who had an idea and wanted to share it with you.
The core idea is mine, but I used AI to rewrite it into a more formal paper format, so if the writing style looks too polished or “AI-ish,” that’s why. The paper is only there to organize the idea better. Excuse me for my laziness, but I really don't have the time to write it myself.
What I want is honest technical criticism.
The goal of the idea is not to “beat Tor” or claim perfect anonymity. It’s a narrower idea: making metadata analysis against one specific person harder by fragmenting what any one ISP can see, as I was annoyed by the idea of everything is going through the ISP even if it is encrypted, still annoying me.
I believe this could also reduce the Metadata analysis and Metadata fingerprint.
I described it in two levels: a cheaper/easier version using one main machine plus either one relay machine or one machine with isolated networks, multiple physical WANs, and multiple ISPs a stronger but more expensive version using multiple devices in different geographic places, each with different ISPs.
The idea is basically to divide requests/flows so that no single provider sees the full pattern. I already know the obvious objections are probably things like: traffic correlation still exists complexity may create more leaks the setup itself may become a fingerprint strong observers may still reconstruct a lot So I’m posting this to ask: where exactly is the biggest weakness? does this give any real privacy benefit at all? which threat models would it actually help against? is the complexity not worth the gain? I’d genuinely appreciate criticism from people who understand Tor, traffic analysis, metadata, and network architecture better than I do.
The file with details will be in the attached link.