I think SimpleX is the way to go. XMPP and Session are ultra retarded options, objectively.
Haha good point,
Although it is safe to say SimpleX is one of the safest if not most chatting platforms there is with fully open source code, full decentralization, zero metadata and full Tor support. I haven't done thorough research myself but the people that care most about privacy certainly did and they decided to use it.
Although it is safe to say SimpleX is one of the safest if not most chatting platforms there is with fully open source code, full decentralization, zero metadata and full Tor support. I haven't done thorough research myself but the people that care most about privacy certainly did and they decided to use it.
XMPP/OMEMO has the advantage of being old, battle-tested, and decentralized. You can pick a server or run your own, and it doesn’t tie you to a number or email. OMEMO is just Double Ratchet on top of XMPP, so you get forward secrecy and multi-device sync. SimpleX doesn’t fully have that yet. The downside is you’re relying on servers, and metadata like who talks to who and when is visible to the host unless you take extra steps like running your own onion-only server.
SimpleX is interesting because it’s trying to remove the whole “user ID on a server” concept by doing message queues with random addresses. That’s where the hype comes from, since it avoids persistent identifiers and spreads traffic across relays. But it’s newer and hasn’t had the same security auditing or stress testing that XMPP has. Unless you force it through Tor, your traffic isn’t anonymous. Trust in the devs is still a question mark since it hasn’t been around long enough to prove itself. The “no random numbers” criticism is also valid since proper entropy is critical for crypto.
Session looked good on paper with LokiNet routing, no IDs, and no phone number, but the dev team has basically gone quiet. Bug fixes are slow, so betting long-term on it feels risky.
If your main requirements are no phone number, no payment, and long-term survivability, XMPP over Tor with OMEMO is still the most proven option. You can spin up a Tor-only Prosody server and only use onion-federated servers to cut metadata exposure a lot (⚠️i have a guide on that here⚠️). SimpleX might evolve into something stronger, but right now XMPP has a decade of real-world stress testing and audits that SimpleX doesn’t.
SimpleX is interesting because it’s trying to remove the whole “user ID on a server” concept by doing message queues with random addresses. That’s where the hype comes from, since it avoids persistent identifiers and spreads traffic across relays. But it’s newer and hasn’t had the same security auditing or stress testing that XMPP has. Unless you force it through Tor, your traffic isn’t anonymous. Trust in the devs is still a question mark since it hasn’t been around long enough to prove itself. The “no random numbers” criticism is also valid since proper entropy is critical for crypto.
Session looked good on paper with LokiNet routing, no IDs, and no phone number, but the dev team has basically gone quiet. Bug fixes are slow, so betting long-term on it feels risky.
If your main requirements are no phone number, no payment, and long-term survivability, XMPP over Tor with OMEMO is still the most proven option. You can spin up a Tor-only Prosody server and only use onion-federated servers to cut metadata exposure a lot (⚠️i have a guide on that here⚠️). SimpleX might evolve into something stronger, but right now XMPP has a decade of real-world stress testing and audits that SimpleX doesn’t.
SimpleX is a good option if you setup the proxy to point to Tor. The protocol it uses is designed to minimize metadata and trust in the servers. You should really go to their github and read the "Threat Model" section of the whitepaper, it has a section on what the servers you use can see and do.
They remove crucial security properties of the Signal Protocol, which is a huge downgrade compared to Signal itself. I'd rather use Signal for sensitive stuff with my own phone number than use session. Read Session article at soatok dot blog if you wanna learn more about why Session is deficient.
++ On this.
To add to what DFT said, its protocol is nearly a carbon copy of libsignal. (If you want the security properties and related readings, ⚠️check⚠️ ⚠️these⚠️ ⚠️posts⚠️ ⚠️out⚠️.) The only change is that it replaced ML-KEM in Signal's PQXDH with sntrup as a hybrid with x25519. (If you want reading about that, go ⚠️here⚠️ and just keep reading. DJB tears NIST up for how badly they fucked up.
To be more blunt, Signal is arguably the best protocol because of things like Double Ratchet. Evgeny is a reputable guy. You may disagree with his political beliefs but he's very trusted in what he does (as he has worked in Haskell for years now.) I am confident in saying that SimpleX is [b]THE MOST SECURE AND ANONYMOUS MESSENGER AVAILABLE[/]. I recommend it to everyone.
To add to what DFT said, its protocol is nearly a carbon copy of libsignal. (If you want the security properties and related readings, ⚠️check⚠️ ⚠️these⚠️ ⚠️posts⚠️ ⚠️out⚠️.) The only change is that it replaced ML-KEM in Signal's PQXDH with sntrup as a hybrid with x25519. (If you want reading about that, go ⚠️here⚠️ and just keep reading. DJB tears NIST up for how badly they fucked up.
To be more blunt, Signal is arguably the best protocol because of things like Double Ratchet. Evgeny is a reputable guy. You may disagree with his political beliefs but he's very trusted in what he does (as he has worked in Haskell for years now.) I am confident in saying that SimpleX is [b]THE MOST SECURE AND ANONYMOUS MESSENGER AVAILABLE[/]. I recommend it to everyone.
I don't advise recommending OMEMO, as it's far weaker than SimpleX. (I'll get to this in a sec!)
SimpleX doesn't route through TOR, they use 2hop. This is a common misconception I hear but it's able to use TOR through a Socks proxy, but it doesn't require it by default. It's more precise to say SimpleX has ephemeral UIDs which are irrelevant to deanonymization, since I've seen people talk about the connect links.
OMEMO in its own specification ⚠️admits⚠️ to having weak forward secrecy and weak repudaition. This is in contrast to libSignal's (SimpleX's protocol is entirely based off this) strong forward secrecy and repudiation.
⚠️Noise would place OMEMO at around a 2-3 and SimpleX easily at 5.⚠️ Pair that with SimpleX's PQC and it's great.
SimpleX doesn't route through TOR, they use 2hop. This is a common misconception I hear but it's able to use TOR through a Socks proxy, but it doesn't require it by default. It's more precise to say SimpleX has ephemeral UIDs which are irrelevant to deanonymization, since I've seen people talk about the connect links.
OMEMO in its own specification ⚠️admits⚠️ to having weak forward secrecy and weak repudaition. This is in contrast to libSignal's (SimpleX's protocol is entirely based off this) strong forward secrecy and repudiation.
⚠️Noise would place OMEMO at around a 2-3 and SimpleX easily at 5.⚠️ Pair that with SimpleX's PQC and it's great.
The issues I have with OMEMO are its crypto compared to SimpleX.
SimpleX is not just good because of the "no UID hype," it's good because of the libraries behind it. libSignal is arguably the best messaging cryptography library, given its double ratchet. Where it differs from OMEMO is in the strength of such a double ratchet. I've referenced this in a few other posts but ⚠️Noise would grade SimpleX's (and for that matter, Signal's) double ratchet as a 5, with OMEMO at a 2 or a 3⚠️.
(⚠️OMEMO's spec would agree with this judgement.⚠️)
I get the critcism about the test of time regarding both the IM and the devs, but that's just not comparable as of now since our dataset is so small. I would love if you could expand upon the "no random numbers" criticism because this is the first I've heard of it. And yeah, ⚠️I definitely agree with you on the point of entropy⚠️.
Session looks good on paper if you also close your eyes and anus, especially with how it's been recently. A lack of PFS, no repudiation, a swap from Oxen (XMR PoS fork) to SESH (ETH) was just bollocks. Session rubs one the wrong way.
I get that XMPP with OMEMO and Tor is the most technically proven, but long-term, SimpleX will be better due to its provably better crypto (and its hybrid sntrup761 with x25519 mitigates the buzz about it being pure newfag-cryptography hype.) I feel like using the test of time as a spearhead to an argument is also a bit faulty and leads to some slippery slopes.
SimpleX is not just good because of the "no UID hype," it's good because of the libraries behind it. libSignal is arguably the best messaging cryptography library, given its double ratchet. Where it differs from OMEMO is in the strength of such a double ratchet. I've referenced this in a few other posts but ⚠️Noise would grade SimpleX's (and for that matter, Signal's) double ratchet as a 5, with OMEMO at a 2 or a 3⚠️.
(⚠️OMEMO's spec would agree with this judgement.⚠️)
I get the critcism about the test of time regarding both the IM and the devs, but that's just not comparable as of now since our dataset is so small. I would love if you could expand upon the "no random numbers" criticism because this is the first I've heard of it. And yeah, ⚠️I definitely agree with you on the point of entropy⚠️.
Session looks good on paper if you also close your eyes and anus, especially with how it's been recently. A lack of PFS, no repudiation, a swap from Oxen (XMR PoS fork) to SESH (ETH) was just bollocks. Session rubs one the wrong way.
I get that XMPP with OMEMO and Tor is the most technically proven, but long-term, SimpleX will be better due to its provably better crypto (and its hybrid sntrup761 with x25519 mitigates the buzz about it being pure newfag-cryptography hype.) I feel like using the test of time as a spearhead to an argument is also a bit faulty and leads to some slippery slopes.
Members-only continuation
This discussion contains more posts.
Create an account or sign in to continue reading the full conversation. 10 additional posts await inside.