How Aphotic compares
Signal, Telegram, Threema, Session, SimpleX and the peer-to-peer messengers such as Briar are all serious projects. This page looks at what each one guarantees by the way it is built, and at the few things only Aphotic does.
How we did this: we compare only what each app guarantees by the way it is built and leave its advertising claims out. Every mark in the table can be checked against public documentation, and the sources are listed underneath.
Checked in August 2026, SimpleX in September 2026, Session's servers in October 2026 and the source, audit and build rows in November 2026, against public documentation, with the sources listed under the table. Spot something out of date? Tell us and we will correct it in the open. Report an error →
Six verdicts
Signal
The reference everyone else gets measured against: open source, run by a US non-profit. It is tied to your phone number, though, and it connects over the ordinary internet, so the servers can see where you are.
→ Choose Aphotic if you want to sign up without a phone number or an email address, or want the server itself hidden behind Tor and the sender hidden on every direct text message, including the first one from a stranger.
Telegram
A messenger packed with features and enormous groups, where the company can read your ordinary chats. Only the chats it calls secret chats are encrypted end to end.
→ Choose Aphotic if you want everything encrypted end to end without switching anything on, and no sender on the envelope of a direct text message.
Threema
A mature paid messenger from Switzerland that lets you have an anonymous ID. The connection underneath is ordinary, though, so the servers still see your IP address.
→ Choose Aphotic for every message over Tor, senders hidden by default, a Monero payment that no database links to your chat identity, and a vault built for the day somebody makes you unlock your phone.
Session
The closest neighbour in spirit: no phone number, onion-routed by default, spread over many independent nodes. To get there it gave up the protection that keeps old messages safe if your keys are ever stolen.
→ Choose Aphotic if you want keys that keep moving, so a break-in today does not open yesterday and the conversation repairs itself afterwards. You also get self-deleting messages the server cannot interfere with, and a panic PIN.
SimpleX
No identifier of any kind, not even a random one. Every contact gets its own pair of one-way message queues, so there is nothing to look you up by. It is free, open source, has calls and runs on every platform. Its relays sit on the ordinary internet, and Tor is a setting you switch on yourself.
→ Choose Aphotic if you want Tor underneath everything with nothing to configure, and servers that have no ordinary web address at all. You also get communities of up to 5,000 people under one shared key, and a vault with a panic PIN built for the day somebody makes you unlock your phone.
Peer-to-peer messengers such as Briar
There is no server anywhere. Each phone runs its own Tor onion service and talks straight to the other phone, and the apps are free and open source, with a Bluetooth mesh for the days the internet is down. Nothing central exists that could leak or go offline. The price is that both phones have to be online at the same moment for a message to arrive, unless you run your own always-on mailbox device, and the onion address you connect from is your identity, so the device on the other end knows exactly who reached it.
→ Choose Aphotic if a message has to arrive while the other phone is switched off, held encrypted on our own servers in the EU for up to seven days, with sealed sender, so the envelope carries no sender for the relay to read. You also get groups under one shared key that scale to 5,000 people, where a removed member is removed by the encryption itself.
The comparison table
Identifier required
- Aphotic
- None (ID derived from your keys)
- Signal
- Phone number
- Telegram
- Phone number
- Threema
- None (random ID)
- Session
- None (derived ID)
- SimpleX
- None at all, not even a random ID. Each contact gets its own pair of one-way queues
- Peer-to-peer (Briar)
- None; your device's onion address is the identity
Transport
- Aphotic
- Messages over Tor only, hidden service
- Signal
- TLS to central servers
- Telegram
- TLS to cloud DCs
- Threema
- TLS to Swiss servers
- Session
- Onion-routed (Lokinet)
- SimpleX
- TLS to relays that you and your contact each pick. A second relay forwards what you send, so the receiving side never sees your IP address. Your own relay still does, unless you switch Tor on
- Peer-to-peer (Briar)
- A Tor onion service on each phone, device to device, plus a Bluetooth mesh when the internet is down
Offline delivery
- Aphotic
- Held encrypted on the relay for up to 7 days
- Signal
- Queued on the server until you reconnect
- Telegram
- Kept in the cloud
- Threema
- Queued on the server until delivery
- Session
- Held by the storage nodes for a limited window
- SimpleX
- Held encrypted on the relay until delivered, for up to 21 days⁸
- Peer-to-peer (Briar)
- Both phones have to be online at the same moment, unless you run your own always-on mailbox device
Jurisdiction & exposure
- Aphotic
- No ordinary web address; our own servers in the EU
- Signal
- US nonprofit; centralized clearnet servers
- Telegram
- Dubai HQ; closed-source server, distributed DCs
- Threema
- Swiss company; own data centre in Zurich
- Session
- Decentralized node swarm for messages; a central file server for attachments¹⁰
- SimpleX
- UK company running the preset relays on the ordinary internet. You can pick other relays or host your own
- Peer-to-peer (Briar)
- No servers at all, and no company running them
What the operator holds
- Aphotic
- Public keys, Community membership and unreadable data (groups live on the phones); the message queue clears after 7 days, a Community with saved history keeps longer. Shop orders with amount, transaction ID and dates, with no buyer identity. No stored signup or last-seen times
- Signal
- Phone number; registration date + last-connection date¹
- Telegram
- Phone number; ordinary chats, the messages themselves, readable by the company
- Threema
- ID creation date, last login, hash of linked phone/email if any²
- Session
- No central message server; storage nodes hold ciphertext briefly. The Session Technology Foundation runs the file server for encrypted attachments and a push server that sees the account ID of anyone using fast notifications¹⁰
- SimpleX
- No accounts or profiles. The relays hold encrypted queues and do not log IP addresses⁸
- Peer-to-peer (Briar)
- No operator. The data lives on the phones
E2EE by default
- Aphotic
- Always
- Signal
- Always
- Telegram
- Secret chats only³
- Threema
- Always
- Session
- Always
- SimpleX
- Always
- Peer-to-peer (Briar)
- Always
Sender metadata
- Aphotic
- Sealed sender by default
- Signal
- Sealed sender, partly opt-in⁴
- Telegram
- Yes, the operator sees it
- Threema
- Visible, minimal retention
- Session
- No, hidden by onion routing
- SimpleX
- No user IDs exist, and sent and received traffic share nothing a relay could match up
- Peer-to-peer (Briar)
- The device you reach sees which identity connected, and no server sits in between to see anything
Forward secrecy
- Aphotic
- Triple Ratchet (Double + PQ)
- Signal
- Triple Ratchet (Double + PQ)
- Telegram
- Secret chats only
- Threema
- Yes (Ibex)
- Session
- No, given up to allow several devices⁵
- SimpleX
- Double ratchet, with a post-quantum step in direct chats
- Peer-to-peer (Briar)
- A ratchet per message, over a per-contact root key that stays the same across reconnects
Post-quantum key agreement
- Aphotic
- Hybrid PQXDH + continuous PQ ratchet
- Signal
- Hybrid PQXDH + continuous PQ ratchet⁶
- Telegram
- No
- Threema
- No
- Session
- No
- SimpleX
- Hybrid sntrup761 on every ratchet step, in direct chats⁹
- Peer-to-peer (Briar)
- Hybrid ML-KEM-768 with X25519 when you pair
Disappearing messages
- Aphotic
- Direct messages and everyday groups, 30 s to 7 days, timer inside the encryption; none in Communities
- Signal
- Yes, incl. groups
- Telegram
- Secret chats / cloud timers
- Threema
- Limited
- Session
- Yes
- SimpleX
- Yes
- Peer-to-peer (Briar)
- Yes
Payment kept apart from chat identity
- Aphotic
- Yes. Licenses are redeemed as blind-signed tokens, so no stored record or token links the purchase to your chat identity. While users are few, timing could pair them
- Signal
- n/a (free, donations)
- Telegram
- Premium tied to account
- Threema
- The Threema Shop for Android takes Bitcoin or cash by post. App-store purchases run under your billing name
- Session
- n/a (free)
- SimpleX
- n/a (free)
- Peer-to-peer (Briar)
- n/a (free)
Business model
- Aphotic
- You buy a license with crypto. Nothing is made from your data
- Signal
- Non-profit, donations
- Telegram
- Premium + ads
- Threema
- One-time paid app
- Session
- Free, token-incentivized nodes
- SimpleX
- Free and AGPL. A company funded by investors and donations
- Peer-to-peer (Briar)
- Free and GPL, with no company in the loop
Group size
- Aphotic
- 20 per group, encrypted for each member; Communities up to 5,000
- Signal
- 1,000
- Telegram
- 200,000 (no E2EE)
- Threema
- ~256
- Session
- ~100
- SimpleX
- Your phone sends to each member. Large groups are still in development
- Peer-to-peer (Briar)
- Small groups. Your phone sends a copy to each member, every member has to be one of your own contacts, and there is no shared group key, so removing someone rests on the other members obeying it
Calls
- Aphotic
- 1:1 voice through our own relay in the EU. The audio leaves Tor, and the app asks before every call
- Signal
- Voice + video, groups
- Telegram
- Voice + video
- Threema
- Voice + video
- Session
- Limited
- SimpleX
- Voice + video
- Peer-to-peer (Briar)
- Briar has none; some forks add them
Multi-device
- Aphotic
- One device on purpose. Your identity is live on one phone at a time
- Signal
- Native linking
- Telegram
- Full cloud sync
- Threema
- Yes (leader device)
- Session
- Seed-based restore
- SimpleX
- One device holds the profile. The desktop app can drive the one on your phone
- Peer-to-peer (Briar)
- One device, because the identity is that device's onion address
Platforms
- Aphotic
- Android only
- Signal
- iOS, Android, desktop
- Telegram
- Everything + web
- Threema
- iOS, Android, desktop
- Session
- iOS, Android, desktop
- SimpleX
- iOS, Android, desktop, terminal
- Peer-to-peer (Briar)
- Android, and a desktop client
Source code
- Aphotic
- Not public. The protocol page shows the encryption functions
- Signal
- Client and server open source (AGPL-3.0); the server's spam filter is kept private¹¹
- Telegram
- Clients open source (GPL-2.0); server code not published¹²
- Threema
- Apps open source (AGPL-3.0); server code not published
- Session
- Client (GPL-3.0) and the nodes' storage server (MIT) open source
- SimpleX
- Client and server open source (AGPL-3.0)
- Peer-to-peer (Briar)
- Open source (GPL-3.0), with no server
Independent security audit
- Aphotic
- None yet
- Signal
- Independent academic analyses of the protocol (2017, 2023)¹³
- Telegram
- Independent academic analyses of the protocol (2022, 2025)¹⁴
- Threema
- Cure53 (apps 2020, desktop 2024)¹⁵
- Session
- Quarkslab, 2021¹⁶
- SimpleX
- Trail of Bits (implementation 2022, design review 2024)¹⁷
- Peer-to-peer (Briar)
- Radically Open Security (2023), Cure53 (2017)¹⁸
Reproducible Android builds
- Aphotic
- No
- Signal
- Yes, documented¹⁹
- Telegram
- Documented, and called experimental²⁰
- Threema
- Documented, and called experimental²¹
- Session
- None documented²²
- SimpleX
- Yes, for the release APKs on GitHub²³
- Peer-to-peer (Briar)
- Yes, with its own reproducer tool²⁴
| Aphotic | Signal | Telegram | Threema | Session | SimpleX | Peer-to-peer (Briar) | |
|---|---|---|---|---|---|---|---|
| Identifier required | None (ID derived from your keys) | Phone number | Phone number | None (random ID) | None (derived ID) | None at all, not even a random ID. Each contact gets its own pair of one-way queues | None; your device's onion address is the identity |
| Transport | Messages over Tor only, hidden service | TLS to central servers | TLS to cloud DCs | TLS to Swiss servers | Onion-routed (Lokinet) | TLS to relays that you and your contact each pick. A second relay forwards what you send, so the receiving side never sees your IP address. Your own relay still does, unless you switch Tor on | A Tor onion service on each phone, device to device, plus a Bluetooth mesh when the internet is down |
| Offline delivery | Held encrypted on the relay for up to 7 days | Queued on the server until you reconnect | Kept in the cloud | Queued on the server until delivery | Held by the storage nodes for a limited window | Held encrypted on the relay until delivered, for up to 21 days⁸ | Both phones have to be online at the same moment, unless you run your own always-on mailbox device |
| Jurisdiction & exposure | No ordinary web address; our own servers in the EU | US nonprofit; centralized clearnet servers | Dubai HQ; closed-source server, distributed DCs | Swiss company; own data centre in Zurich | Decentralized node swarm for messages; a central file server for attachments¹⁰ | UK company running the preset relays on the ordinary internet. You can pick other relays or host your own | No servers at all, and no company running them |
| What the operator holds | Public keys, Community membership and unreadable data (groups live on the phones); the message queue clears after 7 days, a Community with saved history keeps longer. Shop orders with amount, transaction ID and dates, with no buyer identity. No stored signup or last-seen times | Phone number; registration date + last-connection date¹ | Phone number; ordinary chats, the messages themselves, readable by the company | ID creation date, last login, hash of linked phone/email if any² | No central message server; storage nodes hold ciphertext briefly. The Session Technology Foundation runs the file server for encrypted attachments and a push server that sees the account ID of anyone using fast notifications¹⁰ | No accounts or profiles. The relays hold encrypted queues and do not log IP addresses⁸ | No operator. The data lives on the phones |
| E2EE by default | Always | Always | Secret chats only³ | Always | Always | Always | Always |
| Sender metadata | Sealed sender by default | Sealed sender, partly opt-in⁴ | Yes, the operator sees it | Visible, minimal retention | No, hidden by onion routing | No user IDs exist, and sent and received traffic share nothing a relay could match up | The device you reach sees which identity connected, and no server sits in between to see anything |
| Forward secrecy | Triple Ratchet (Double + PQ) | Triple Ratchet (Double + PQ) | Secret chats only | Yes (Ibex) | No, given up to allow several devices⁵ | Double ratchet, with a post-quantum step in direct chats | A ratchet per message, over a per-contact root key that stays the same across reconnects |
| Post-quantum key agreement | Hybrid PQXDH + continuous PQ ratchet | Hybrid PQXDH + continuous PQ ratchet⁶ | No | No | No | Hybrid sntrup761 on every ratchet step, in direct chats⁹ | Hybrid ML-KEM-768 with X25519 when you pair |
| Disappearing messages | Direct messages and everyday groups, 30 s to 7 days, timer inside the encryption; none in Communities | Yes, incl. groups | Secret chats / cloud timers | Limited | Yes | Yes | Yes |
| Payment kept apart from chat identity | Yes. Licenses are redeemed as blind-signed tokens, so no stored record or token links the purchase to your chat identity. While users are few, timing could pair them | n/a (free, donations) | Premium tied to account | The Threema Shop for Android takes Bitcoin or cash by post. App-store purchases run under your billing name | n/a (free) | n/a (free) | n/a (free) |
| Business model | You buy a license with crypto. Nothing is made from your data | Non-profit, donations | Premium + ads | One-time paid app | Free, token-incentivized nodes | Free and AGPL. A company funded by investors and donations | Free and GPL, with no company in the loop |
| Group size | 20 per group, encrypted for each member; Communities up to 5,000 | 1,000 | 200,000 (no E2EE) | ~256 | ~100 | Your phone sends to each member. Large groups are still in development | Small groups. Your phone sends a copy to each member, every member has to be one of your own contacts, and there is no shared group key, so removing someone rests on the other members obeying it |
| Calls | 1:1 voice through our own relay in the EU. The audio leaves Tor, and the app asks before every call | Voice + video, groups | Voice + video | Voice + video | Limited | Voice + video | Briar has none; some forks add them |
| Multi-device | One device on purpose. Your identity is live on one phone at a time | Native linking | Full cloud sync | Yes (leader device) | Seed-based restore | One device holds the profile. The desktop app can drive the one on your phone | One device, because the identity is that device's onion address |
| Platforms | Android only | iOS, Android, desktop | Everything + web | iOS, Android, desktop | iOS, Android, desktop | iOS, Android, desktop, terminal | Android, and a desktop client |
| Source code | Not public. The protocol page shows the encryption functions | Client and server open source (AGPL-3.0); the server's spam filter is kept private¹¹ | Clients open source (GPL-2.0); server code not published¹² | Apps open source (AGPL-3.0); server code not published | Client (GPL-3.0) and the nodes' storage server (MIT) open source | Client and server open source (AGPL-3.0) | Open source (GPL-3.0), with no server |
| Independent security audit | None yet | Independent academic analyses of the protocol (2017, 2023)¹³ | Independent academic analyses of the protocol (2022, 2025)¹⁴ | Cure53 (apps 2020, desktop 2024)¹⁵ | Quarkslab, 2021¹⁶ | Trail of Bits (implementation 2022, design review 2024)¹⁷ | Radically Open Security (2023), Cure53 (2017)¹⁸ |
| Reproducible Android builds | No | Yes, documented¹⁹ | Documented, and called experimental²⁰ | Documented, and called experimental²¹ | None documented²² | Yes, for the release APKs on GitHub²³ | Yes, with its own reproducer tool²⁴ |
- 1Signal's published record of government requests: asked about a phone number, the only account data it can produce is the date the account was registered and the date it last connected. signal.org/bigbrother/
- 2Threema's transparency report lists what it can hand over: when the ID was created, when it last logged in, and a scrambled form of any linked phone number or email address. threema.com/en/transparency-report
- 3Telegram's own FAQ: end-to-end encryption covers secret chats and calls. Ordinary chats are encrypted only between your phone and Telegram's servers, which means Telegram can read them. telegram.org/faq#q-how-are-secret-chats-different
- 4Signal's own announcement of sealed sender: on by default between saved contacts, while accepting it from strangers is a setting you have to switch on. signal.org/blog/sealed-sender/
- 5Session's whitepaper describes the deliberate move to long-lived keys, giving up the per-message protection in exchange for working on several devices and delivering while you are offline. arxiv.org/abs/2002.04609
- 6The PQXDH specification, which both Aphotic and Signal implement, is public. signal.org/docs/specifications/pqxdh/
- 7Briar's own documentation on how messages travel: phone to phone over Tor, Wi-Fi or Bluetooth, with an optional mailbox device you host yourself for delivery while a contact is offline. briarproject.org/how-it-works/
- 8SimpleX's privacy policy: no user accounts, messages removed from the preset relays on delivery or after 21 days, no IP address logging, operated by a UK company. simplex.chat/privacy/
- 9SimpleX's own announcement of its post-quantum double ratchet: Streamlined NTRU Prime (sntrup761) alongside the classical exchange, in direct chats. simplex.chat/blog/20240314-simplex-chat-v5-6-quantum-resistance-signal-double-ratchet-algorithm.html
- 10Session's FAQ: attachments go to the Session File Server, run by the Session Technology Foundation, and fast-mode notifications pass the account ID and push token to a push server the foundation runs. getsession.org/faq
- 11Signal publishes its server, except the spam-filtering part, whose interfaces are public and whose implementation is not. signal.org/blog/keeping-spam-off-signal/
- 12Telegram's FAQ on why it does not publish its server code. telegram.org/faq
- 13Independent formal analyses of the Signal protocol: Cohn-Gordon et al. (IEEE EuroS&P 2017) and Cremers et al. (USENIX Security 2023). eprint.iacr.org/2016/1013
- 14Independent analyses of Telegram's MTProto by Albrecht et al.: IEEE S&P 2022 and Eurocrypt 2025. eprint.iacr.org/2023/469
- 15Threema's list of code audits, with the Cure53 reports. threema.com/en/faq/code-audit
- 16Session's FAQ links its 2021 audit by Quarkslab. getsession.org/faq
- 17SimpleX's security page lists its reviews by Trail of Bits. simplex.chat/security/
- 18Briar's audit by Radically Open Security, carried out in 2023 and published by the Open Technology Fund. opentech.fund/wp-content/uploads/2024/06/report_briar-android.pdf
- 19Signal's instructions for reproducing its Android builds. github.com/signalapp/Signal-Android/tree/main/reproducible-builds
- 20Telegram's page on reproducible builds, which calls the feature experimental. core.telegram.org/reproducible-builds
- 21Threema's instructions for reproducing its Android builds, described as experimental. threema.com/en/open-source/reproducible-builds
- 22Session's release notes for Android describe signature and checksum checks of the release files and no way to rebuild them. github.com/session-foundation/session-android/blob/dev/RELEASE.md
- 23SimpleX's instructions for reproducing its Android release APKs. simplex.chat/reproduce/
- 24Briar's reproducer tool, which rebuilds the official APK and compares it bit for bit. code.briarproject.org/briar/briar-reproducer
Built to leave the smallest possible data trail
Every design decision above serves that goal. If it's your goal too, it takes one license key to start.