#tlschannelbinding — Public Fediverse posts on home.social

Erik van Straten @[email protected] · 2025-12-04 · 12:33 UTC

@pake_preacher : I forgot the details of PAKE and SRP, but in the end the most secure client authentication requires:

1️⃣ Strong, long term, human comprehensible, *serving endpoint* authentication;
*AND*
2️⃣ TLS channel binding (enforcing known endpoints).

(Apart from those, both serving endpoint AND client MUST be trustworthy).

🚨 The -corrupt- CA/B forum breaks 1️⃣ by:
a) Advocating anonymous Domain Validated certificates, which render secure account creation IMPOSSIBLE;
b) Continuously decreasing certificate lifetime.

🚨 Furthermore, "legitimate" MitM's * break 2️⃣.

* Man in the Middle, like on-device virusscanners and firewalls that "open" TLS tunnels (both requiring installation of a dedicated root certificate) and proxies such as (definitely not limited to) Cloudflare and Fastly.

😱 Passkeys enforce NEITHER 1️⃣ NOR 2️⃣.

😱😱 Worse, because passkeys (or FIDO2 hardware keys) can be easily irretrievably "lost", servers typically provide WAY EASIER phishable authentication methods (such as "rescue codes").

@cendyne @soatok @chazh

#AitM #MitM #SecureOnlineAuthIsHARD #SecureAuthentication #OnlineAuthentication #Authentication #Impersonation #ChannelBinding #TLSchannelBinding #UTM #TLS #TLSinterception #TLSscanning #Proxy #Proxies #GoogleIsEvil #CloudflareIsEvil

#aitm #mitm #secureonlineauthishard #secureauthentication #onlineauthentication #authentication

Erik van Straten @[email protected] · 2025-12-04 · 12:33 UTC

@pake_preacher : I forgot the details of PAKE and SRP, but in the end the most secure client authentication requires:

1️⃣ Strong, long term, human comprehensible, *serving endpoint* authentication;
*AND*
2️⃣ TLS channel binding (enforcing known endpoints).

(Apart from those, both serving endpoint AND client MUST be trustworthy).

🚨 The -corrupt- CA/B forum breaks 1️⃣ by:
a) Advocating anonymous Domain Validated certificates, which render secure account creation IMPOSSIBLE;
b) Continuously decreasing certificate lifetime.

🚨 Furthermore, "legitimate" MitM's * break 2️⃣.

* Man in the Middle, like on-device virusscanners and firewalls that "open" TLS tunnels (both requiring installation of a dedicated root certificate) and proxies such as (definitely not limited to) Cloudflare and Fastly.

😱 Passkeys enforce NEITHER 1️⃣ NOR 2️⃣.

😱😱 Worse, because passkeys (or FIDO2 hardware keys) can be easily irretrievably "lost", servers typically provide WAY EASIER phishable authentication methods (such as "rescue codes").

@cendyne @soatok @chazh

#AitM #MitM #SecureOnlineAuthIsHARD #SecureAuthentication #OnlineAuthentication #Authentication #Impersonation #ChannelBinding #TLSchannelBinding #UTM #TLS #TLSinterception #TLSscanning #Proxy #Proxies #GoogleIsEvil #CloudflareIsEvil

#aitm #mitm #secureonlineauthishard #secureauthentication #onlineauthentication #authentication

Erik van Straten @[email protected] · 2025-12-04 · 12:33 UTC

@pake_preacher : I forgot the details of PAKE and SRP, but in the end the most secure client authentication requires:

1️⃣ Strong, long term, human comprehensible, *serving endpoint* authentication;
*AND*
2️⃣ TLS channel binding (enforcing known endpoints).

(Apart from those, both serving endpoint AND client MUST be trustworthy).

🚨 The -corrupt- CA/B forum breaks 1️⃣ by:
a) Advocating anonymous Domain Validated certificates, which render secure account creation IMPOSSIBLE;
b) Continuously decreasing certificate lifetime.

🚨 Furthermore, "legitimate" MitM's * break 2️⃣.

* Man in the Middle, like on-device virusscanners and firewalls that "open" TLS tunnels (both requiring installation of a dedicated root certificate) and proxies such as (definitely not limited to) Cloudflare and Fastly.

😱 Passkeys enforce NEITHER 1️⃣ NOR 2️⃣.

😱😱 Worse, because passkeys (or FIDO2 hardware keys) can be easily irretrievably "lost", servers typically provide WAY EASIER phishable authentication methods (such as "rescue codes").

@cendyne @soatok @chazh

#AitM #MitM #SecureOnlineAuthIsHARD #SecureAuthentication #OnlineAuthentication #Authentication #Impersonation #ChannelBinding #TLSchannelBinding #UTM #TLS #TLSinterception #TLSscanning #Proxy #Proxies #GoogleIsEvil #CloudflareIsEvil

#aitm #mitm #secureonlineauthishard #secureauthentication #onlineauthentication #authentication

Erik van Straten @[email protected] · 2025-12-04 · 12:33 UTC

@pake_preacher : I forgot the details of PAKE and SRP, but in the end the most secure client authentication requires:

1️⃣ Strong, long term, human comprehensible, *serving endpoint* authentication;
*AND*
2️⃣ TLS channel binding (enforcing known endpoints).

(Apart from those, both serving endpoint AND client MUST be trustworthy).

🚨 The -corrupt- CA/B forum breaks 1️⃣ by:
a) Advocating anonymous Domain Validated certificates, which render secure account creation IMPOSSIBLE;
b) Continuously decreasing certificate lifetime.

🚨 Furthermore, "legitimate" MitM's * break 2️⃣.

* Man in the Middle, like on-device virusscanners and firewalls that "open" TLS tunnels (both requiring installation of a dedicated root certificate) and proxies such as (definitely not limited to) Cloudflare and Fastly.

😱 Passkeys enforce NEITHER 1️⃣ NOR 2️⃣.

😱😱 Worse, because passkeys (or FIDO2 hardware keys) can be easily irretrievably "lost", servers typically provide WAY EASIER phishable authentication methods (such as "rescue codes").

@cendyne @soatok @chazh

#AitM #MitM #SecureOnlineAuthIsHARD #SecureAuthentication #OnlineAuthentication #Authentication #Impersonation #ChannelBinding #TLSchannelBinding #UTM #TLS #TLSinterception #TLSscanning #Proxy #Proxies #GoogleIsEvil #CloudflareIsEvil