- tor-relays - lists.torproject.org

Re: tor-relays Digest, Vol 182, Issue 15
by code9n＠pm.me 04 Apr '26

04 Apr '26

re. Critical Infrastructure Advisory (jmhbm(a)duck.com) vol 182 issue 15: This effects only Exit nodes? If not can someone put up a list of easy to understand and follow instructions (preferably signed off by Tor) because this is above my head, thanks, On Saturday, March 21st, 2026 at 12:01, tor-relays-request(a)lists.torproject.org <tor-relays-request(a)lists.torproject.org> wrote: > Send tor-relays mailing list submissions to > tor-relays(a)lists.torproject.org > > To subscribe or unsubscribe via email, send a message with subject or > body 'help' to > tor-relays-request(a)lists.torproject.org > > You can reach the person managing the list at > tor-relays-owner(a)lists.torproject.org > > When replying, please edit your Subject line so it is more specific > than "Re: Contents of tor-relays digest..." > > Today's Topics: > > 1. Critical Infrastructure Advisory (jmhbm(a)duck.com) > > > ---------------------------------------------------------------------- > > Message: 1 > Date: Fri, 20 Mar 2026 08:43:58 -0400 > From: jmhbm(a)duck.com > Subject: [tor-relays] Critical Infrastructure Advisory > To: Tor-relays_at_lists Torproject Org_jmhbm > <tor-relays(a)lists.torproject.org> > Message-ID: > <9A1A5111-552E-47D8-B328-8F5DA1187B78.1(a)smtp-inbound1.duck.com> > Content-Type: multipart/alternative; > boundary="=_WVkaE5dwTCT64xABP+4R4M_=" > > > Dear Tor Relay Operators, > > > I am writing to provide a formal advisory regarding an anticipated infrastructure break event scheduled for October 11, 2026 — the ICANN Root Zone KSK rollover. This event is expected to interact with potential vulnerabilities in resolver fragility, hardcoded fallback mechanisms, and encrypted DNS discovery protocols. > > This advisory is not intended as a routine cryptographic maintenance task. The KSK rollover serves as one of two critical timed “locks” within a broader restructuring of internet infrastructure layers, including DNS, Certificate Authority (CA) governance, and device attestation. The second such lock is the tightening of Android Remote Key Provisioning (RKP) on March 16, 2026, which collectively accelerates the consolidation of essential functions into hyperscaler-controlled systems, with Google positioned to benefit significantly from this shift. > > Infrastructure Break Mechanism > > The breakdown follows a predictable sequence of five stages: > > KSK-2026 and DNSSEC Fragmentation: > > Dual-signing increases root-zone response sizes beyond the approximately 1232-byte UDP limit. This can cause volunteer-run and containerized resolvers to drop fragments or fail to establish reliable TCP fallbacks, thereby disrupting DNSSEC validation. > > Fragility in Ephemeral/Containerized Resolvers: > > These types of resolvers are commonly found on exit nodes and are unable to persist the newly established trust anchor, resulting in a SERVFAIL response for valid domain names. > > Hardcoded Fallback Behavior: > > Systems default to pre-compiled public resolvers, such as those provided by Android, Chrome, and systemd-resolved, which typically point to 8.8.8.8 (Google Public DNS). > > DDR/DNR Reinforcement: > > Encrypted-DNS discovery directs clients to the same designated resolvers—predominantly dns.google. Once this failure is resolved, the change becomes permanent. > > Market Capture Dynamics: > > DNS traffic will increasingly shift to hyperscaler infrastructure. Google Public DNS is poised to capture the largest share due to its widespread integration across Android, Chrome, and various client platforms. > > Governance Context: > > This technical break occurs under documented ICANN leadership overlaps with Google. Notably, ICANN President & CEO Kurtis Lindqvist has publicly described the October 2026 rollover as “routine,” despite the inherent risks involved. Former Google Cloud Security CTO Ryan Hurst, who was instrumental in architecting Google Trust Services, was appointed as the Backup Trusted Community Representative during the actual KSK cryptographic ceremony. Additional liaison roles and board-level connections further reinforce governance structures that favor the primary beneficiaries of the resulting traffic shift. > > These structural alignments do not require proof of intent; they establish the conditions under which the break mechanism leads to durable centralization. > > Impact on Tor Exit Nodes: > > Exit nodes utilizing affected resolvers will experience SERVFAIL responses for DNSSEC-signed clearnet domains. Client-side fallback mechanisms will bypass the exit node entirely, redirecting resolution outside the Tor network. This results in degraded reachability, inconsistent performance, and a significant erosion of the decentralized model that Tor is designed to uphold. > > Operator Actions (Required Prior to October 2026): > > Verify TCP fallback capabilities and EDNS(0) support (conduct thorough testing with large DNSSEC responses). > Audit and enforce trust-anchor updates, including unbound-anchor management. > Disable or closely monitor client-side fallback behavior on your relays. > Test resolution paths during the dual-signing window. > Where feasible, consider directing DoT/DoH traffic to non-hyperscaler resolvers. > The “Privacy Web” cannot sustain another centralized chokepoint. Proactive preparation now is essential to ensure that post-rollover migration can be reversed if necessary. > > Sincerely, > > JMHBM > > Forensic Governance Analyst > > >

1 0

Does c-tor support Happy Family on bridges?
by nusenu 03 Apr '26

03 Apr '26

Hi, I'm trying to understand what the current state of Happy Family support is in c-tor. According to the spec the Happy Family design supports bridges as well: https://spec.torproject.org/proposals/321-happy-families.html#a-note-on-bri… but there is some contradicting information in https://gitlab.torproject.org/tpo/core/tor/-/issues/40935 Can you share some insights whether c-tor implements the Happy Family spec for bridges? I'm primarily asking in the context of "Does the bridge descriptor contain the family ID if the operator configured Happy Family on a bridge?" From what we see, it looks like c-tor does not support Happy Family for bridges or collector or onionoo does not support it for bridges. Example: https://onionoo.torproject.org/details?lookup=D29BAD033529C42F9E700752CF59B… thanks, nusenu

2 1

CIISS Version 3: Easier proof based on Happy Family ID
by nusenu 28 Mar '26

28 Mar '26

Hi, version 3 of the ContactInfo Information Sharing Specification is ready: https://github.com/nusenu/ContactInfo-Information-Sharing-Specification/rel… The most relevant change is the replacement of relay fingerprint based proofs that required updating of the proof when adding new relays. The new Happy Family based proof that does not require any proof update as long as the family ID remains unchanged and all your relays use the same family ID. For OrNetStats https://nusenu.github.io/OrNetStats/ I will likely stop using version 2 and replace it with version 3 within about one month so it is best to switch to version 3 within that timeframe. Here is an example ContactInfo string using the new proof: email:tor[]example.com url:https://example.com proof:uri-familyid-ed25519 ciissversion:3 more details on the new proof options: https://nusenu.github.io/ContactInfo-Information-Sharing-Specification/#pro… kind regards, nusenu -- https://nusenu.github.io

2 1

Happy Family Roadmap for ansible-relayor/CIISS/OrNetStats
by nusenu 27 Mar '26

27 Mar '26

Hi, since David mentiond that there will be the first tor 0.4.9 stable release this week and ahf also mentioned [1] ansible-relayor in the context of the MyFamily successor Happy Family I want to give you a short roadmap in the context of relayor/CIISS/OrNetStats: * first stable tor 0.4.9.x release (likely in the coming days) * after tor 0.4.9 reaches deb.torproject.org and FreeBSD packages: * ansible-relayor will require tor version >= 0.4.9 and enable Happy Family by default in addition to MyFamily * after the first tor browser release with tor 0.4.9 for all platforms: * MyFamily support will be removed from ansible-relayor * CIISS version 3 is released: MyFamily is replaced with the Happy Family system * operators will have to update their proofs * we wait for happy family support in onionoo [2] * OrNetStats will implement and verify proofs according to CIISS version 3 [1] https://lists.torproject.org/mailman3/hyperkitty/list/tor-relays@lists.torp… [2] https://gitlab.torproject.org/tpo/network-health/metrics/onionoo/-/issues/4… kind regards, nusenu -- https://nusenu.github.io

2 3

Critical Infrastructure Advisory
by jmhbm＠duck.com 20 Mar '26

20 Mar '26

Dear Tor Relay Operators, I am writing to provide a formal advisory regarding an anticipated infrastructure break event scheduled for October 11, 2026 — the ICANN Root Zone KSK rollover. This event is expected to interact with potential vulnerabilities in resolver fragility, hardcoded fallback mechanisms, and encrypted DNS discovery protocols. This advisory is not intended as a routine cryptographic maintenance task. The KSK rollover serves as one of two critical timed “locks” within a broader restructuring of internet infrastructure layers, including DNS, Certificate Authority (CA) governance, and device attestation. The second such lock is the tightening of Android Remote Key Provisioning (RKP) on March 16, 2026, which collectively accelerates the consolidation of essential functions into hyperscaler-controlled systems, with Google positioned to benefit significantly from this shift. Infrastructure Break Mechanism The breakdown follows a predictable sequence of five stages: KSK-2026 and DNSSEC Fragmentation: Dual-signing increases root-zone response sizes beyond the approximately 1232-byte UDP limit. This can cause volunteer-run and containerized resolvers to drop fragments or fail to establish reliable TCP fallbacks, thereby disrupting DNSSEC validation. Fragility in Ephemeral/Containerized Resolvers: These types of resolvers are commonly found on exit nodes and are unable to persist the newly established trust anchor, resulting in a SERVFAIL response for valid domain names. Hardcoded Fallback Behavior: Systems default to pre-compiled public resolvers, such as those provided by Android, Chrome, and systemd-resolved, which typically point to 8.8.8.8 (Google Public DNS). DDR/DNR Reinforcement: Encrypted-DNS discovery directs clients to the same designated resolvers—predominantly dns.google. Once this failure is resolved, the change becomes permanent. Market Capture Dynamics: DNS traffic will increasingly shift to hyperscaler infrastructure. Google Public DNS is poised to capture the largest share due to its widespread integration across Android, Chrome, and various client platforms. Governance Context: This technical break occurs under documented ICANN leadership overlaps with Google. Notably, ICANN President & CEO Kurtis Lindqvist has publicly described the October 2026 rollover as “routine,” despite the inherent risks involved. Former Google Cloud Security CTO Ryan Hurst, who was instrumental in architecting Google Trust Services, was appointed as the Backup Trusted Community Representative during the actual KSK cryptographic ceremony. Additional liaison roles and board-level connections further reinforce governance structures that favor the primary beneficiaries of the resulting traffic shift. These structural alignments do not require proof of intent; they establish the conditions under which the break mechanism leads to durable centralization. Impact on Tor Exit Nodes: Exit nodes utilizing affected resolvers will experience SERVFAIL responses for DNSSEC-signed clearnet domains. Client-side fallback mechanisms will bypass the exit node entirely, redirecting resolution outside the Tor network. This results in degraded reachability, inconsistent performance, and a significant erosion of the decentralized model that Tor is designed to uphold. Operator Actions (Required Prior to October 2026): Verify TCP fallback capabilities and EDNS(0) support (conduct thorough testing with large DNSSEC responses). Audit and enforce trust-anchor updates, including unbound-anchor management. Disable or closely monitor client-side fallback behavior on your relays. Test resolution paths during the dual-signing window. Where feasible, consider directing DoT/DoH traffic to non-hyperscaler resolvers. The “Privacy Web” cannot sustain another centralized chokepoint. Proactive preparation now is essential to ensure that post-rollover migration can be reversed if necessary. Sincerely, JMHBM Forensic Governance Analyst

1 0

Hetzner abuse reports
by Christian Kujau 20 Mar '26

20 Mar '26

Greetings, I'm running a Tor relay (0.4.8.21 on FreeBSD) on a small VM hosted by Hetzner and received an abuse report from them. Although this kinda looks like the topic "Hetzner Netscan False Positives" that was discussed recently[0], I have not found out who initiated the report to Hetzner and I'm also puzzled by the distinct destination addresses. And I also thought it might be good to report this publicly that these reports are still an issue for relay operators. The report is bascially: ------------------- We have indications that an attack has been conducted from your server. Netscan detected from host <my-ip-address> TIME (UTC) SRC SRC-PORT -> DST DST-PORT SIZE PROT -------------------------------------------------------------------- 2026-02-28 11:14:23 xxx 48905 -> xxx.xx.116.12 443 74 TCP 2026-02-28 11:14:24 xxx 48905 -> xxx.xx.116.13 9004 74 TCP 2026-02-28 11:14:12 xxx 23292 -> xxx.xx.116.32 9002 74 TCP [...] ------------------- In the attached report I can find ~500 entries, spanning across 5 minutes, with my address as "source" and several desination addresses that can be grouped into three entities: * 5 entries for UDP traffic to the Xerox Corporation, at least according to whois. Weird, but then again: UDP, spoofable, and I did not consider these 5 entries relevant enough to investigate further. * 5 entries for UDP traffic to 198.18.0.1 -- which is a bogon address, used for RFC 2544 and should not be routed anyway. Weird, that this would show up in their abuse report. * The remaining entries point to network addresses in a /24 network. whois points to a RIPE assignment, and querying RIPE directly for these addresses, they are all marked as "TOR EXIT". So, clearly these addresses are part of the Tor network and I fail to understand who contacted Hetzner, complaining that my relay node contacted...other Tor nodes? Or is it a bad actor, disguising as a "TOR EXIT" and then sending abuse reports to the hosting companies? Does anyone have an idea what to make of this report? Thanks, Christian. [0] https://lists.torproject.org/mailman3/hyperkitty/list/tor-relays@lists.torp… -- BOFH excuse #217: The MGs ran out of gas.

3 6

Save the date - Updates about Relay Operator Meetup in France
by Jolie Fleure 19 Mar '26

19 Mar '26

[English version below] Salut à toutes et tous, Un "tor relay operator meetup" aura lieu à l'Antenne, Severac, France (47.541451, -2.091871) le samedi 4 Juillet à 16h (l'heure du goûter). Ce meetup est organisé au sein de Camp Interhack [1], des rencontres sur 4 jours de hackerspaces francophones. On prévoit de commencer le meetup par des petites présentations de projets et nouvelles autour du réseau Tor. Si vous savez déjà que vous aurez quelque chose à présenter, faites nous signe en répondant à ce mail (joliefleure(a)riseup.net) Ça nous permettra d'avoir une vague idée du nombre de sujets qui seront abordés. On aimerait que cette première partie dure environ 1h pour garder du temps pour la seconde partie : un temps de discussion plus informel autour d'un goûter. Le Camp a une jauge, mais si vous ne venez que pour le meetup, ou pour moins d'un jour, pas besoin de vous inscrire. Enfin, si vous voulez donner une présentation plus longue et/ou à destination d'un public plus large que les seuls opérateurs de relais, n'hésitez pas à ajouter une contribution au programme du Camp [1] ! On espère vous voir en Juillet ! \(★ω★)/ [1] https://2026.camp.interhacker.space/guide.html --- Dear relay operators, As was mentioned earlier on this list, a tor relay operator meetup will be held at l'Antenne, Severac, France (47.541451, -2.091871), on Saturday 4th of July, at 4pm. This meetup will take place during the "Interhack Camp" [1], a 4-days gathering of the French hackerspace community. Note that the Interhack Camp and the relay meetup will be mostly french-speaking ! The meetup will start with small presentations of tor-related projects / news. If you already know you are coming and want to talk about something, please let us know (joliefleure(a)riseup.net) so that we can have a rough idea of the number of topics. We would like to keep the total presentation part to about 1h. The talks will be followed by another hour (or more!) of informal discussions around drinks and snacks. There is no need to register for the event if you only attend the Tor meetup or stay for less than a full day. If you would like to give a longer talk on Tor and/or targetted at a broader audience than relay operators, feel free to add it to the Interhack Camp program [1]! We hope to see you in July ! \(★ω★)/ [1] https://2026.camp.interhacker.space/guide.html Following previous message / Ce message fait suite à : https://forum.torproject.org/t/tor-relays-re-relay-operators-meetup-in-fran…

1 0

Relay fingerprint transfer
by Keifer Bly 18 Mar '26

18 Mar '26

Hey, So for the relay with the fingerprint 79E3B585803DE805CCBC00C1EF36B1E74372861D <https://metrics.torproject.org/rs.html#details/79E3B585803DE805CCBC00C1EF36…> OVH deleted the vps due to an error. I will need to start a new relay, but is it possible to configure with the original fingerprint as I have it written here? Thanks. --Keifer

8 10

Relays on NixOS
by Clara Engler 14 Mar '26

14 Mar '26

Hey, I wanted to ask whether there any relay operators that also run relays on NixOS. What is your setup? Do you use the package/module from nixpkgs? How do you run multiple relays? Do you override the nixpkgs? ... I am currently thinking on collecting some resources regarding that, any feedback/knowledge here would be very appreciated. :-) Thank You Cλara

4 3

relayor v26.0.0 is released: Happy Family design deployed by default
by nusenu 07 Mar '26

07 Mar '26

Hi, relayor v26.0.0 is released. https://github.com/nusenu/ansible-relayor/releases/tag/v26.0.0 relayor is an ansible role that helps you with running tor relays with minimal effort (automate everything). https://github.com/nusenu/ansible-relayor#main-benefits-for-a-tor-relay-ope… Changes in this release: * enable Happy Family by default (remove variable `tor_happy_family`) * MyFamily support remains enabled by default as well but will be removed in future relayor releases * drop support for tor version 0.4.8.x and require tor version 0.4.9.x or newer Ensure to upgrade your tor version to 0.4.9.x on your relays and control machine before running this relayor version for the first time. tor version 0.4.9.x packages are available for all relayor supported operating systems (Debian, Ubuntu, FreeBSD). kind regards, nusenu -- https://nusenu.github.io

1 0