Re: [tor-relays] Why MyFamily?
Hi nusenu Am Fr., 21. Feb. 2020 um 22:24 Uhr schrieb nusenu <nusenu-lists@riseup.net>:
Hi [REDACTED],
Last week i got an email with a warning that some of my relays are missing the correct MyFamily setup and that i am a risk to do end-to-end correlation attacks together with a list of all relays i operate plus one relay which uses the same name than i use but is not operated by me.
the email [REDACTED] is referring to for the interested readers: [1]
I already knew that not all of my relays have a correct MyFamily setup because as long as i am not sure if they will stay i usually dont include them in MyFamily because it is a pain to edit every torrc
Yes, manually managing MyFamily is a pain with that many relays. It is best to automate it so you don't have to worry about it no matter how long your relays might run.
It is also relevant to note that we are not talking about fresh relays (born days or weeks ago) but >6 months.
I think the relay you are talking about is the relay in Yekaterinburg. This relay is not listed as with no MyFamily in the first email but after receiving the first email i decided that i am okay with the newer hoster more than the older hoster so i included the newer one in MyFamily and excluded the older one and will probably let it expire so the relays mentioned with no MyFamily setup in the first and the ones in the second email should be slightly different. But thats not important anyway. I usually wait at least one billing period. Some of my relays are paid for three years in advance which means there might be cases where a relay will be in a state of unknown for three years till i know if they will stay.
A few days later i got a message that some of my relays will soon get rejected because i did not responded to the previous email.
A more correct version is: some relays were proposed for removal on the bad-relays@ list should there not be any reaction by the operator [2].
That is something different than informing an operator about an upcoming removal since everybody can propose removals and only dir auths can actually vote for the removal.
Understood.
[2] For the readers on this list, this is the second mentioned email:
I'm proposing the removal of the first 5 entries in the following table (end-to-end correlation risk) should there not be any reaction to this email from the operator.
A previous email from 2020-02-15 did not result in a reply so far.
+---------------------+--------+----------------------+------------------+-------------------+-----------------+---------------------------------------------------+------------------------------------------+
| first_seen | member | contact | nickname | tor_version | IP | as_name | fingerprint |
+---------------------+--------+----------------------+------------------+-------------------+-----------------+---------------------------------------------------+------------------------------------------+
| 2020-02-01 18:00:00 | 1 | NULL | angeltest33 | 0.4.2.6 | 139.99.238.17 | OVH SAS | 4BF3D299BC500C350868F078749291C766C7AA6F | | 2020-01-11 16:00:00 | 1 | NULL | angeltest5test | 0.4.2.6 | 51.38.147.96 | OVH SAS | 951307BA74E44A9C9C208B2F134CDA2409944075 | | 2019-08-06 11:00:00 | 1 | NULL | angeltest27 | 0.4.2.6 | 185.173.177.153 | GalaxyStar LLC | 95C8B9418E74F3FF80E5C3D3AF7F03156FFBBFBE | | 2019-08-31 09:00:00 | 1 | NULL | angeltest9 | 0.2.9.14 | 104.244.76.190 | FranTech Solutions | F1D5C0F5157D9B24014BE8C7A1D878AEA6843B42 | | 2019-11-21 12:00:00 | 1 | NULL | angeltest26test | 0.4.2.6 | 91.243.50.239 | Petersburg Internet Network ltd. | F51A927E34662D6005393F2327C870FB0D0D7FE0 |
- The bad-relays team expect an answer to their emails even if they do not tell you that in the first email and rather send you a second email that they will soon reject your relays if you dont answer them.
I think you are confusing the "bad-relays team" with subscribers or people sending emails to the bad-relays@ list. If in doubt require the sender to have a @torproject.org address (unfortunately there is no actual sender address for the bad-relays team since it is just a mailing list)
The senders name was "tor-bad-relays@riseup.net" which looked official to me. Then Georg Koppen "validated" the second email. For the next time i know that i only need to respond to valid emails.
So for what reason do i set the MyFamily option beside making a Hidden Service Guard discovery attack more easy?
- risk reduction for tor users MyFamily declarations allow the tor client software to automatically detect relay families when creating circuits to avoid using multiple relays from the same operator in a single circuit.
This should not matter if the operator is not malicious and like i already said an malicious operator will not use the same contact info or relay name. - reducing the risk for tor users that might become victims if some
operator gets compromized (with all its relays)
This is a reason i can understand. Not sure how much that would really help in practice but i can understand it. - transparency
Every relay operator should declare their relay group to allow everybody to measure their network fraction (Sybil detection).
Should... But i understand this one too. But as long as my family is still a small one with only one exit compared to others i am not a Sybil attack risk and even if i would would i get any special treatment then? - risk reduction for relay operators
MyFamily also provides risk reduction for operators since they are less valuable as an attack target if they can not technically be used for e2e correlation attacks
I think this is similar to your first point but i think that should be the operators choice if he want to take steps against this case. - allow the identification of "false-friends" and actual malicious relays
By setting MyFamily you make it easier to detect relays that claim to be you since MyFamily requires mutual configuration malicious entities can not add their relays to your MyFamily.
This is what happened in your case (which was a mixture of misconfiguration and actual "false-friends").
Talking about that particular possibly malicious FranTech relay the operator and me set their MyFamily correctly. I did not included it in my family because i am not the operator and he did not tried to include it in my family so relying on MyFamily was useless anyway. I knew about that relay for some while already because i first thought i had accidentially made one of my relays an exit but decided to not report it because relay names are free to choose. Thats the reason why i was so mad about the fact that someone assume this particular malicious relay is operated by me because what the other possibly malicious operator tried to achieve (to look like me) was exactly what happened. I think MyFamily greatly fails in trying to solve a problem and is making life for honest operators more difficult but not for mailicious operators.
If you are really interested into the MyFamily topic you can find a few tickets on trac.torproject.org about it (including arguments against it).
I will look into it.
[1]:
Hello,
This email wants to make you aware that you are probably putting tor users at risk by not properly setting MyFamily on your tor relays.
If the relays using your contactInfo are not actually yours please send an email to bad-relays@lists.torproject.org so they can be removed from the network for impersonating your contactInfo.
https://nusenu.github.io/OrNetStats/endtoend-correlation-groups#torrpi1405gm...
Thanks for taking care of this.
+---------------------+------+------------------+--------+----------------------+-------------------+-----------------+------------------------------------------+
| first_seen | exit | nickname | member | contact | tor_version | IP | fingerprint |
+---------------------+------+------------------+--------+----------------------+-------------------+-----------------+------------------------------------------+
| 2018-08-28 21:00:00 | 0 | angeltest2 | 27 | [REDACTED] | 0.4.4.0-alpha-dev | 5.39.60.243 | 3B07C500AC17E7B5A1EE616613E104A094AB87F3 | | 2018-09-05 17:00:00 | 0 | angeltest7 | 27 | [REDACTED] | 0.4.2.6 | 37.252.187.111 | EE4AF632058F0734C1426B1AD689F47445CA2056 | | 2018-09-05 20:00:00 | 0 | angeltest8 | 27 | [REDACTED] | 0.4.2.6 | 185.112.82.50 | 7AAF5597B18D82CC90CA95FB7976A1CEA4A32E06 | | 2018-09-07 23:00:00 | 0 | angeltest9 | 27 | [REDACTED] | 0.4.2.6 | 92.38.163.21 | 9288B75B5FF8861EFF32A6BE8825CC38A4F9F8C2 | | 2018-09-27 00:00:00 | 0 | angeltest11 | 27 | [REDACTED] | 0.4.2.6 | 213.183.60.21 | 39F91959416763AFD34DBEEC05474411B964B2DC | | 2018-09-27 19:00:00 | 0 | angeltest12 | 27 | [REDACTED] | 0.4.2.6 | 91.201.65.91 | 57C6DF5B93E54EB9C8DB90029D9E9A1111BD34D2 | | 2018-12-15 00:00:00 | 0 | angeltest14 | 27 | [REDACTED] | 0.4.2.6 | 195.123.245.141 | 465D17C6FC297E3857B5C6F152006A1E212944EA | | 2019-01-10 00:00:00 | 0 | angeltest18 | 27 | [REDACTED] | 0.4.2.6 | 94.140.125.122 | B517198B86B3859C307857C59F6660A281FC8B47 | | 2019-01-10 22:00:00 | 0 | angeltest19 | 27 | [REDACTED] | 0.4.2.6 | 185.246.152.22 | A86EC24F5B8B964F67AC7C27CE92842025983274 | | 2019-02-26 10:00:00 | 0 | angeltest20 | 27 | [REDACTED] | 0.4.2.6 | 178.17.170.103 | ADE6AB2BFBD7A5780B321DC33BBACCD0D777C94D | | 2019-04-26 12:00:00 | 0 | angeltest23 | 27 | [REDACTED] | 0.4.2.6 | 81.169.235.154 | EFA2E7B073AA4CE2DAF7160F23C90DB805948F4A | | 2019-05-29 07:00:00 | 0 | angeltest26 | 27 | [REDACTED] | 0.4.2.6 | 89.223.100.121 | 40108FDFA40EDB013F7291F3B4DA3D412ED3A5EF | | 2019-08-06 11:00:00 | 0 | angeltest27 | 27 | [REDACTED] | 0.4.2.6 | 185.173.177.153 | 95C8B9418E74F3FF80E5C3D3AF7F03156FFBBFBE | | 2019-08-13 12:00:00 | 0 | angeltest6 | 27 | [REDACTED] | 0.4.2.6 | 185.61.149.67 | 295F1BD8995A12ECC77E050CCF6EC641572739E9 | | 2019-08-19 04:00:00 | 0 | angeltest28 | 27 | [REDACTED] | 0.4.2.6 | 31.207.89.49 | 1A7A2516A961F2838F7F94786A8811BE82F9CFFE | | 2019-09-19 14:00:00 | 0 | angeltest3 | 27 | [REDACTED] | 0.4.2.6 | 62.141.38.69 | FF9FC6D130FA26AE3AE8B23688691DC419F0F22E | | 2019-10-02 19:00:00 | 0 | angeltest10 | 27 | [REDACTED] | 0.4.2.6 | 178.254.20.159 | C1939D36649DE98A202429631D8EFC70128D5F5F | | 2019-11-01 05:00:00 | 0 | angeltest5 | 27 | [REDACTED] | 0.4.2.6 | 51.38.134.104 | 39C6F833D4B09524770D3655DF825A11213CA0A9 | | 2019-11-01 08:00:00 | 0 | angeltest27test | 1 | [REDACTED] | 0.4.2.6 | 92.223.109.71 | 1323D34C2FA4AE0EC4EEA9853F3464693EF428E7 | | 2019-11-02 14:00:00 | 0 | angeltest17 | 27 | [REDACTED] | 0.4.2.6 | 5.34.183.29 | F8AA8D8CCBA0C5F2836DE6315CDFA6E4A31A0890 | | 2019-11-05 14:00:00 | 0 | angeltest13 | 27 | [REDACTED] | 0.4.2.6 | 185.225.17.173 | A4CC39184AD287D72C2247738835811C7A7ECB8E | | 2019-11-21 12:00:00 | 0 | angeltest26test | 1 | [REDACTED] | 0.4.2.6 | 91.243.50.239 | F51A927E34662D6005393F2327C870FB0D0D7FE0 | | 2019-12-11 22:00:00 | 0 | angeltest29 | 27 | [REDACTED] | 0.4.2.6 | 87.106.152.102 | 73283C4DEBC01D3E4A5FD1BB1F2B50D927379F59 | | 2019-12-20 04:00:00 | 0 | angeltest24 | 27 | [REDACTED] | 0.4.2.6 | 2.56.241.243 | 401A66747713038CEEF6ED28C8AFEB70570EEBCC | | 2019-12-20 06:00:00 | 0 | angeltest25 | 27 | [REDACTED] | 0.4.2.6 | 185.118.164.41 | C9BC841E180B35F229FD47664F84CF8A8ADB3F68 | | 2019-12-24 16:00:00 | 0 | angeltest30 | 27 | [REDACTED] | 0.4.2.6 | 185.4.135.157 | B93503D458D9FE97DE5C12D211082871D08F1284 | | 2020-01-11 16:00:00 | 0 | angeltest5test | 1 | [REDACTED] | 0.4.2.6 | 51.38.147.96 | 951307BA74E44A9C9C208B2F134CDA2409944075 | | 2020-01-12 17:00:00 | 0 | angeltest31 | 27 | [REDACTED] | 0.4.2.6 | 185.101.35.219 | 2F8B9500DC98C13FD28CC51E47D3416DE423ED78 | | 2020-01-14 15:00:00 | 0 | angeltest32 | 27 | [REDACTED] | 0.4.2.6 | 185.99.2.178 | 12B1A5769D38FF47CF68C2235E1BDA315DF400F2 | | 2020-01-23 04:00:00 | 0 | angeltest16 | 27 | [REDACTED] | 0.4.4.0-alpha-dev | 195.123.238.164 | D5812BAB52820A4D448E5F16EE363A0F4CEEF691 | | 2020-02-01 18:00:00 | 0 | angeltest33 | 1 | [REDACTED] | 0.4.2.6 | 139.99.238.17 | 4BF3D299BC500C350868F078749291C766C7AA6F | | 2020-02-11 14:00:00 | 1 | angeltestwindows | 1 | [REDACTED] | 0.4.2.5 | 91.132.147.168 | DC81AA3B1D51566DBF27BFA562E4047AEB1C52DA |
+---------------------+------+------------------+--------+----------------------+-------------------+-----------------+------------------------------------------+
_______________________________________________ tor-relays mailing list tor-relays@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays
- risk reduction for tor users MyFamily declarations allow the tor client software to automatically detect relay families when creating circuits to avoid using multiple relays from the same operator in a single circuit.
This should not matter if the operator is not malicious
That is a big if and impossible to detect automatically. If we accept operators to run end-to-end correlation relay groups by receiving "you can trust me" emails you can guess what malicious actors will do next. The only way the tor client software can detect relay groups across multiple /16 blocks automatically and at scale is currently by MyFamily declaration. There is no "dude don't worry, you can trust me" flag.
and like i already said an malicious operator will not use the same contact info or relay name.
We've had that already.
But as long as my family is still a small
It is rather hard, time consuming and error prone to asses group sizes without proper MyFamily declarations.
I think MyFamily greatly fails in trying to solve a problem
I agree, but it is currently the only option how operators can tell tor clients about their relay group in an automated way. To summarize: Multiple recommendations (with and without configuration management) have been pointed out to practically solve the hassle of MyFamily across multiple relays with a growing group of relays without requiring to mess with all torrc files manually whenever a new relay gets added to a group. Using one of them should be in the interest of relay operators to help protect tor users (and indirectly help with malicious relay detection).
Am Sa., 22. Feb. 2020 um 15:17 Uhr schrieb nusenu <nusenu-lists@riseup.net>:
- risk reduction for tor users MyFamily declarations allow the tor client software to automatically detect relay families when creating circuits to avoid using multiple relays from the same operator in a single circuit.
This should not matter if the operator is not malicious
That is a big if and impossible to detect automatically. If we accept operators to run end-to-end correlation relay groups by receiving "you can trust me" emails you can guess what malicious actors will do next.
Of course would they do.
The only way the tor client software can detect relay groups across multiple /16 blocks automatically and at scale is currently by MyFamily declaration. There is no "dude don't worry, you can trust me" flag.
And if there would be then this would be the worst possible solution.
and like i already said an malicious operator will not use the same contact info or relay name.
We've had that already.
I know. Thats why i point that out again because now i am somehow affected too and can better understand what they mean with that sentence.
But as long as my family is still a small
It is rather hard, time consuming and error prone to asses group sizes without proper MyFamily declarations.
I am the operator of my relays so if i for whatever reason decide to not publish that i run a bigger family then this should be my own decision. If the torproject needs these information urgently they need to force it for example with a relay registration or should find a better soultion which is not depending on a trust level.
I think MyFamily greatly fails in trying to solve a problem
I agree, but it is currently the only option how operators can tell tor clients about their relay group in an automated way.
To summarize:
Multiple recommendations (with and without configuration management) have been pointed out to practically solve the hassle of MyFamily across multiple relays with a growing group of relays without requiring to mess with all torrc files manually whenever a new relay gets added to a group.
Understood.
Using one of them should be in the interest of relay operators to help protect tor users (and indirectly help with malicious relay detection).
Not proposing relays of honest operators for removal should be in the interest of all to help protect tor users but an opt-in solution for MyFamily which gets forced by random people on a public tor-bad-relays mailinglist is not the right way in my opinion because obviously at least in my case these people might lack information. I understand that this is only obvious for me but then these people should think twice before they propose relays for removal.
_______________________________________________ tor-relays mailing list tor-relays@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays
[REDACTED]:
But as long as my family is still a small It is rather hard, time consuming and error prone to asses group sizes without proper MyFamily declarations.
I am the operator of my relays so if i for whatever reason decide to not publish that i run a bigger family then this should be my own decision.
There are two notions to this, depending on what you mean by 'publish'. 'publish' in the sense of linkability relays <-> operator identity: Correct there is no need for that. 'publish' in the sense of declaring a proper MyFamily set: from the tor manual page: "If you run more than one relay, the MyFamily option on each relay **must** list all other relays"
If the torproject needs these information urgently they need to force it
The Torproject Inc does not run the Tor network, nor the majority of Tor directory authorities, but yes, some Torproject employees play a key role on what gets actually enforced on the network and what not and The Torproject produces the software that dir auths run so they have at least partial/indirect control over the imposed rules and the network. As far as I know there is no formal or informal written agreement between Tor directory authorities as to how they run the network. Past attempts by a Torproject employee an me, to establish something like that unfortunately failed [1].
Not proposing relays of honest operators for removal should be in the interest of all to help protect tor users
It is hard to tell honest operators from others if the relay has no ContactInfo or does not reply to emails. Even if they reply it can be non-trivial. So if there were actual technical rules they should apply to all relays equally and not just to dishonest operators because how do you define and measure "honest" operator? Should an operator who confirms to bad-relays@ that he setup modified relays to collect onion addresses be allowed on the network because he is at least honest about it?
but an opt-in solution for MyFamily which gets forced by random people on a public [tor-]bad-relays mailinglist
bad-relays@ is not public in the sense that everyone can read it, but everyone can send to it, which is its main purpose. [1] https://medium.com/@nusenu/the-growing-problem-of-malicious-relays-on-the-to...
Am Sa., 22. Feb. 2020 um 17:11 Uhr schrieb nusenu <nusenu-lists@riseup.net>:
[REDACTED]:
But as long as my family is still a small It is rather hard, time consuming and error prone to asses group sizes without proper MyFamily declarations.
I am the operator of my relays so if i for whatever reason decide to not publish that i run a bigger family then this should be my own decision.
There are two notions to this, depending on what you mean by 'publish'.
'publish' in the sense of linkability relays <-> operator identity: Correct there is no need for that.
'publish' in the sense of declaring a proper MyFamily set:
from the tor manual page: "If you run more than one relay, the MyFamily option on each relay **must** list all other relays"
I will list them or shut them down. Just not right now. I thought about why i do not want to list them right now and this reason might sound stupid to others but for me including a new relay into MyFamily is some sort of "celebration". When i include a new relay i commit myself to care for it for the next time period (no matter how long that means). For me that means checking nyx and the logfile everyday and taking a quick look into nmon. So as an operator who paies the bills for my new children i expect the torproject and all affected people to wait till i did my "celebration" or take the necessary steps and reject them so that i understand this as a message that the celebration took too long and that these relays are not wanted anymore. I think i do not want to automate this because it would destroy my celebration. I already automated upgrades because i see a purpose in this but from my point of view i can not see any porpose to automate MyFamily. I also thought about why i include them in MyFamily at all and i think the reason is because i want that others have the possibility to exclude my relays if they see a need to do so.
If the torproject needs these information urgently they need to force it
The Torproject Inc does not run the Tor network, nor the majority of Tor directory authorities, but yes, some Torproject employees play a key role on what gets actually enforced on the network and what not and The Torproject produces the software that dir auths run so they have at least partial/indirect control over the imposed rules and the network. As far as I know there is no formal or informal written agreement between Tor directory authorities as to how they run the network. Past attempts by a Torproject employee an me, to establish something like that unfortunately failed [1].
I think i remember something where nick explained that he (or any other torproject member listed on the torprojects peoples page) can not directly tell the authorities operators what they should do. This made me think about for the first time what the torproject actually is. And i think that the way it is right now is actually a good thing. Maybe it even must be like that to ensure free speech as good as possible and to not make some people a big target. Bu its funny to hear from one of the main designers of Tor that he can actually not really decide what happens. I think there is the difference between a normal company and Tor and thats the reason why i am okay paying bills without getting something countable back (beside the fact that i learned a lot in many ways since i started my first relay). I think it is impressive how good this project works and i think you would put that at risk if you try to force standards too easily and telling directory authorities operators how to run them is one of these examples. Of course i only see what i can see and i see that you are more involved than i am but i think as long as its not broken dont try to fix it.
Not proposing relays of honest operators for removal should be in the interest of all to help protect tor users
It is hard to tell honest operators from others if the relay has no ContactInfo or does not reply to emails. Even if they reply it can be non-trivial. So if there were actual technical rules they should apply to all relays equally and not just to dishonest operators because how do you define and measure "honest" operator? Should an operator who confirms to bad-relays@ that he setup modified relays to collect onion addresses be allowed on the network because he is at least honest about it?
(Just for the record i had contact info on all my relays and check that email address weekly. Since my first exit even daily.) Yes of course this is a problem and the only "solution" is to raise the bar for all. But this is not what MyFamily is doing. It is raising the bar for the honest ones but not for the dishonest ones. If this problem need to get fixed i propose using a contact email address as a replacement of MyFamily and sending a validation email. Maybe even send another validation email after some randomized timeframe from time to time to validate that the operator is still caring and reachable. This way you at least raise the bar for all and most honest operators will just use one email address for it so automatically using this one as "the family" is easy. If a malicious entity wants to put for example 20 relays on the network with different contact email addresses this is still manageable but 200 of them is a whole other thing. Sending validation emails is generally stupid but if this would solve a problem which needs a fix then why not? People are used to it because whatever they do on the internet they get a validation email and the torproject already uses this for its mailing lists. The idea with somehow using the comtact info was raised here already but never really discussed: https://trac.torproject.org/projects/tor/ticket/6676 About the question if someone should be allowed to for example collect onion addresses with a modified tor client/relay: Actually in the first moment i would say yes. Where is the difference between a crazy user writing down every onion he can find an an automatic system doing this? If you are talking about HSdirs collecting the hidden services they got i would say "better not" because this is not what they were made for and i can not see a benefit for the network. But this should be technically solved and not by blocking them and as far as i know v3 onion addresses are solving that problem so an hidden service operator can now choose if they want to take that risk or they could use client authentication right away. Another reason to allow it would be because if you dont allow it but it is still possible and someone do this and you try to solve that by blocking them that means you are constantly failing in what you do because if they want to do it then they will just come back. But if you say "Yes it is allowed! (... if you are able to do it) then you have a competition and an overall improvement or at least a bigger viewpoint about what to fix, what to allow and where to aggree.
but an opt-in solution for MyFamily which gets forced by random people on a public [tor-]bad-relays mailinglist
bad-relays@ is not public in the sense that everyone can read it, but everyone can send to it, which is its main purpose.
[1] https://medium.com/@nusenu/the-growing-problem-of-malicious-relays-on-the-to...
_______________________________________________ tor-relays mailing list tor-relays@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays
On 22.02.20 15:51, [REDACTED] wrote:
I am the operator of my relays so if i for whatever reason decide to not publish that i run a bigger family then this should be my own decision.> If the torproject needs these information urgently they need to force it for example with a relay registration or should find a better soultion which is not depending on a trust level.
I am sorry, but this is an ignorant perspective. Even though the Tor network has no means to force it on to you, you really should configure your nodes correctly. This includes a correct MyFamily statement, even if it means more work. If you don't want to do that work, then you should ask yourself why you contribute relays in the first place. Do you really want to do it to weaken the network? Probably not. It is really not that much effort to synchronize the statement, even with a large number of relays and without willingness to work with "configuration management" tools. It took me only a few minutes to put together a bash script that logs in, grabs fingerprints, assembles them to a unified MyFamily statement, and pushes the updated line to all relays again. [1] On a more general level, do you really want to argue than any rule or law that is not enforceable is completely pointless in society? You seem to think MyFamily is not that relevant because its correct configuration relies on the same operator that you need to trust not to perform end-to-end correlation in the first place. This is only a minor aspect. As an operator, you and your infrastructure becomes a potential target. By not configuring MyFamily correctly, you invite attackers, and make their lives easier. I can pown you, steal your keys, exploit a weakness in your configuration, get a court to give me a wiretapping order for a single individual much easier than for many, etc etc, all much more interesting if I _know_ that you are a careless operator that does not configure their relays correctly. You should make your relays less interesting, also for others, not only for yourself. Cheers, and thanks for trying to run relays in a good fashion :) Moritz [1] https://github.com/torservers/myfamilyupdater
Am So., 23. Feb. 2020 um 11:51 Uhr schrieb Moritz Bartl < moritz@torservers.net>:
On 22.02.20 15:51, [REDACTED] wrote:
I am the operator of my relays so if i for whatever reason decide to not publish that i run a bigger family then this should be my own decision.> If the torproject needs these information urgently they need to force it for example with a relay registration or should find a better soultion which is not depending on a trust level.
I am sorry, but this is an ignorant perspective. Even though the Tor network has no means to force it on to you, you really should configure your nodes correctly. This includes a correct MyFamily statement, even if it means more work. If you don't want to do that work, then you should ask yourself why you contribute relays in the first place. Do you really want to do it to weaken the network? Probably not. It is really not that much effort to synchronize the statement, even with a large number of relays and without willingness to work with "configuration management" tools. It took me only a few minutes to put together a bash script that logs in, grabs fingerprints, assembles them to a unified MyFamily statement, and pushes the updated line to all relays again. [1]
Not going with the stream is an ignorant perspective most of the time. The reason why i run relays is because in my opinion tor is doing exactly that. You want my IP address? NO! We rather build a big non-profit organization, find developers, search donations, encourage people all over the world to run relays, resist against all governmental censorship tries and do everything we can because we believe our IP address is ours. This is ignorance at its finest and thats one of the reasons why i run relays.
On a more general level, do you really want to argue than any rule or law that is not enforceable is completely pointless in society?
No, no that was not what i meant. I just didnt understood why i should set MyFamily and brought up my personal points against it so that hopefully someone can explain me why other points are more important than mine. teor explained me that with words i understood so for the future i will set MyFamily correctly now.
You seem to think MyFamily is not that relevant because its correct configuration relies on the same operator that you need to trust not to perform end-to-end correlation in the first place. This is only a minor aspect. As an operator, you and your infrastructure becomes a potential target. By not configuring MyFamily correctly, you invite attackers, and make their lives easier. I can pown you, steal your keys, exploit a weakness in your configuration, get a court to give me a wiretapping order for a single individual much easier than for many, etc etc, all much more interesting if I _know_ that you are a careless operator that does not configure their relays correctly. You should make your relays less interesting, also for others, not only for yourself.
Cheers, and thanks for trying to run relays in a good fashion :)
If my words sounded ignorant or rude or egoistic that was not my intention. I just wanted to understand why i should waste energy to do steps which i dont understand. Now i understand them and i will go with the stream and set MyFamily correctly today. Thank you all for that interesting conversation
I just found out that i can have more than one MyFamily line specified in the torrc. nusenu could you please check with your tool that everything is correct now? Greatz Michael
Hi, I've gone a few emails back up the thread, because the risk analysis is missing some really important factors. And just some reminders: Some users depend on the tor network for their safety. Relay operators take some risks, but we do our best to reduce those risks. MyFamily is about user and operator safety. We pay more attention to arguments based on safety.
On 22 Feb 2020, at 23:02, [REDACTED] wrote:
So for what reason do i set the MyFamily option beside making a Hidden Service Guard discovery attack more easy?
- risk reduction for tor users MyFamily declarations allow the tor client software to automatically detect relay families when creating circuits to avoid using multiple relays from the same operator in a single circuit.
This should not matter if the operator is not malicious and like i already said an malicious operator will not use the same contact info or relay name.
- reducing the risk for tor users that might become victims if some operator gets compromized (with all its relays)
This is a reason i can understand. Not sure how much that would really help in practice but i can understand it.
In practice, relay operators become targets for compromise when they don't set MyFamily. Because those relays can be used to attack a Tor users. If relay operators correctly set MyFamily, then an attacker needs to compromise multiple operators to see a single user's traffic. In this case, it doesn't matter if the operator is malicious.
- transparency Every relay operator should declare their relay group to allow everybody to measure their network fraction (Sybil detection).
Should... But i understand this one too. But as long as my family is still a small one with only one exit compared to others i am not a Sybil attack risk and even if i would would i get any special treatment then?
It doesn't matter how small your relays are. Some clients will choose your relays as guards. You're putting those users in danger.
- risk reduction for relay operators MyFamily also provides risk reduction for operators since they are less valuable as an attack target if they can not technically be used for e2e correlation attacks
I think this is similar to your first point but i think that should be the operators choice if he want to take steps against this case.
There's also a network effect here. If almost all operators set MyFamily, then the Tor Network becomes a less valuable target for attacks. So attackers use other methods, like attacking Tor Browser, or offline attacks. But if a lot of operators don't set MyFamily, then attackers develop tools and techniques to attack the network. Then they can repeat these attacks easily whenever they get a new target. I guess you could call that a market effect. So if you're not going to set MyFamily for yourself, do it for Tor users, and do it for Luther relay operators. We prioritise the safety of users and relay operators here. T
Am So., 23. Feb. 2020 um 01:55 Uhr schrieb teor <teor@riseup.net>:
Hi,
I've gone a few emails back up the thread, because the risk analysis is missing some really important factors.
And just some reminders:
Some users depend on the tor network for their safety.
Relay operators take some risks, but we do our best to reduce those risks.
MyFamily is about user and operator safety. We pay more attention to arguments based on safety.
On 22 Feb 2020, at 23:02, [REDACTED] < [REDACTED]> wrote:
So for what reason do i set the MyFamily option beside making a Hidden
Service Guard discovery attack more easy?
- risk reduction for tor users MyFamily declarations allow the tor client software to automatically detect relay families when creating circuits to avoid using multiple relays from the same operator in a single circuit.
This should not matter if the operator is not malicious and like i already said an malicious operator will not use the same contact info or relay name.
- reducing the risk for tor users that might become victims if some
operator gets compromized (with all its relays)
This is a reason i can understand. Not sure how much that would really help in practice but i can understand it.
In practice, relay operators become targets for compromise when they don't set MyFamily. Because those relays can be used to attack a Tor users.
If relay operators correctly set MyFamily, then an attacker needs to compromise multiple operators to see a single user's traffic.
In this case, it doesn't matter if the operator is malicious.
Understood. So for example if someone compromise multiple of my relays without me noticing it and installs software on them (or the providers network) to do a traffic correlations attack i am a less interesting target when i have set MyFamily. Another benefit of a proper MyFamily setting in this case would be that he first would need to remove the MyFamily to see any interesting traffic which i would most likely realize faster than without a proper MyFamily setting. This is indeed something what makes me very uncomfortable because it would be my fault if someones privacy would get affected by this.
- transparency
Every relay operator should declare their relay group to allow everybody to measure their network fraction (Sybil detection).
Should... But i understand this one too. But as long as my family is still a small one with only one exit compared to others i am not a Sybil attack risk and even if i would would i get any special treatment then?
It doesn't matter how small your relays are. Some clients will choose your relays as guards. You're putting those users in danger.
I understand this one as related to the first one.
- risk reduction for relay operators
MyFamily also provides risk reduction for operators since they are less valuable as an attack target if they can not technically be used for e2e correlation attacks
I think this is similar to your first point but i think that should be the operators choice if he want to take steps against this case.
There's also a network effect here. If almost all operators set MyFamily, then the Tor Network becomes a less valuable target for attacks. So attackers use other methods, like attacking Tor Browser, or offline attacks.
But if a lot of operators don't set MyFamily, then attackers develop tools and techniques to attack the network. Then they can repeat these attacks easily whenever they get a new target. I guess you could call that a market effect.
Understood.
So if you're not going to set MyFamily for yourself, do it for Tor users, and do it for Luther relay operators.
Will try to do it tomorrow.
We prioritise the safety of users and relay operators here.
T _______________________________________________ tor-relays mailing list tor-relays@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays
participants (5)
-
[REDACTED] -
Moritz Bartl -
nusenu -
teor -
Toralf Förster