[Script Info]
Title: 
[Events]
Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text
Dialogue: 0,0:00:00.00,0:00:07.94,Default,,0000,0000,0000,,{\i1}intro music{\i0}
Dialogue: 0,0:00:07.94,0:00:13.07,Default,,0000,0000,0000,,Herald: This is now \N"Towards a more trustworthy Tor network"
Dialogue: 0,0:00:13.07,0:00:15.11,Default,,0000,0000,0000,,by Nusenu
Dialogue: 0,0:00:15.11,0:00:21.58,Default,,0000,0000,0000,,The talk will give examples of malicious\Nrelay groups and current issues and how to
Dialogue: 0,0:00:21.58,0:00:28.25,Default,,0000,0000,0000,,tackle those to empower Tor users for\Nself-defense, so they don't necessarily
Dialogue: 0,0:00:28.25,0:00:31.82,Default,,0000,0000,0000,,need to rely on the detection and removal\Nof those groups.
Dialogue: 0,0:00:31.82,0:00:36.28,Default,,0000,0000,0000,,So without further ado, enjoy!
Dialogue: 0,0:00:36.28,0:00:40.77,Default,,0000,0000,0000,,And we will see each other \Nfor Q&A afterwards.
Dialogue: 0,0:00:45.53,0:00:49.62,Default,,0000,0000,0000,,Nusenu: Thanks for inviting me to give a\Ntalk about something I deeply care about:
Dialogue: 0,0:00:49.62,0:00:51.75,Default,,0000,0000,0000,,The Tor network.
Dialogue: 0,0:00:51.75,0:00:54.27,Default,,0000,0000,0000,,The Tor network \Nis a crucial privacy infrastructure,
Dialogue: 0,0:00:54.27,0:00:57.34,Default,,0000,0000,0000,,without which, \Nwe could not use Tor Browser.
Dialogue: 0,0:00:57.34,0:01:02.26,Default,,0000,0000,0000,,I like to uncover malicious Tor relays \Nto help protect Tor users.
Dialogue: 0,0:01:02.26,0:01:06.71,Default,,0000,0000,0000,,But since that does not come without \Npersonal risks, I'm taking steps
Dialogue: 0,0:01:06.71,0:01:10.24,Default,,0000,0000,0000,,to protect myself from those \Nrunning those malicious nodes,
Dialogue: 0,0:01:10.24,0:01:12.82,Default,,0000,0000,0000,,so I can continue to fight them.
Dialogue: 0,0:01:12.82,0:01:17.66,Default,,0000,0000,0000,,For this reason, this is a prerecorded \Ntalk without using my own voice.
Dialogue: 0,0:01:18.38,0:01:20.64,Default,,0000,0000,0000,,Thanks to the people behind the scenes
Dialogue: 0,0:01:20.64,0:01:24.77,Default,,0000,0000,0000,,who made it possible to \Npresent this talk in a safe way.
Dialogue: 0,0:01:26.88,0:01:28.88,Default,,0000,0000,0000,,A few words about me.
Dialogue: 0,0:01:28.88,0:01:32.15,Default,,0000,0000,0000,,I have a long-standing interest \Nin the state of the Tor network.
Dialogue: 0,0:01:32.15,0:01:37.89,Default,,0000,0000,0000,,In 2015, I started OrNetRadar, \Nwhich is a public mailing list and
Dialogue: 0,0:01:37.89,0:01:43.48,Default,,0000,0000,0000,,website showing reports about new\Nrelay groups and possible Sybil attacks.
Dialogue: 0,0:01:43.48,0:01:50.18,Default,,0000,0000,0000,,In 2017, I was asked to join the private \Nbad-relays Tor Project mailing list
Dialogue: 0,0:01:50.18,0:01:55.03,Default,,0000,0000,0000,,to help analyze and confirm reports\Nabout malicious relays.
Dialogue: 0,0:01:56.43,0:02:01.09,Default,,0000,0000,0000,,To get a better understanding of who runs\Nwhat fraction of the Tor network over time
Dialogue: 0,0:02:01.09,0:02:07.54,Default,,0000,0000,0000,,I started OrNetStats. It shows you also\Nwhich operators could de-anonymize Tor
Dialogue: 0,0:02:07.54,0:02:12.55,Default,,0000,0000,0000,,users because they are in a position \Nto perform end-to-end correlation attacks,
Dialogue: 0,0:02:12.55,0:02:14.82,Default,,0000,0000,0000,,something we will describe later.
Dialogue: 0,0:02:14.82,0:02:19.92,Default,,0000,0000,0000,,I'm also the maintainer of \Nansible-relayor, which is an Ansible role
Dialogue: 0,0:02:19.92,0:02:22.93,Default,,0000,0000,0000,,used by many large relay operators.
Dialogue: 0,0:02:23.41,0:02:27.69,Default,,0000,0000,0000,,Out of curiosity, I also like\Nengaging in some limited open-source
Dialogue: 0,0:02:27.69,0:02:32.27,Default,,0000,0000,0000,,intelligence gathering on malicious \NTor network actors, especially when
Dialogue: 0,0:02:32.27,0:02:35.76,Default,,0000,0000,0000,,their motivation for running relays \Nhas not been well understood.
Dialogue: 0,0:02:36.60,0:02:39.62,Default,,0000,0000,0000,,To avoid confusions, \Nwith regards to the Tor Project:
Dialogue: 0,0:02:39.62,0:02:45.02,Default,,0000,0000,0000,,I am not employed by the Tor Project \Nand I do not speak for the Tor Project.
Dialogue: 0,0:02:47.89,0:02:52.84,Default,,0000,0000,0000,,In this presentation, we will go through\Nsome examples of malicious actors on
Dialogue: 0,0:02:52.84,0:02:58.56,Default,,0000,0000,0000,,the Tor network. They basically represent\Nour problem statement that motivates us to
Dialogue: 0,0:02:58.56,0:03:04.03,Default,,0000,0000,0000,,improve the "status quo". After describing\Nsome issues with current approaches to
Dialogue: 0,0:03:04.03,0:03:09.06,Default,,0000,0000,0000,,fight malicious relays, we present a new,\Nadditional approach aiming at achieving a
Dialogue: 0,0:03:09.06,0:03:13.74,Default,,0000,0000,0000,,safer Tor experience using trusted relays\Nto some extent.
Dialogue: 0,0:03:14.76,0:03:17.87,Default,,0000,0000,0000,,The primary target audience \Nof this presentation are:
Dialogue: 0,0:03:17.87,0:03:24.74,Default,,0000,0000,0000,,Tor users, like Tor Browser users, \Nrelay operators,
Dialogue: 0,0:03:24.74,0:03:29.17,Default,,0000,0000,0000,,onion service operators \Nlike, for example, SecureDrop
Dialogue: 0,0:03:29.17,0:03:33.03,Default,,0000,0000,0000,,and anyone else that cares about Tor.
Dialogue: 0,0:03:35.80,0:03:40.60,Default,,0000,0000,0000,,To get everyone on the same page, \Na quick refresher on how Tor works
Dialogue: 0,0:03:40.60,0:03:45.19,Default,,0000,0000,0000,,and what type of relays – also\Ncalled nodes – there are.
Dialogue: 0,0:03:45.19,0:03:50.19,Default,,0000,0000,0000,,When Alice uses Tor Browser \Nto visit Bob's website,
Dialogue: 0,0:03:50.19,0:03:56.01,Default,,0000,0000,0000,,her Tor client selects three Tor relays \Nto construct a circuit that will be used
Dialogue: 0,0:03:56.01,0:04:00.47,Default,,0000,0000,0000,,to route her traffic through the \NTor network before it reaches Bob.
Dialogue: 0,0:04:00.47,0:04:03.87,Default,,0000,0000,0000,,This gives Alice location anonymity.
Dialogue: 0,0:04:03.87,0:04:08.83,Default,,0000,0000,0000,,The first relay in such a circuit \Nis called an entry guard relay.
Dialogue: 0,0:04:08.83,0:04:14.80,Default,,0000,0000,0000,,This relay is the only relay seeing \NAlice's real IP address and is therefore
Dialogue: 0,0:04:14.80,0:04:20.59,Default,,0000,0000,0000,,considered a more sensitive type of relay. \NThe guard relay does not learn that Alice
Dialogue: 0,0:04:20.59,0:04:25.74,Default,,0000,0000,0000,,is connecting to Bob, though. It \Nonly sees the next relay as destination.
Dialogue: 0,0:04:25.74,0:04:30.93,Default,,0000,0000,0000,,Guard relays are not changed frequently,\Nand Alice's Tor client waits up to 12
Dialogue: 0,0:04:30.93,0:04:35.77,Default,,0000,0000,0000,,weeks before choosing a new guard \Nto make some attacks less effective.
Dialogue: 0,0:04:35.77,0:04:42.40,Default,,0000,0000,0000,,The second relay is called a middle \Nor middle only relay. This relay
Dialogue: 0,0:04:42.40,0:04:46.96,Default,,0000,0000,0000,,is the least sensitive position, since it\Nonly sees other relays, but does not know
Dialogue: 0,0:04:46.96,0:04:51.14,Default,,0000,0000,0000,,anything about Alice or Bob because it\Njust forwards encrypted traffic.
Dialogue: 0,0:04:51.60,0:04:56.39,Default,,0000,0000,0000,,And, \Nthe final relay is called an exit relay.
Dialogue: 0,0:04:56.39,0:05:00.99,Default,,0000,0000,0000,,The exit relay gets to learn the \Ndestination, Bob, but does not know
Dialogue: 0,0:05:00.99,0:05:06.20,Default,,0000,0000,0000,,who is connecting to Bob. \NThe exit relay is also considered
Dialogue: 0,0:05:06.20,0:05:11.35,Default,,0000,0000,0000,,a more sensitive relay type, since it\Npotentially gets to see and manipulate
Dialogue: 0,0:05:11.35,0:05:20.13,Default,,0000,0000,0000,,clear text traffic (if Alice is not using\Nan encrypted protocol like HTTPS.)
Dialogue: 0,0:05:20.13,0:05:25.69,Default,,0000,0000,0000,,Although exit relays see the destination,\Nthey can not link all sites Alice visits
Dialogue: 0,0:05:25.69,0:05:31.86,Default,,0000,0000,0000,,at a given point in time to the same Tor\Nclient, to profile her, because Alice's
Dialogue: 0,0:05:31.86,0:05:36.31,Default,,0000,0000,0000,,Tor Browser instructs the Tor client to\Ncreate and use distinct circuits for
Dialogue: 0,0:05:36.31,0:05:42.93,Default,,0000,0000,0000,,distinct URL bar domains. So, although\Nthis diagram shows a single circuit only,
Dialogue: 0,0:05:42.93,0:05:49.03,Default,,0000,0000,0000,,a Tor client usually has multiple open Tor\Ncircuits at the same time. In networks
Dialogue: 0,0:05:49.03,0:05:56.50,Default,,0000,0000,0000,,where Tor is censored, users make use of a\Nspecial node type, which is called Bridge.
Dialogue: 0,0:05:56.50,0:06:01.64,Default,,0000,0000,0000,,Their primary difference is that they are\Nnot included in the public list of relays,
Dialogue: 0,0:06:01.64,0:06:07.19,Default,,0000,0000,0000,,to make it harder to censor them. Alice\Nhas to manually configure Tor Browser if
Dialogue: 0,0:06:07.19,0:06:13.28,Default,,0000,0000,0000,,she wants to use a bridge. For redundancy,\Nit is good to have more than one bridge in
Dialogue: 0,0:06:13.28,0:06:16.31,Default,,0000,0000,0000,,case a bridge goes down or gets censored.
Dialogue: 0,0:06:16.31,0:06:22.03,Default,,0000,0000,0000,,The used bridge also gets to see Alice's \Nreal IP address, but not the destination.
Dialogue: 0,0:06:24.68,0:06:28.33,Default,,0000,0000,0000,,Now that we have a basic \Nunderstanding of Tor's design,
Dialogue: 0,0:06:28.33,0:06:30.67,Default,,0000,0000,0000,,we might wonder,\Nwhy do we need to trust the network,
Dialogue: 0,0:06:30.67,0:06:35.55,Default,,0000,0000,0000,,when roles are distributed \Nacross multiple relays?
Dialogue: 0,0:06:35.55,0:06:38.85,Default,,0000,0000,0000,,So let's look into some attack scenarios.
Dialogue: 0,0:06:40.72,0:06:44.72,Default,,0000,0000,0000,,If an attacker controls \NAlice's guard and exit relay,
Dialogue: 0,0:06:44.72,0:06:47.46,Default,,0000,0000,0000,,they can learn that Alice connected to Bob
Dialogue: 0,0:06:47.46,0:06:51.15,Default,,0000,0000,0000,,by performing \Nend-to-end correlation attacks.
Dialogue: 0,0:06:51.15,0:06:56.56,Default,,0000,0000,0000,,Such attacks can be passive, \Nmeaning no traffic is manipulated
Dialogue: 0,0:06:56.56,0:07:02.11,Default,,0000,0000,0000,,and therefore cannot be detected by \Nprobing suspect relays with test traffic.
Dialogue: 0,0:07:03.20,0:07:09.63,Default,,0000,0000,0000,,OrNetStats gives you a daily updated list \Nof potential operators in such a position.
Dialogue: 0,0:07:09.63,0:07:15.40,Default,,0000,0000,0000,,There are some restrictions a default\NTor client follows when building circuits
Dialogue: 0,0:07:15.40,0:07:19.02,Default,,0000,0000,0000,,to reduce the likelihood of this occurring
Dialogue: 0,0:07:19.02,0:07:24.61,Default,,0000,0000,0000,,For example, a Tor client does not use\Nmore than one relay in the same /16 IPv4
Dialogue: 0,0:07:24.61,0:07:30.93,Default,,0000,0000,0000,,network block when building circuits. For\Nexample, Alice's Tor client would never
Dialogue: 0,0:07:30.93,0:07:36.43,Default,,0000,0000,0000,,create this circuit because guard and exit\Nrelays are in the same net block one
Dialogue: 0,0:07:36.43,0:07:45.62,Default,,0000,0000,0000,,192.0./16. For this reason, the number of\Ndistinct /16 network blocks an attacker
Dialogue: 0,0:07:45.62,0:07:51.78,Default,,0000,0000,0000,,distributed its relays across is relevant\Nwhen evaluating this kind of risk.
Dialogue: 0,0:07:51.78,0:07:59.40,Default,,0000,0000,0000,,Honest relay operators declare their group\Nof relays in the so-called "MyFamily"
Dialogue: 0,0:07:59.40,0:08:04.64,Default,,0000,0000,0000,,setting. This way they are transparent\Nabout their set of relays and Tor clients
Dialogue: 0,0:08:04.64,0:08:09.09,Default,,0000,0000,0000,,automatically avoid using more than a\Nsingle relay from any given family in a
Dialogue: 0,0:08:09.09,0:08:14.77,Default,,0000,0000,0000,,single circuit. Malicious actors will\Neither not declare relay families or
Dialogue: 0,0:08:14.77,0:08:17.61,Default,,0000,0000,0000,,pretend to be in more than one family.
Dialogue: 0,0:08:19.89,0:08:24.68,Default,,0000,0000,0000,,Another variant of the end-to-end \Ncorrelation attack is possible
Dialogue: 0,0:08:24.68,0:08:28.95,Default,,0000,0000,0000,,when Bob is the attacker or \Nhas been compromised by the attacker,
Dialogue: 0,0:08:28.95,0:08:34.87,Default,,0000,0000,0000,,and the attacker also happens to run \NAlice's guard relay. In this case,
Dialogue: 0,0:08:34.87,0:08:40.88,Default,,0000,0000,0000,,the attacker can also determine\Nthe actual source IP address used by Alice
Dialogue: 0,0:08:40.88,0:08:43.33,Default,,0000,0000,0000,,when she visits Bob's website.
Dialogue: 0,0:08:45.50,0:08:50.47,Default,,0000,0000,0000,,In cases of large, suspicious, non-exit \Nrelay groups, it is also plausible that
Dialogue: 0,0:08:50.47,0:08:54.41,Default,,0000,0000,0000,,they are after onion services, because \Ncircuits for onion services do not require
Dialogue: 0,0:08:54.41,0:09:01.96,Default,,0000,0000,0000,,exit relays. Onion services provide \Nlocation anonymity to the server side.
Dialogue: 0,0:09:01.96,0:09:04.18,Default,,0000,0000,0000,,By running many non-exits,
Dialogue: 0,0:09:04.18,0:09:10.96,Default,,0000,0000,0000,,an attacker could aim at finding the real\NIP address / location of an onion service.
Dialogue: 0,0:09:13.04,0:09:17.55,Default,,0000,0000,0000,,Manipulating exit relays are probably\Nthe most common attack type
Dialogue: 0,0:09:17.55,0:09:23.03,Default,,0000,0000,0000,,detected in the wild. That is also\Nthe easiest-to-perform attack type.
Dialogue: 0,0:09:23.03,0:09:29.74,Default,,0000,0000,0000,,Malicious exits usually do not care who\NAlice is or what her actual IP address is.
Dialogue: 0,0:09:29.74,0:09:34.50,Default,,0000,0000,0000,,They are mainly interested to \Nprofit from traffic manipulation.
Dialogue: 0,0:09:35.49,0:09:40.74,Default,,0000,0000,0000,,This type of attack can be detected \Nby probing exits with decoy traffic,
Dialogue: 0,0:09:40.74,0:09:45.16,Default,,0000,0000,0000,,but since malicious exits moved \Nto more targeted approaches
Dialogue: 0,0:09:45.16,0:09:50.18,Default,,0000,0000,0000,,(specific domains only), detection \Nis less trivial than one might think.
Dialogue: 0,0:09:51.38,0:09:56.07,Default,,0000,0000,0000,,The best protection against this \Nkind of attack is using encryption.
Dialogue: 0,0:09:56.07,0:10:00.100,Default,,0000,0000,0000,,Malicious exit relays cannot harm \Nconnections going to onion services.
Dialogue: 0,0:10:02.89,0:10:06.63,Default,,0000,0000,0000,,Now, let's look into \Ntwo real-world examples
Dialogue: 0,0:10:06.63,0:10:10.47,Default,,0000,0000,0000,,of large scale and persistent \Nmalicious actors on the Tor network.
Dialogue: 0,0:10:12.51,0:10:20.06,Default,,0000,0000,0000,,The first example, tracked as BTCMITM20,\Nis in the malicious exit's business and
Dialogue: 0,0:10:20.06,0:10:26.98,Default,,0000,0000,0000,,performs SSL strip attacks on exit relays \Nto manipulate plaintext HTTP traffic,
Dialogue: 0,0:10:26.98,0:10:31.90,Default,,0000,0000,0000,,like Bitcoin addresses,\Nto divert Bitcoin transactions to them.
Dialogue: 0,0:10:31.90,0:10:37.98,Default,,0000,0000,0000,,They have been detected for the first time\Nin 2020, and had some pretty large relay
Dialogue: 0,0:10:37.98,0:10:44.33,Default,,0000,0000,0000,,groups. On this graph, you can see how \Nmuch of the Tor exit fraction was under
Dialogue: 0,0:10:44.33,0:10:50.23,Default,,0000,0000,0000,,their control in the first half of 2020. \NThe different colors represent different
Dialogue: 0,0:10:50.23,0:10:56.25,Default,,0000,0000,0000,,contact infos they gave on their relays \Nto pretend they are distinct groups.
Dialogue: 0,0:10:56.25,0:11:00.03,Default,,0000,0000,0000,,The sharp drops show events when \Nthey were removed from the network,
Dialogue: 0,0:11:00.03,0:11:02.66,Default,,0000,0000,0000,,before adding relays again.
Dialogue: 0,0:11:03.97,0:11:11.65,Default,,0000,0000,0000,,In February 2021, they managed over 27% \Nof the Tor network's exit capacity,
Dialogue: 0,0:11:11.65,0:11:15.84,Default,,0000,0000,0000,,despite multiple removal attempts \Nover almost a year.
Dialogue: 0,0:11:16.72,0:11:23.14,Default,,0000,0000,0000,,At some point in the future, \Nwe will hopefully have HTTPS-Only mode
Dialogue: 0,0:11:23.14,0:11:28.45,Default,,0000,0000,0000,,enabled by default in Tor Browser \Nto kill this entire attack vector for good
Dialogue: 0,0:11:28.45,0:11:32.20,Default,,0000,0000,0000,,and make malicious exits less lucrative.
Dialogue: 0,0:11:32.20,0:11:37.24,Default,,0000,0000,0000,,I encourage you to test \NHTTPS-Only mode in Tor Browser
Dialogue: 0,0:11:37.24,0:11:42.14,Default,,0000,0000,0000,,and notify website operators \Nthat do not work in that mode.
Dialogue: 0,0:11:42.14,0:11:45.59,Default,,0000,0000,0000,,If a website does not work \Nin HTTPS-Only mode,
Dialogue: 0,0:11:45.59,0:11:50.08,Default,,0000,0000,0000,,you also know it is probably \Nnot safe to use in the first place.
Dialogue: 0,0:11:51.81,0:11:57.11,Default,,0000,0000,0000,,The second example actor, \Ntracked as KAX17,
Dialogue: 0,0:11:57.11,0:12:02.73,Default,,0000,0000,0000,,is still somewhat of a mystery. And \Nthat is not the best situation to be in.
Dialogue: 0,0:12:02.73,0:12:07.03,Default,,0000,0000,0000,,They are remarkable for: \Ntheir focus on non-exit relays,
Dialogue: 0,0:12:07.03,0:12:13.04,Default,,0000,0000,0000,,their network diversity, \Nwith over 200 distinct /16 subnets,
Dialogue: 0,0:12:13.04,0:12:19.43,Default,,0000,0000,0000,,their size – it is the first actor I know\Nof that peaked at over 100 Gbit/s
Dialogue: 0,0:12:19.43,0:12:25.68,Default,,0000,0000,0000,,advertised non-exit bandwidth – and\Nthey are active since a very long time.
Dialogue: 0,0:12:27.42,0:12:32.06,Default,,0000,0000,0000,,Let's have a look at some KAX17 \Nrelated events in the past two years.
Dialogue: 0,0:12:32.06,0:12:38.86,Default,,0000,0000,0000,,I first detected and reported them \Nto the Tor Project in September 2019.
Dialogue: 0,0:12:38.86,0:12:44.08,Default,,0000,0000,0000,,In October 2019, \NKAX17 relays got removed
Dialogue: 0,0:12:44.08,0:12:47.46,Default,,0000,0000,0000,,by the Tor directory \Nauthorities for the first time.
Dialogue: 0,0:12:49.76,0:12:54.21,Default,,0000,0000,0000,,In December 2019,\NI published the first blog post about them
Dialogue: 0,0:12:54.21,0:12:57.78,Default,,0000,0000,0000,,At that point, they were already \Nrebuilding their infrastructure
Dialogue: 0,0:12:57.78,0:13:00.58,Default,,0000,0000,0000,,by adding new relays.
Dialogue: 0,0:13:01.83,0:13:07.79,Default,,0000,0000,0000,,In February 2020, I contacted an email \Naddress that was used on some relays that
Dialogue: 0,0:13:07.79,0:13:13.34,Default,,0000,0000,0000,,did not properly declare their relay group\Nusing the "MyFamily" setting. At the time,
Dialogue: 0,0:13:13.34,0:13:18.64,Default,,0000,0000,0000,,they said they would run bridges instead, \Nso they do not have to set MyFamily.
Dialogue: 0,0:13:18.64,0:13:23.81,Default,,0000,0000,0000,,Side note: \NMyFamily is not supported for bridges.
Dialogue: 0,0:13:23.81,0:13:30.29,Default,,0000,0000,0000,,I was not aware that this email address\Nis linked to KAX17 until October 2021.
Dialogue: 0,0:13:31.21,0:13:37.71,Default,,0000,0000,0000,,In the first half of 2020,\NI regularly reported large quantities of
Dialogue: 0,0:13:37.71,0:13:43.91,Default,,0000,0000,0000,,relays to the Tor Project, and they got\Nremoved at high pace until June 2020,
Dialogue: 0,0:13:43.91,0:13:48.44,Default,,0000,0000,0000,,when directory authorities changed their\Npractices and stopped removing them
Dialogue: 0,0:13:48.44,0:13:53.60,Default,,0000,0000,0000,,because they didn't want to "scare away" \Npotential new relay operators.
Dialogue: 0,0:13:54.86,0:13:58.52,Default,,0000,0000,0000,,In July 2020, an email address joined \Na tor-relays mailing list discussion
Dialogue: 0,0:13:58.52,0:14:07.18,Default,,0000,0000,0000,,I started about a proposal to limit \Nlarge-scale attacks on the network.
Dialogue: 0,0:14:07.18,0:14:12.72,Default,,0000,0000,0000,,Now we know \Nthat email address is linked to KAX17.
Dialogue: 0,0:14:13.92,0:14:18.44,Default,,0000,0000,0000,,Since the Tor directory authorities \Nno longer removed the relay groups
Dialogue: 0,0:14:18.44,0:14:24.09,Default,,0000,0000,0000,,showing up, I sent the information \Nof over 600 KAX17 relays
Dialogue: 0,0:14:24.09,0:14:26.52,Default,,0000,0000,0000,,to the public tor-talk mailing list.
Dialogue: 0,0:14:27.70,0:14:32.32,Default,,0000,0000,0000,,In October 2021, someone who asked for \Nanonymity reached out to me and provided a
Dialogue: 0,0:14:32.32,0:14:39.29,Default,,0000,0000,0000,,new way to detect Tor relay groups that \Ndo not run the official Tor software.
Dialogue: 0,0:14:40.52,0:14:45.35,Default,,0000,0000,0000,,Using this methodology, \Nwe were able to detect KAX17
Dialogue: 0,0:14:45.35,0:14:47.72,Default,,0000,0000,0000,,using a second detection method.
Dialogue: 0,0:14:47.72,0:14:52.22,Default,,0000,0000,0000,,This also apparently convinced \Nthe Tor directory authorities,
Dialogue: 0,0:14:52.22,0:14:57.10,Default,,0000,0000,0000,,and in November this year, \Na major removal event took place.
Dialogue: 0,0:14:58.53,0:15:03.89,Default,,0000,0000,0000,,Sadly, the time span during which\NKAX17 was running relays without
Dialogue: 0,0:15:03.89,0:15:11.48,Default,,0000,0000,0000,,limitations was rather long. \NThis motivated us to come up with a
Dialogue: 0,0:15:11.48,0:15:16.60,Default,,0000,0000,0000,,design that avoids this kind of complete \Ndependency on Tor directory authorities
Dialogue: 0,0:15:16.60,0:15:19.26,Default,,0000,0000,0000,,when it comes to safety issues.
Dialogue: 0,0:15:20.03,0:15:22.81,Default,,0000,0000,0000,,And, as you might guess,
Dialogue: 0,0:15:22.81,0:15:27.37,Default,,0000,0000,0000,,KAX17 is already attempting \Nto restore their foothold again.
Dialogue: 0,0:15:30.01,0:15:33.40,Default,,0000,0000,0000,,Here are some KAX17 properties.
Dialogue: 0,0:15:33.40,0:15:39.50,Default,,0000,0000,0000,,After the release of my second \NKAX17 blog post in November 2021,
Dialogue: 0,0:15:39.50,0:15:43.82,Default,,0000,0000,0000,,the media was quick with using \Nwords like "nation-state" and
Dialogue: 0,0:15:43.82,0:15:46.86,Default,,0000,0000,0000,,"Advanced Persistent Threat".
Dialogue: 0,0:15:46.86,0:15:53.18,Default,,0000,0000,0000,,But I find it hard to believe such such \Nserious entities would be so sloppy.
Dialogue: 0,0:15:53.18,0:15:58.06,Default,,0000,0000,0000,,Since they claim to work for an ISP \Nin every other email…
Dialogue: 0,0:15:58.80,0:16:02.14,Default,,0000,0000,0000,,I looked into their AS distribution.
Dialogue: 0,0:16:02.80,0:16:06.88,Default,,0000,0000,0000,,I guess they work for more than one ISP.
Dialogue: 0,0:16:07.69,0:16:10.62,Default,,0000,0000,0000,,This chart shows used Autonomous System,
Dialogue: 0,0:16:10.62,0:16:15.97,Default,,0000,0000,0000,,sorted by the unique IP addresses \Nused at that hoster. So, for example,
Dialogue: 0,0:16:15.97,0:16:21.18,Default,,0000,0000,0000,,They used more than 400 IP \Naddresses at Microsoft to run relays.
Dialogue: 0,0:16:21.18,0:16:27.01,Default,,0000,0000,0000,,These are not exact numbers,\Nsince it only includes relays since 2019,
Dialogue: 0,0:16:27.01,0:16:33.02,Default,,0000,0000,0000,,and there are likely more.\NIf we map their IP addresses
Dialogue: 0,0:16:33.02,0:16:39.05,Default,,0000,0000,0000,,to countries, we get this. Do not take\Nthis map too seriously, as the used GEOIP
Dialogue: 0,0:16:39.05,0:16:44.82,Default,,0000,0000,0000,,database was severely outdated and such\Ndatabases are never completely accurate,
Dialogue: 0,0:16:44.82,0:16:54.98,Default,,0000,0000,0000,,but it gives us a rough idea. To be clear,\NI have no evidence that KAX17 is
Dialogue: 0,0:16:54.98,0:17:00.67,Default,,0000,0000,0000,,performing any kind of attacks against Tor\Nusers, but in our threat model it is
Dialogue: 0,0:17:00.67,0:17:06.07,Default,,0000,0000,0000,,already a considerable risk if even a\Nbenevolent operator is not declaring their
Dialogue: 0,0:17:06.07,0:17:12.04,Default,,0000,0000,0000,,more than 800 relays as a family. Good\Nprotections should protect against
Dialogue: 0,0:17:12.04,0:17:17.81,Default,,0000,0000,0000,,benevolent and malicious Sybil attacks\Nequally. The strongest input factor for
Dialogue: 0,0:17:17.81,0:17:23.24,Default,,0000,0000,0000,,the risk assessment of this actor is the\Nfact they do not run the official Tor
Dialogue: 0,0:17:23.24,0:17:28.77,Default,,0000,0000,0000,,software on their relays. There are still\Nmany open questions, and the analysis into
Dialogue: 0,0:17:28.77,0:17:39.37,Default,,0000,0000,0000,,KAX17 is ongoing. If you have any input,\Nfeel free to reach out to me. After
Dialogue: 0,0:17:39.37,0:17:44.42,Default,,0000,0000,0000,,looking at some examples of malicious\Nactors, I want to shortly summarize some
Dialogue: 0,0:17:44.42,0:17:51.52,Default,,0000,0000,0000,,of the issues in how the malicious relays\Nproblem is currently approached. It is
Dialogue: 0,0:17:51.52,0:17:56.69,Default,,0000,0000,0000,,pretty much like playing Whack-A-Mole. You\Nhit them and they come back. You hit them
Dialogue: 0,0:17:56.69,0:18:02.62,Default,,0000,0000,0000,,again, and they come back again, over and\Nover and while you're at it, you're also
Dialogue: 0,0:18:02.62,0:18:08.58,Default,,0000,0000,0000,,training them to come back stronger next\Ntime. Malicious actors can run relays
Dialogue: 0,0:18:08.58,0:18:14.67,Default,,0000,0000,0000,,until they get caught/detected or are\Nconsidered suspicious enough for removal
Dialogue: 0,0:18:14.67,0:18:20.63,Default,,0000,0000,0000,,by a Tor directory authorities. If your\Nthreat model does not match the Tor
Dialogue: 0,0:18:20.63,0:18:25.18,Default,,0000,0000,0000,,directory's threat model, you are out of\Nluck or have to maintain your own
Dialogue: 0,0:18:25.18,0:18:31.37,Default,,0000,0000,0000,,exclusion lists. Attempts to define a\Nformer set of "do not do" requirements for
Dialogue: 0,0:18:31.37,0:18:36.99,Default,,0000,0000,0000,,relays that Tor directory authorities\Ncommit to enforce, have failed, even with
Dialogue: 0,0:18:36.99,0:18:45.76,Default,,0000,0000,0000,,the involvement of a core Tor developer.\NIt is time for a paradigm change. The
Dialogue: 0,0:18:45.76,0:18:50.80,Default,,0000,0000,0000,,current processes for detecting and\Nremoving malicious Tor relays are failing
Dialogue: 0,0:18:50.80,0:18:56.66,Default,,0000,0000,0000,,us and are not sustainable in the long\Nrun. In recent years, malicious groups
Dialogue: 0,0:18:56.66,0:19:05.66,Default,,0000,0000,0000,,have become larger, harder to detect,\Nharder to get removed and more persistent.
Dialogue: 0,0:19:05.66,0:19:11.97,Default,,0000,0000,0000,,Here are some of our design goals. Instead\Nof continuing the single sided arms race
Dialogue: 0,0:19:11.97,0:19:17.39,Default,,0000,0000,0000,,with malicious actors. We aim to empower\NTor users for self-defense without
Dialogue: 0,0:19:17.39,0:19:22.26,Default,,0000,0000,0000,,requiring the detection of malicious Tor\Nrelays and without, solely, depending on
Dialogue: 0,0:19:22.26,0:19:28.39,Default,,0000,0000,0000,,Tor directly authorities for protecting us\Nfrom malicious relays. We aim to reduce
Dialogue: 0,0:19:28.39,0:19:36.90,Default,,0000,0000,0000,,the risk of de-anonymization by using at\Nleast a trusted guard or exit or both. We
Dialogue: 0,0:19:36.90,0:19:41.44,Default,,0000,0000,0000,,also acknowledge it is increasingly\Nimpossible to detect all malicious relays
Dialogue: 0,0:19:41.44,0:19:46.99,Default,,0000,0000,0000,,using decoy traffic, therefore, we stop\Ndepending on the detectability of
Dialogue: 0,0:19:46.99,0:19:56.27,Default,,0000,0000,0000,,malicious relays to protect users. In\Ntoday's Tor network, we hope to not choose
Dialogue: 0,0:19:56.27,0:20:01.52,Default,,0000,0000,0000,,a malicious guard when we pick one. In the\Nproposed design, we would pick a trusted
Dialogue: 0,0:20:01.52,0:20:07.62,Default,,0000,0000,0000,,guard instead. In fact, this can be done\Nwith today's Tor browser, if you set any
Dialogue: 0,0:20:07.62,0:20:14.29,Default,,0000,0000,0000,,trusted relays as your bridge. Another\Nsupported configuration would be to use
Dialogue: 0,0:20:14.29,0:20:20.17,Default,,0000,0000,0000,,trusted guards and trusted exits. Such\Ndesigns are possible without requiring
Dialogue: 0,0:20:20.17,0:20:25.37,Default,,0000,0000,0000,,code changes in Tor, but are cumbersome to\Nconfigure manually, since Tor only
Dialogue: 0,0:20:25.37,0:20:33.31,Default,,0000,0000,0000,,supports relay fingerprints and does not\Nknow about relay operator identifiers. But
Dialogue: 0,0:20:33.31,0:20:39.46,Default,,0000,0000,0000,,what do we actually mean by trusted\Nrelays? Trusted relays are operated by
Dialogue: 0,0:20:39.46,0:20:45.38,Default,,0000,0000,0000,,trusted operators. These operators are\Nbelieved to run relays without malicious
Dialogue: 0,0:20:45.38,0:20:51.56,Default,,0000,0000,0000,,intent. Trusted operators are specified by\Nthe user. Users assign trust at the
Dialogue: 0,0:20:51.56,0:20:57.45,Default,,0000,0000,0000,,operator, not the relay level, for\Nscalability reasons, and to avoid
Dialogue: 0,0:20:57.45,0:21:05.99,Default,,0000,0000,0000,,configuration changes when an operator\Nchanges their relays. Since users should
Dialogue: 0,0:21:05.99,0:21:12.10,Default,,0000,0000,0000,,be able to specify trusted operators, we\Nneed human-readable, authenticated and
Dialogue: 0,0:21:12.10,0:21:19.03,Default,,0000,0000,0000,,globally unique operator identifiers. By\Nauthenticated, we mean they should not be
Dialogue: 0,0:21:19.03,0:21:26.70,Default,,0000,0000,0000,,spoofable arbitrarily like current relay\Ncontact infos. For simplicity, we use DNS
Dialogue: 0,0:21:26.70,0:21:34.14,Default,,0000,0000,0000,,domains as relay operator identifiers, and\Nwe will probably restrict them to 40
Dialogue: 0,0:21:34.14,0:21:47.00,Default,,0000,0000,0000,,characters in length. How do Authenticated\NRelay Operator IDs, short AROI, work. From
Dialogue: 0,0:21:47.00,0:21:52.49,Default,,0000,0000,0000,,an operator point of view, configuring an\NAROI is easy. Step one: The operator
Dialogue: 0,0:21:52.49,0:21:59.56,Default,,0000,0000,0000,,specifies the desired domain under her\Ncontrol using Tor's ContactInfo option.
Dialogue: 0,0:21:59.56,0:22:06.85,Default,,0000,0000,0000,,Step two: The operator publishes a simple\Ntext file using the IANA well-known URI
Dialogue: 0,0:22:06.85,0:22:13.10,Default,,0000,0000,0000,,containing all relay fingerprints. If no\Nweb server is available or if the web
Dialogue: 0,0:22:13.10,0:22:18.71,Default,,0000,0000,0000,,server is not considered safe enough,\NDNSSEC-signed TXT records are also an
Dialogue: 0,0:22:18.71,0:22:25.58,Default,,0000,0000,0000,,option for authentication. Using DNS is\Ngreat for scalability and availability due
Dialogue: 0,0:22:25.58,0:22:31.30,Default,,0000,0000,0000,,to DNS caching, but since every relay\Nrequires its own TXT record, it will take
Dialogue: 0,0:22:31.30,0:22:37.11,Default,,0000,0000,0000,,longer than the URI type proof when\Nperforming proof validation. Operators
Dialogue: 0,0:22:37.11,0:22:42.37,Default,,0000,0000,0000,,that have no domain at all can use free\Nservices like GitHub pages or similar to
Dialogue: 0,0:22:42.37,0:22:49.87,Default,,0000,0000,0000,,serve the text file. For convenience, Eran\NSandler created this simple to use
Dialogue: 0,0:22:49.87,0:22:55.10,Default,,0000,0000,0000,,ContactInfo generator, so relay operators\Ndon't have to read the specification to
Dialogue: 0,0:22:55.10,0:23:00.53,Default,,0000,0000,0000,,generate the required ContactInfo string\Nfor their configuration. For the
Dialogue: 0,0:23:00.53,0:23:06.29,Default,,0000,0000,0000,,Authenticated Relay Operator ID the "url"\Nand "proof" fields are the only relevant
Dialogue: 0,0:23:06.29,0:23:13.72,Default,,0000,0000,0000,,fields. There are already over 1000 relays\Nthat have implemented the Authenticated
Dialogue: 0,0:23:13.72,0:23:21.69,Default,,0000,0000,0000,,Relay Operator ID. OrNetStats displays an\Nicon in case the operator implemented it
Dialogue: 0,0:23:21.69,0:23:29.11,Default,,0000,0000,0000,,correctly. Out of the top 24 largest\Nfamilies by bandwidth, all but eight
Dialogue: 0,0:23:29.11,0:23:34.72,Default,,0000,0000,0000,,operators have implemented the\NAuthenticated Relay Operator ID already.
Dialogue: 0,0:23:34.72,0:23:40.38,Default,,0000,0000,0000,,On the right-hand side, you can see a few\Nlogos of organizations running relays with
Dialogue: 0,0:23:40.38,0:23:46.64,Default,,0000,0000,0000,,a properly set up AROI. The most relevant\Ndistinction between lines having that
Dialogue: 0,0:23:46.64,0:23:51.79,Default,,0000,0000,0000,,checkmark icon and those that do not have\Nit is the fact that the string in lines
Dialogue: 0,0:23:51.79,0:23:59.41,Default,,0000,0000,0000,,that do not include the icon can be\Narbitrarily spoofed. This graph shows the
Dialogue: 0,0:23:59.41,0:24:05.55,Default,,0000,0000,0000,,largest exit operators that implemented\Nthe AROI. I want to stress one crucial
Dialogue: 0,0:24:05.55,0:24:12.66,Default,,0000,0000,0000,,point about AROIs though, authenticated\Nmust not be confused with trusted.
Dialogue: 0,0:24:12.66,0:24:19.20,Default,,0000,0000,0000,,Malicious operators can also authenticate\Ntheir domain and they do. A given AROI can
Dialogue: 0,0:24:19.20,0:24:25.53,Default,,0000,0000,0000,,be trusted or not. It is up to the user,\Nbut using AROIs instead of ContactInfo for
Dialogue: 0,0:24:25.53,0:24:30.67,Default,,0000,0000,0000,,assigning trust is crucial because\NContactInfo can not be trusted directly
Dialogue: 0,0:24:30.67,0:24:37.84,Default,,0000,0000,0000,,without further checks. This graph shows\Nwhat fraction of the Tor network's exit
Dialogue: 0,0:24:37.84,0:24:43.77,Default,,0000,0000,0000,,capacity implemented the Authenticated\NRelay Operator ID over time. Currently, we
Dialogue: 0,0:24:43.77,0:24:49.53,Default,,0000,0000,0000,,are at around 60 percent already, but\Nguard capacity is a lot lower, around 15
Dialogue: 0,0:24:49.53,0:24:55.88,Default,,0000,0000,0000,,percent. The reason for that is that exits\Nare operated mostly by large operators and
Dialogue: 0,0:24:55.88,0:25:00.91,Default,,0000,0000,0000,,organizations, while guards are\Ndistributed across a lot more operators.
Dialogue: 0,0:25:00.91,0:25:11.45,Default,,0000,0000,0000,,There are over 1800 guard families, but\Nonly around 400 exit families. How does a
Dialogue: 0,0:25:11.45,0:25:19.38,Default,,0000,0000,0000,,Tor client make use of AROIs, current Tor\Nversions do not know what AROIs are and
Dialogue: 0,0:25:19.38,0:25:24.69,Default,,0000,0000,0000,,primarily take relay fingerprints as\Nconfiguration inputs. So, we need some
Dialogue: 0,0:25:24.69,0:25:28.98,Default,,0000,0000,0000,,tooling to generate a list of relay\Nfingerprints starting from a list of
Dialogue: 0,0:25:28.98,0:25:36.92,Default,,0000,0000,0000,,trusted AROIs. We have implemented a quick\Nand dirty proof of concept that puts
Dialogue: 0,0:25:36.92,0:25:41.37,Default,,0000,0000,0000,,everything together and performs all the\Nsteps shown on this slide, to demonstrate
Dialogue: 0,0:25:41.37,0:25:47.10,Default,,0000,0000,0000,,the concept of using trusted AROIs to\Nconfigure Tor client to use trusted exit
Dialogue: 0,0:25:47.10,0:25:53.64,Default,,0000,0000,0000,,relays. It is not meant to be used by end-\Nusers, it merely is a preview for the
Dialogue: 0,0:25:53.64,0:25:57.57,Default,,0000,0000,0000,,technical audience who would like to see\Nit in action to achieve a better
Dialogue: 0,0:25:57.57,0:26:03.87,Default,,0000,0000,0000,,understanding of the design. The current\Nproof of concept performs all proof checks
Dialogue: 0,0:26:03.87,0:26:09.80,Default,,0000,0000,0000,,itself without relying on third parties,\Nbut since there are a lot of reasons for
Dialogue: 0,0:26:09.80,0:26:15.08,Default,,0000,0000,0000,,doing proof-checks centrally instead, for\Nexample, by directory authorities. I
Dialogue: 0,0:26:15.08,0:26:21.08,Default,,0000,0000,0000,,recently submitted a partial proposal for\Nit to the Tor development mailing list to
Dialogue: 0,0:26:21.08,0:26:25.08,Default,,0000,0000,0000,,see whether they would consider it before\Nproceeding with a more serious
Dialogue: 0,0:26:25.08,0:26:30.95,Default,,0000,0000,0000,,implementation than the current proof of\Nconcept. I find it important to always try
Dialogue: 0,0:26:30.95,0:26:35.82,Default,,0000,0000,0000,,achieving a common goal together with\Nupstream first before creating solutions
Dialogue: 0,0:26:35.82,0:26:40.77,Default,,0000,0000,0000,,that are maintained outside of upstream\Nbecause it will lead to better maintained
Dialogue: 0,0:26:40.77,0:26:46.18,Default,,0000,0000,0000,,improvements and likely a more user-\Nfriendly experience if they are integrated
Dialogue: 0,0:26:46.18,0:26:52.91,Default,,0000,0000,0000,,in upstream. Here is a link to the\Nmentioned tor-dev email, for those who
Dialogue: 0,0:26:52.91,0:27:01.79,Default,,0000,0000,0000,,would like to follow along. To summarize,\Nafter reviewing some real world examples
Dialogue: 0,0:27:01.79,0:27:08.20,Default,,0000,0000,0000,,of malicious actors on the Tor network, we\Nconcluded that current approaches to limit
Dialogue: 0,0:27:08.20,0:27:16.27,Default,,0000,0000,0000,,risks by bad relays on Tor users might not\Nlive up to Tor users expectations, are not
Dialogue: 0,0:27:16.27,0:27:22.49,Default,,0000,0000,0000,,sustainable in the long run and need an\Nupgrade to avoid depending on the
Dialogue: 0,0:27:22.49,0:27:28.61,Default,,0000,0000,0000,,detectability of malicious relays, which\Nis becoming increasingly hard. We
Dialogue: 0,0:27:28.61,0:27:34.91,Default,,0000,0000,0000,,presented a design to extend current anti\Nbad relay approaches that does not rely on
Dialogue: 0,0:27:34.91,0:27:41.34,Default,,0000,0000,0000,,the detection of malicious relays using\Ntrusted Authenticated Relay Operator IDs.
Dialogue: 0,0:27:41.34,0:27:47.08,Default,,0000,0000,0000,,We have shown that most exit capacity has\Nimplemented AROIs already, while guard
Dialogue: 0,0:27:47.08,0:27:53.29,Default,,0000,0000,0000,,capacity is currently significantly lower,\Nshowing a lack of insights on who operates
Dialogue: 0,0:27:53.29,0:28:00.10,Default,,0000,0000,0000,,Tor's guard capacity. When publicly\Nspeaking about modifying Tor's path
Dialogue: 0,0:28:00.10,0:28:06.59,Default,,0000,0000,0000,,selection in front of a wide audience, I\Nalso consider it to be my responsibility
Dialogue: 0,0:28:06.59,0:28:12.76,Default,,0000,0000,0000,,to explicitly state that you should not\Nchange your Tor configuration options that
Dialogue: 0,0:28:12.76,0:28:18.38,Default,,0000,0000,0000,,influenced path selection behavior without\Na clear need, according to your threat
Dialogue: 0,0:28:18.38,0:28:26.23,Default,,0000,0000,0000,,model to avoid potentially standing out.\NUsing trusted AROIs certainly comes with
Dialogue: 0,0:28:26.23,0:28:31.99,Default,,0000,0000,0000,,some tradeoffs of its own, like for\Nexample, network load balancing, to name
Dialogue: 0,0:28:31.99,0:28:38.57,Default,,0000,0000,0000,,only one. Thanks to many large, trusted\Nexit operators, it should be feasible in
Dialogue: 0,0:28:38.57,0:28:43.09,Default,,0000,0000,0000,,the near future to use trusted exits\Nwithout standing out in a trivially
Dialogue: 0,0:28:43.09,0:28:49.26,Default,,0000,0000,0000,,detectable way because it is harder in the\Nsense of takes longer to statistically
Dialogue: 0,0:28:49.26,0:28:55.11,Default,,0000,0000,0000,,detect a Tor client changed its possible\Npool of exits, if it only excluded a
Dialogue: 0,0:28:55.11,0:29:02.52,Default,,0000,0000,0000,,smaller fraction of exits. Detecting Tor\Nclients using only a subset of all guards
Dialogue: 0,0:29:02.52,0:29:08.54,Default,,0000,0000,0000,,takes a lot longer than detecting custom\Nexit sets because guards are not changed
Dialogue: 0,0:29:08.54,0:29:16.12,Default,,0000,0000,0000,,over a longer period of time when compared\Nwith exits. And finally, Tor clients that
Dialogue: 0,0:29:16.12,0:29:22.86,Default,,0000,0000,0000,,make use of trusted AROIs will need a way\Nto find trusted AROIs, ideally, they could
Dialogue: 0,0:29:22.86,0:29:29.79,Default,,0000,0000,0000,,learn about them dynamically in a safe\Nway. There is an early work in progress
Dialogue: 0,0:29:29.79,0:29:40.40,Default,,0000,0000,0000,,draft specification linked on this slide.\NI want to dedicate this talk to Karsten
Dialogue: 0,0:29:40.40,0:29:47.31,Default,,0000,0000,0000,,Loesing who passed away last year. He was\Nthe kindest person I got to interact with
Dialogue: 0,0:29:47.31,0:29:54.06,Default,,0000,0000,0000,,in the Tor community. Karsten was the Tor\Nmetrics team lead and without his work, my
Dialogue: 0,0:29:54.06,0:29:59.83,Default,,0000,0000,0000,,projects, OrNetStats and OrNetRadar would\Nnot exist. Every time you use
Dialogue: 0,0:29:59.83,0:30:07.39,Default,,0000,0000,0000,,metrics.torproject.org, for example, the\Nso-called "Relay Search", you are using
Dialogue: 0,0:30:07.39,0:30:14.73,Default,,0000,0000,0000,,his legacy. Thank you for listening, and\NI'm really looking forward to your
Dialogue: 0,0:30:14.73,0:30:19.63,Default,,0000,0000,0000,,questions. I'm not sure I'll be able to\Nrespond to questions after the talk in
Dialogue: 0,0:30:19.63,0:30:23.98,Default,,0000,0000,0000,,real time, but it would be nice to have\Nthem read out. So they are part of the
Dialogue: 0,0:30:23.98,0:30:28.66,Default,,0000,0000,0000,,recording and I'll make an effort to\Npublish answers to all of them via
Dialogue: 0,0:30:28.66,0:30:35.68,Default,,0000,0000,0000,,Mastodon, should I not be able to respond\Nin real time. I'm also happy to take tips
Dialogue: 0,0:30:35.68,0:30:40.72,Default,,0000,0000,0000,,about unusual things you observed on the\NTor network. Do not underestimate your
Dialogue: 0,0:30:40.72,0:30:48.20,Default,,0000,0000,0000,,power as Tor user to contribute to a safer\NTor network by reporting unusual things.
Dialogue: 0,0:30:48.20,0:30:55.05,Default,,0000,0000,0000,,Most major hits against bad relay actors\Nwere the result of Tor user reports.
Dialogue: 0,0:30:55.05,0:31:18.54,Default,,0000,0000,0000,,{\i1}quietness{\i0}
Dialogue: 0,0:31:18.54,0:31:27.81,Default,,0000,0000,0000,,Herald: OK. Thank you very much for this\Nvery informative talk and yes so we will
Dialogue: 0,0:31:27.81,0:31:41.06,Default,,0000,0000,0000,,switch over to the Q&A now. Yeah, thanks\Nagain. Very fascinating. So we have
Dialogue: 0,0:31:41.06,0:31:50.61,Default,,0000,0000,0000,,collected several questions from our IRC\Nchat, so I'm just going to start. If
Dialogue: 0,0:31:50.61,0:31:57.27,Default,,0000,0000,0000,,bridges don't need the MyFamily setting\Nisn't this a wide open gap for end-to-end
Dialogue: 0,0:31:57.27,0:32:03.84,Default,,0000,0000,0000,,correlation attacks, for example if a\Nmalicious actor can somehow make the relay
Dialogue: 0,0:32:03.84,0:32:10.18,Default,,0000,0000,0000,,popular as bridge?\NNusenu: Yes, bridges are a concern in the
Dialogue: 0,0:32:10.18,0:32:15.42,Default,,0000,0000,0000,,context of MyFamily, for that reason, it\Nis not recommended to run bridges and
Dialogue: 0,0:32:15.42,0:32:20.20,Default,,0000,0000,0000,,exits at the same time in current versions\Nof Tor, but future versions of Tor will
Dialogue: 0,0:32:20.20,0:32:26.20,Default,,0000,0000,0000,,get a new and more relay operator friendly\NMyFamily setting. That new MyFamily design
Dialogue: 0,0:32:26.20,0:32:33.54,Default,,0000,0000,0000,,will also support bridges. This will\Nlikely be in Tor 0.4.8.x at some point in
Dialogue: 0,0:32:33.54,0:32:45.11,Default,,0000,0000,0000,,2022.\NHerald: OK, thanks. Despite what kind of
Dialogue: 0,0:32:45.11,0:32:55.12,Default,,0000,0000,0000,,attack, are there statistics who or from\Nwhich country these attacks are coming
Dialogue: 0,0:32:55.12,0:33:02.80,Default,,0000,0000,0000,,most? Background here is there are rumors\Nabout NSA driven and exit notes.
Dialogue: 0,0:33:02.80,0:33:08.39,Default,,0000,0000,0000,,Nusenu: I don't know about any general\Nstatistics, but I usually include used
Dialogue: 0,0:33:08.39,0:33:13.61,Default,,0000,0000,0000,,autonomous systems by certain groups when\Nblogging about them. There are some
Dialogue: 0,0:33:13.61,0:33:17.90,Default,,0000,0000,0000,,autonomous systems that are notorious for\Nbeing used by malicious groups, but
Dialogue: 0,0:33:17.90,0:33:23.20,Default,,0000,0000,0000,,malicious groups also try to blend in with\Nthe rest by using large ISPs like Hetzner
Dialogue: 0,0:33:23.20,0:33:30.75,Default,,0000,0000,0000,,and OVH.\NHerald: Thanks. Is using a bridge that I
Dialogue: 0,0:33:30.75,0:33:35.01,Default,,0000,0000,0000,,host also safer than using a random guard\Nnode?
Dialogue: 0,0:33:35.01,0:33:41.79,Default,,0000,0000,0000,,Nusenu: This is a tricky question, since\Nit also depends on whether it is a private
Dialogue: 0,0:33:41.79,0:33:47.38,Default,,0000,0000,0000,,bridge, a bridge that is not distributed\Nto other uses by a bridgeDB. I would say
Dialogue: 0,0:33:47.38,0:33:53.47,Default,,0000,0000,0000,,it is better to not run the bridges you\Nuse yourself.
Dialogue: 0,0:33:53.47,0:34:02.76,Default,,0000,0000,0000,,Herald: OK. What is worse? KAX17 or a well\Nknown trusted operators running 20 percent
Dialogue: 0,0:34:02.76,0:34:06.85,Default,,0000,0000,0000,,of Tor's exits?\NNusenu: Currently, I would say KAX17.
Dialogue: 0,0:34:06.85,0:34:17.46,Default,,0000,0000,0000,,Herald: OK. I think that's the last one\Nfor now: Isn't the anonymity, not
Dialogue: 0,0:34:17.46,0:34:22.21,Default,,0000,0000,0000,,decreased or changed while using trusted\Nrelay list?
Dialogue: 0,0:34:22.21,0:34:27.00,Default,,0000,0000,0000,,Nusenu: Yes, this is a trade-off that\Nusers will need to make. This heavily
Dialogue: 0,0:34:27.00,0:34:36.86,Default,,0000,0000,0000,,depends on the threat model.\NHerald: OK. So I think we have gathered
Dialogue: 0,0:34:36.86,0:34:42.51,Default,,0000,0000,0000,,all the questions and they were all\Nanswered. So thank you again for. Yes,
Dialogue: 0,0:34:42.51,0:34:46.22,Default,,0000,0000,0000,,thank you again.
Dialogue: 0,0:34:46.22,0:34:59.18,Default,,0000,0000,0000,,{\i1}rc3 postroll music{\i0}
Dialogue: 0,0:34:59.18,0:35:03.00,Default,,0000,0000,0000,,Subtitles created by c3subtitles.de\Nin the year 2022. Join, and help us!#