- tor-dev - lists.torproject.org

Network team: New status page for 0.4.1; help needed on ticket triage
by Nick Mathewson 30 May '19

30 May '19

Hi! This mail is mainly for the network-team folks, but I'm sending it to tor-dev anyway in case it's helpful to anybody else. I've made a new page for our 0.4.1 status, based on Teor's previous 040 status page. Here it is: https://trac.torproject.org/projects/tor/wiki/org/teams/NetworkTeam/CoreTor… Note that this time I'm using two different keywords: 041-must and 041-should. The difference here is that 041 stuff should be things that would really block a release candidate, whereas 041-should stuff is stuff that we might in theory release without fixing. (It might annoy our users or violate our best practices, so we _should_ try to fix it, at least.) At the bottom of the page are all the tickets in the milestone that are *not* marked 041-must or 041-should. Some of these are things we could still fix anyway in 041; many will be removed. Let's triage these together! As discussed at the last meeting, we are currently in feature freeze for 0.4.1. Other than tickets that are marked "postfreeze-ok", we're planning not to take any more features in Tor 0.4.1. (Network team folks, please remember to spend your time on 041-must and 041-should tickets, and on Sponsor 19 tickets. If we make enough progress on these, we can open up 0.4.2 for merging early.) happy hacking, -- Nick

1 1

Proposal 302: Hiding onion service clients using WTF-PAD
by George Kadianakis 27 May '19

27 May '19

Filename: 302-padding-machines-for-onion-clients.txt Title: Hiding onion service clients using padding Author: George Kadianakis, Mike Perry Created: Thursday 16 May 2019 Status: Accepted Ticket: #28634 0. Overview Tor clients use "circuits" to do anonymous communications. There are various types of circuits. Some of them are for navigating the normal Internet, others are for fetching Tor directory information, others are for connecting to onion services, while others are simply for measurements and testing. It's currently possible for MITM type of adversaries (like tor-network-level and local-area-network adversaries) to distinguish Tor circuit types from each other using a wide array of metadata and distinguishers. In this proposal, we study various techniques that can be used to distinguish client-side onion service circuits and provide WTF-PAD circuit padding machines (using prop#254) to hide them against certain adversaries. 1. Motivation We are writing this proposal for various reasons: 1) We believe that in an ideal setting MITM adversaries should not be able to distinguish circuit types by inspecting traffic. Tor traffic should look amorphous to an outside observer to maximize uncertainty and anonymity properties. Client-side onion service circuits are an easy target for this proposal, because we believe we can improve their privacy with low bandwidth overhead. 2) We want to start experimenting with the WTF-PAD subsystem of Tor, and this use-case provides us with a good testbed. 3) We hope that by actually starting to use the WTF-PAD subsystem of Tor, we will encourage more researchers to start experimenting with it. 2. Scope of the proposal [SCOPE] Given the above, this proposal sets forth to use the WTF-PAD system to hide client-side onion service circuits against the classifiers of paper by Kwon et al. above. By client-side onion service circuits we refer to these two types of circuits: - Client-side introduction circuits: Circuit from client to the introduction point - Client-side rendezvous circuits: Circuit from client to the rendezvous point Service-side onion service circuits are not in scope for this proposal, and this is because hiding those would require more bandwidth and also more advanced WTF-PAD features. Furthermore, this proposal only aims to cloak the naive distinguishing features mentioned in the [KNOWN_DISTINGUISHERS] section, and can by no means guarantee that client-side onion service circuits are totally indistinguishable by other means. The machines specified in this proposal are meant to be lightweight and created for a specific purpose. This means that they can be easily extended with additional states to do more advanced hiding. 3. Known distinguishers against onion service circuits [KNOWN_DISTINGUISHERS] Over the past years it's been assumed that motivated adversaries can distinguish onion-service traffic from normal Tor traffic given their special characteristics. As far as we know, there has been relatively little research-level work done to this direction. The main article published in this area is the USENIX paper "Circuit Fingerprinting Attacks: Passive Deanonymization of Tor Hidden Services" by Kwon et al. [0] The above paper deals with onion service circuits in sections 3.2 and 5.1. It uses the following three "naive" circuit features to distinguish circuits: 1) Circuit construction sequence 2) Number of incoming and outgoing cells 3) Duration of Activity ("DoA") All onion service circuits have particularly loud signatures to the above characteristics, but WTF-PAD (prop#254) gives us tools to effectively silence those signatures to the point where the paper's classifiers won't work. 4. Hiding circuit features using WTF-PAD According to section [KNOWN_DISTINGUISHERS] there are three circuit features we are attempting to hide. Here is how we plan to do this using the WTF-PAD system: 1) Circuit construction sequence The USENIX paper uses the directions of the first 10 cells sent in a circuit to fingerprint them. Client-side onion service circuits have unique circuit construction sequences and hence they can be fingeprinted using just the first 10 cells. We use WTF-PAD to destroy this feature of onion service circuits by carefully sending padding cells (relay DROP cells) during circuit construction and making them look exactly like most general tor circuits up till the end of the circuit construction sequence. 2) Number of incoming and outgoing cells The USENIX paper uses the amount of incoming and outgoing cells to distinguish circuit types. For example, client-side introduction circuits have the same amount of incoming and outgoing cells, whereas client-side rendezvous circuits have more incoming than outgoing cells. We use WTF-PAD to destroy this feature by changing the number of cells sent in introduction circuits. We leave rendezvous circuits as is, since the actual rendezvous traffic flow usually resembles well normal Tor circuits. 3) Duration of Activity ("DoA") The USENIX paper uses the period of time during which circuits send and receive cells to distinguish circuit types. For example, client-side introduction circuits are really short lived, wheras service-side introduction circuits are very long lived. OTOH, rendezvous circuits have the same median lifetime as general Tor circuits which is 10 minutes. We use WTF-PAD to destroy this feature of client-side introduction circuits by setting a special WTF-PAD option, which keeps the circuits open for 10 minutes completely mimicking the DoA of general Tor circuits. 4.1. A dive into general circuit construction sequences [CIRCCONSTRUCTION] In this section we give an overview of how circuit construction looks like to a network or guard-level adversary. We use this knowledge to make the right padding machines that can make intro and rend circuits look like these general circuits. In particular, most general Tor circuits used to surf the web or download directory information, start with the following 6-cell relay cell sequence (cells surrounded in [brackets] are outgoing, the others are incoming): [EXTEND2] -> EXTENDED2 -> [EXTEND2] -> EXTENDED2 -> [BEGIN] -> CONNECTED When this is done, the client has established a 3-hop circuit and also opened a stream to the other end. Usually after this comes a series of DATA cell that either fetches pages, establishes an SSL connection or fetches directory information: [DATA] -> [DATA] -> DATA -> DATA The above stream of 10 relay cells defines the grand majority of general circuits that come out of Tor browser during our testing, and it's what we are gonna use to make introduction and rednezvous circuits blend in. Please note that in this section we only investigate relay cells and not connection-level cells like CREATE/CREATED or AUTHENTICATE/etc. that are used during the link-layer handshake. The rationale is that connection-level cells depend on the type of guard used and are not an effective fingerprint for a network/guard-level adversary. 5. WTF-PAD machines For the purposes of this proposal we will make use of four WTF-PAD machines as follows: - Client-side introduction circuit hiding machine (origin-side) - Client-side introduction circuit hiding machine (relay-side) - Client-side rendezvous circuit hiding machine (origin-side) - Client-side rendezvous circuit hiding machine (relay-side) In the following sections we will analyze these machines. 5.1. Client-side introduction circuit hiding machines [INTRO_CIRC_HIDING] These two machines are meant to hide client-side introduction circuits. The origin-side machine sits on the client and sends padding towards the introduction circuit, whereas the relay-side machine sits on the middle-hop (second hop of the circuit) and sends padding towards the client. The padding from the origin-side machine terminates at the middle-hop and does not get forwarded to the actual introduction point. Both of these machines only get activated for introduction circuits, and only after an INTRODUCE1 cell has been sent out. This means that before the machine gets activated our cell flow looks like this: [EXTEND2] -> EXTENDED2 -> [EXTEND2] -> EXTENDED2 -> [EXTEND2] -> EXTENDED2 -> [INTRODUCE1] Comparing the above with section [CIRCCONSTRUCTION], we see that the above cell sequence matches the one from general circuits up to the first 7 cells. However, in normal introduction circuits this is followed by an INTRODUCE_ACK and then the circuit gets teared down, which does not match the sequence from [CIRCCONSTRUCTION]. Hence when our machine is used, after sending an [INTRODUCE1] cell, we also send a [PADDING_NEGOTIATE] cell, which gets answered by a PADDING_NEGOTIATED cell and an INTRODUCE_ACKED cell. This makes us match the [CIRCCONSTRUCTION] sequence up to the first 10 cells. After that, we continue sending padding from the relay-side machine so as to fake a directory download, or an SSL connection setup. We also want to continue sending padding so that the connection stays up longer to destroy the "Duration of Activity" fingerprint. To calculate the padding overhead, we see that the origin-side machine just sends a single [PADDING_NEGOATIATE] cell, wheras the origin-side machine sends a PADDING_NEGOTIATED cell and between 7 to 10 DROP cells. This means that the average overhead of this machine is 11 padding cells. In terms of WTF-PAD terminology, these machines have three states (START, OBF, END). They move from the START to OBF state when the first non-padding cell is received on the circuit, and they stay in the OBF state until all the padding gets depleted. The OBF state is controlled by a histogram which specifies the parameters described in the paragraphs above. After all the padding finishes, it moves to END state. We also set a special WTF-PAD flag which keeps the circuit open even after the introduction is performed. In particular, with this feature the circuit will stay alive for the same durations as normal web circuits before they expire (usually 10 minutes). 5.2. Client-side rendezvous circuit hiding machines The rendezvous circuit machines apply on client-side rendezvous circuits and only after the rendezvous point has been established (REND_ESTABLISHED has been received). Up to that point, the following cell sequence has been observed on the circuit: [EXTEND2] -> EXTENDED2 -> [EXTEND2] -> EXTENDED2 -> [ESTABLISH_REND] -> REND_ESTABLISHED which matches the general circuit construction sequence [CIRCCONSTRUCTION] up to the first 6 cells. However after that, normal rendezvous circuits receive a RENDEZVOUS2 cell followed by a [BEGIN] and a CONNECTED, which does not fit the circuit construction sequence we are trying to imitate. Hence our machine gets activated right after REND_ESTABLISHED is received, and continues by sending a [PADDING_NEGOTIATE] and a [DROP] cell, before receiving a PADDING_NEGOTIATED and a DROP cell, effectively blending into the general circuit construction sequence on the first 10 cells. After that our machine gets deactivated, and we let the actual rendezvous circuit shape the traffic flow. Since rendezvous circuits usually immitate general circuits (their purpose is to surf the web), we can expect that they will look alike. In terms of overhead, this machine is quite light. Both sides send 2 padding cells, for a total of 4 padding cells. 6. Overhead analysis Given the parameters above, intro circuit machines have an overhead of 11 padding cells, and rendezvous circuit machines have an overhead of 4 cpadding ells. . This means that for every intro and rendezvous circuit there will be an overhead of 15 padding cells in average, which is about 7.5kb. In the PrivCount paper [1] we learn that the Tor network sees about 12 million successful descriptor fetches per day. We can use this figure to assume that the Tor network also sees about 12 million intro and rendezvous circuits per day. Given the 7.5kb overhead of each of these circuits, we get that our padding machines infer an additional 94GB overhead per day on the network, which is about 3.9GB per hour. XXX Isn't this kinda intense????? Using the graphs from metrics we see that the Tor network has total capacity of 300 Gbit/s which is about 135000GB per hour, so 3.9GB per hour is not that much, but still... 7. Discussion 7.1. Alternative approaches These machines try to hide onion service client-side circuits by obfuscating their looks. This is a reasonable approach, but if the resulting circuits look unlike any other Tor circuits, they would still be fingerprintable just by that fact. Another approach we could take is make normal client circuits look like onion service circuits, or just make normal clients establish fake onion service circuits periodically. The hope here is that the adversary won't be able to distinguish fake onion service circuits from real ones. This approach has not been taken yet, mainly because it requires additional WTF-PAD features and poses greater overhead risks. 7.2. Future work As discussed in [SCOPE], this proposal only aims to hide some very specific features of client-side onion service circuits. There is lots of work to be done here to see what other features can be used to distinguish such circuits, and also what other classifiers can be built using deep learning and whatnot. --- [0]: https://www.usenix.org/node/190967 https://blog.torproject.org/technical-summary-usenix-fingerprinting-paper [1]: "Understanding Tor Usage with Privacy-Preserving Measurement" by Akshaya Mani, T Wilson-Brown, Rob Jansen, Aaron Johnson, and Micah Sherr In Proceedings of the Internet Measurement Conference 2018 (IMC 2018).

4 5

Next Network Team Meeting Times
by teor 23 May '19

23 May '19

Hi, Monday 27 May is a US public holiday, so we've changed the next two meeting times: May schedule: * Tuesday 28 May at 2300 UTC (Monday is a US public holiday) June schedule: * Monday 3 June at 1700 UTC (The 28 May meeting was at the patch party time) We're back to our regular schedule after that: * Monday 10 June at 1700 UTC * Monday 17 June at 1700 UTC * Monday 24 June at 1700 UTC We meet in the #tor-meeting channel on irc.oftc.net. T -- teor ----------------------------------------------------------------------

1 0

TrackTor - GUI Based Implementation for Monitoring Tor
by Harsh Gandhi 21 May '19

21 May '19

Hello, We are under-graduate students from Malaviya National Institute of Technology currently pursuing Computer Science & Engineering. We have developed a GUI Based Tool for monitoring TOR (inspired by Nyx). We have packaged and uploaded the same on, PyPI - https://pypi.org/project/TrackTor/ Please do give it a try and suggest for improvements. Also, please let us know on how can we incorporate/integrate this with The Tor Project? Thank You! Cheers! -- Regards, Harsh Rajiv Gandhi B.Tech Final Year Department of Computer Science and Engineering Malaviya National Institute of Technology Jaipur - 302017, Rajasthan +91-9983543656

2 1

Network Health Monitoring
by Iain Learmonth 21 May '19

21 May '19

Hi All, I'm working on #28322 to improve the monitoring of Tor Metrics services, but this also has the side effect of monitoring network health. For example, we'd like to know when Onionoo messes up and starts reporting zero relays, but we also get to learn for free in the same check how many relays we have and alert if that number does something weird. What would be the most useful checks to add here? * Range of expected total relays * Range of expected relays with Guard flag * Range of expected relays with Exit flag * Range of expected consensus weight in each position Each one of these is basically an if statement in the script so I'm happy to add these. I can do this by trial and error but if someone has already thought about it then please reply or comment on the ticket. Thanks, Iain.

3 2

Submit to FOCI '19: deadline in ten days
by Roger Dingledine 13 May '19

13 May '19

The submission deadline for this year's Free and Open Communications on the Internet (FOCI) workshop is coming up in ten days. The workshop will be co-located with Usenix Security in Santa Clara in August. https://www.usenix.org/conference/foci19 Here's the blurb: "The Free and Open Communications on the Internet (FOCI) workshop, now in its 9th year, is a vibrant, peer-reviewed event dedicated to bringing together researchers and practitioners from technology, law, and policy who are working on a wide variety of censorship-related topics. This includes direct censorship mechanisms---both technical and legal---as well as indirect methods of censorship through coordinated harassment campaigns and online astroturfing that may be used to drown out and/or chill free expression." We encouraged Susan and Michael, the chairs this year, to make it their own by adding the areas they find most fascinating, and they picked the harassment and astroturfing angle, which I agree needs much more attention from both the research and practitioner worlds. So, you should submit your workshop paper! And even if you don't submit, you should come to the workshop. There will be a bunch of great censorship circumvention and censorship measurement people there, including as usual some Tor people. --Roger

1 0

Re: [tor-dev] tor-dev Digest, Vol 100, Issue 8
by amar memisevic 08 May '19

08 May '19

1 0

My implementation of hash for controller password - torhash
by Damon (TheDcoder) 08 May '19

08 May '19

Hello everyone! I have written a very simple tool called torhash <https://github.com/TheDcoder/torhash> to generate hashed passwords/strings according to the instructions in control spec to authenticate with the controller interface (TC?). I did not actually study the S2K algorithm since I found it hard to understand RFC 2440, I studied source code and bit by bit I figured out what was happening behind the scenes! The reason I did this is that I am working on another program called ProxAllium <https://proxallium.dtw.tools/> which acts as a graphical user interface for Tor, I am currently in the process of rewriting <https://github.com/DcodingTheWeb/ProxAllium/tree/next-gen> it in C to make it cross-platform (along with a few other reasons...). I hit a roadblock while I was implementing support for interacting with the controller interface, I had to make a choice between calling Tor to generate the hashed password or to hash the password in the program itself. I felt like hashing the password in the program itself was the right choice, I had a few other weak reasons to not call Tor for hashing the password but my gut was the main motivator behind this choice. I created torhash (apologies about the bad name choice) as a proof-of-concept and to improve my general programming skills, I had no prior experience working with cryptography or hash functions on this level, so I went with the simplest library I could find for hashing the data. I could have gone with OpenSSL, as I believe that it is the de-facto cross-platform standard for cryptography and cryptographic hashing, I am currently looking into using it, but I am not yet sure if this is a good idea. I would be grateful if some of you can take out some time to have a look, I am very excited to hear your opinions and any advice that you may have to help me improve, the code itself is very simple and short (114 lines of code). Pardon any mistakes or bad code that I may have written, I have only begun programming relatively recently and my experience with low-level development (with C) is very recent. Thank you for reading and for the input in advance! Best Regards, Damon H. (TheDcoder)

2 1

Solving World's Tor Users Being Blocked by Websites (was: Tor exit bridges)
by grarpamp 08 May '19

08 May '19

On 5/7/19, nusenu <nusenu-lists(a)riseup.net> wrote: > > juanjo: >> Tor relays are public and easily blocked by IP. To connect to Tor >> network users where Tor is censored have to use bridges and even PTs. >> But, what happens on the exit? Many websites block Tor IPs so using >> it to access "clearweb" is not possible. Should we allow and start >> using "exit bridges"? In I2P we have not this problem since there is >> no central IP list of relays. > > there is no need to extend to one more hope to achieve this > > https://lists.torproject.org/pipermail/tor-dev/2018-March/013036.html > > https://lists.torproject.org/pipermail/tor-relays/2019-May/017273.html It's possible to augment such outbound solution offerings even further by running an OpenVPN termination service so users can transport UDP between clearnet as well. VPNGate.net project has an idea there too. Even large regional IPv6 pools could be bought and shared by operators and rotated through. More tor relay operators should consider some of the above options, whether as a requested "bridge" service mechanism, or listed in the consensus "contact" field, or as more of a standalone VPNGate support, or "ExitGate" project sort of arrangement. Using only tor right now, a user needs to use a clearnet service that does not scrape consensus, or one not fronted by services doing similar to CloudFlare's spiteful default tor blocking policy, or find a lucky exit within whatever geolocation game the clearnet service uses, or get lucky through traditional vpn or proxy. But those are only fun statistical hacks, not real long term solutions to the underlying problem. It's unfortunate that such braindead blocking, stupid policy regimes, sites refusal to developing creative solutions [1] for so many world's users legitimate privacy, info risk, anonymity needs... often results in users accounts being locked out and escalated into forcing disclosure of users private info and ID to sites to unlock them, thus exposing users to ongoing long term fraud, cost, and stress when that info (in most cases truly unnecessary to collect) is eventually shared misused and stolen by both sites and criminals... or outright auto deletion of user's valued account, built up social networks, etc... all for doing nothing wrong, and harming no one or thing. Death by DriveByExit :( And really shameful to deny world's users the right to simply read a website, be it social, commercial, information, etc or even sadly their own tax-theft funded governmental public sites doing this blocking too. There are some related projects, best practice, as well... https://trac.torproject.org/projects/tor/wiki/org/projects/WeSupportTor https://trac.torproject.org/projects/tor/wiki/org/projects/DontBlockMe https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-access Positive outreach and direct engagement by Tor community is key here, and perhaps not enough of that is happening, at least not publicly. It's a big enough issue that it really needs a dedicated, active, allied, and even funded subproject... a MegaProject that needs to happen. [1] Such as forfeitable cryptocurrency and mailed-in cash deposits refundable in time, increasing account priviledges and features based on account age and activity, community moderation and behaviour support within the sites, opensource third party tracking-free local SecurImage style captcha throughout a sites features, privacy preserving non-SMS non-Google/Apple pure TOTP authenticator protocols, PGP recovery, letting users simply *read* websites without any hindrance, while utilizing these methods only for *write* operations, etc and so many more ways you can envision... Cc'd for awareness and inclusion. *Please remove tor-dev and tor-relays, and move this to tor-talk or tor-access for ongoing discussion and progress. Thanks.

1 0

Controller: text connection vs SNMP
by grarpamp 07 May '19

07 May '19

Speaking of MIBs and management, was SNMP ever seriously looked at back when desire for a control mechanism evolved? If I recall, agent libs and clients weren't wished as middleware, thus demurring to a text connection shell interface. Though commercial routers have both, the shell connection is usually richer and more capable, but harder to parse (Juniper being more programmatic), and requires downstream development to speak to each vendor's shell in automated fashion.

1 0