Molly White

Computer science student at Northeastern University, programmer, Wikipedian, feminist, Twitter bot commander.

Transcript of "Free software for freedom, surveillance and you"

Transcript of "Free software for freedom, surveillance and you," a talk by Jacob Appelbaum at LibrePlanet 2014. This transcript is also available as subtitles to the video, viewable on Wikimedia Commons or Amara. This video was made available under the Creative Commons Attribution-ShareAlike 3.0 Unported license. © 2014, Jacob Appelbaum.


Screenshot of Jacob Appelbaum's webstream

It's really awkward to talk directly into a camera. I spend most of my life trying to avoid surveillance, so... I'm sorry to say I can't be with you in person, so we'll have to do with this video feed. Basically, I guess if John has given a good intro, and I suspect that he has, then you understand that the situation is a little precarious for me, and returning to the U.S. at the moment for me due to my journalistic work with Der Speigel is a little difficult.

So I split my life into two basic parts. One is that I work as a free software developer, and generally as a free software advocate, with the Tor project. It's also the case that I work with Der Spiegel, and also as a freelance journalist with several other publications, and lately I've spent my time doing research into the NSA's surveillance. I've interviewed Edward Snowden, and I've published basically a bunch of information. Stuff that previously we used to think was paranoid crazy-talk, and now we learned that the paranoid crazy-talkers were not paranoid enough.

So I guess for me the goal of this conversation, if we could call it that when I'm just talking directly into a camera awkwardly, is that I wanted to say that the people that are writing free software are actually writing the future. It's a little awkward to say that, but it's true. As an example, I use a free software laptop. It's an X60 and this laptop is, as far as I can tell, about as free as you can get a laptop. It has coreboot, it has Debian GNU/Linux, it is not using any binary blobs, it is, to the best of my ability, liberated from proprietary software. It's pretty good, but it's also many years old, and it's also the case that it's very difficult to do forensics on a machine like this to know if it has been compromised. It is the case, though, that I can do that at all because of free software. It's possible to begin to build something secure with free software, where I can verify and build things from source, where I can look at the source to see exactly what an attacker might try to do with the programs that are running on my computer. So the Four Freedoms here are incredibly important, especially for the work that I'm doing. Now, given the difficulty of setting up this webstream, I think it's clear that I've done a pretty good job of making sure there's no microphones. I removed the microphone from this laptop for example, until I plugged another one in. I've done a pretty good job of that. But it turns out that this is actually really hard for people who have never programmed in C, or for people that do not really understand how to program their home computer and beam themselves into the future.

So, part of what I wanted to do was to inspire some people today to think about what they can do to make this problem of mass surveillance, for example, something that is different. That is, something where there isn't so much hopelessness. It's a little strange, because I feel like, for many years, the free software community is the only one that has really cared about privacy. The free software community builds decentralized or federated solutions and works really on solving problems of the Four Freedoms, but how those Four Freedoms touch the rest of our lives. And so, when I think about what's possible, I think the main problem right now is usability. So, in this regard, to set up this stream, which is broadcasting over Tor, or if you connect to the Tor hidden service which I've also set up, you can watch this video directly, without knowing my location. That was so difficult to set up that it took me the better part of the day to actually accomplish that. It's not that GStreamer is a bad piece of software, it's just that it's extremely complicated. A simple user interface, for example, would have made a world of difference, such that people who are using other video streaming solutions, such as proprietary solutions like the NSA PRISM partner Skype, those people might choose something else. If this was integrated into Debian GNU/Linux, and all you needed to do was download Tails or to install Debian, then it would be the case that you could very easily convince people to do it securely, in a decentralized fashion. That's something that at the moment, I think, is very far away. If it takes me the better part of the day, that means that it will take Glenn Greenwald an infinite amount of time. It will probably never happen, actually, with all due respect to Glenn. When the software is so complicated, it turns out that he'll choose something that's less complicated. And so usability is a fundamentally important way that we can ensure that users will basically care about the Four Freedoms, because once everything they do is working well in free software, why would they choose proprietary software? Usually it is the case that they won't. In fact, there are many people who are Tails users, but they have to switch to proprietary systems, they feel, for the usability of something like Skype.

So, I mean, why? What's going on that all of these things are necessary? I suppose you've all heard the bad news. And the bad news is that the Internet is under, essentially, total surveillance. And what I mean by that is not, let's say, the traditional version of surveillance, where you have a person, and the person is inspecting things, it's looking at things, it's taking notes. Rather, the mass surveillance that is happening now is in the form of deep packet inspection. Now, all around the world, there exists a series of systems that are run by the National Security Agency, GCHQ, CSE, DSD, and other agencies. Those are just the ones we could call "friendly" agencies, if you can call mass surveillance friendly, but you could say it's a kind of social concern, so let's say that they're friendly. Then there's the rest of the world, and what's happening with those systems. We don't know a lot about the rest of the world's systems, but thanks to the courage of Edward Snowden, we do, for example, understand a great deal about the NSA's surveillance. And one of the things that we understand is that it violates the Fourth Amendment. At least on its face it seems clear that it is an unreasonable search and a seizure. That is, it takes data from the Internet, sometimes through fiber optic splitters, things like Glimmerglass-related devices, where they copy information from fiber optic lines onto another line. It then inspects the data, that is, it has seized the data and then it, I believe, unreasonably searches through the data. It does this looking for selectors. So, for example, if you happen to be Chancellor Merkel, that is the German Chancellor, you would have had your phone number as one of those selectors, and anytime any flow of traffic passed by one of these sensors—that is the TURBINE, TURMOIL, Turbulence architecture of sensors—your data would be selected, and that selection would ensure that your data flows that are associated with that would be recorded forever. So if you know about the Bluffdale, Utah complex, you'll know for example that the NSA is building very large computation centers, not just for attacking cryptographic systems, but also for recording data that they cannot currently attack, in hopes that when that encrypted data is interesting to them, or when they have a cryptographic breakthrough, they'll be able to do something with data that previously was just noise to them.

So this tells us a couple things. One of the things that it tells us is that we are in a lot of trouble. There are proprietary software solutions which, according to some of the documents leaked by Edward Snowden and published by Glenn Greenwald, there are proprietary software companies that are what are called SIGINT enabled. That is to say that they believe that when they sell you a security product that it's somehow not false advertising to have willingly broken the system. So that's a really serious problem. It tells us, for example, that proprietary software definitely has backdoors. Unfortunately Glenn did not release the name of that company, or the companies that are involved, but it should be extremely obvious that, if you were to guess, you probably wouldn't guess incorrectly. And if you were to guess, you would probably understand that it's not a free software product. That is, if you look at free software, you can find the problems. If someone were to add a backdoor, while it might not always be obvious if they're very sneaky about it, it is significantly more difficult to add a backdoor to a free software project than it is to add one to a proprietary hardware or software device. So, when we start to see that our security solutions are actually not security solutions overall, one of the things we'll notice is that it's not just the NSA or GCHQ that will be exploiting these systems. It's actually lots of different people. So for example, now that you guys know about the Dual_EC_DRBG backdoor, there is a very good chance that someone in the audience is working on attacking it and breaking it. And if you happen to have, for example, copies of traffic that were encrypted with that, as the seed for the random number generator, you may be able to exploit it.

So this is ultimately a kind of security by obscurity, in hopes that by adding this backdoor, only the good guys, allegedly, they will be the only ones to break it. In reality, it doesn't work that way. So we have a sort of tension here between signals intelligence and communication security. On the one hand, we have signals intelligence collecting as much data as we can from those aforementioned deep packet inspection systems. And on the other side, we have communications security tools which allegedly are protecting us, but it turns out some of them are SIGINT-enabled.

So what is there to be done about this? It seems quite clear to me that free software and free hardware, that is free and open hardware, where things are freely specified, where it's possible for you to fab your own hardware in a factory of your choosing, in a country of your choosing, with parts of your choosing, where you can verify them, that is clearly the right direction to go. Because it is not just, let's say, again, so-called legitimate authorities that are doing this. It's probably a lot of other people as well. We just happen to know there is absolutely, for certain, one set of people that are doing it. And so free software in particular has given, I would say, some leverage. That is, it has allowed regular people to be able to communicate securely in a time of complete and total mass surveillance. So, for example, things that work: cryptography does actually work. If it is not signals intelligence-enabled, that is SIGINT-enabled, it is the case that the mathematics behind DSA, RSA, Diffie–Hellman, those seem to not be broken when you use appropriate key sizes. That's very good news, because in an age of mass surveillance, the only thing that stops the surveillance is making the surveillance worthwhile. That is, if it's extremely valuable for everyone to spy, they will spy. If it's a lot of noise, if the searching for selectors doesn't work, it changes the game significantly. At the moment, not every single byte of data is recorded forever. There is clearly some attempts to do that, and that is something that we need to cope with.

From a cryptographic perspective, we need to think about it, I think, on the hundred-year crypto timeline. This is something that Zooko has been working on and I think is a really good idea. But we have to imagine that all of the things we do are being recorded for all time, and with that, we need to react appropriately. So if you, for example, work on a free software chat client, it should have Off-the-Record Messaging built into it. If, for example, you build a web browser, it should be compatible with the Tor network. It should not be compatible, for example, with the proprietary Flash Player first. Unfortunately, on balance, what we see is that people are often more concerned with making things work with Flash than with allegedly paranoic tools like the Tor project and all of its software, like the Tor network. This, I think, is kind of sad, but I think that we can turn this around and change it, and we need to look at the crypto that we actually use. So, for example, for a Jabber server, it should be using forward secret crypto. So, if you have TLS, it should be used in a forward secret mode. Because it isn't just passive surveillance, though that is a very core and serious thing, it's also active surveillance.

The Turbulence architecture that I mentioned, I detailed a little bit in my reporting at the 30C3 and also in Der Spiegel at the end of last year. And what we see is those selectors that I mentioned before, let's say your phone number or your email address... I wish I could see the audience because I'd ask you to raise your hand, and say "how many of you are sysadmins?" Obviously some of you are, someone raised their hand. So the NSA is probably targeting you, if you're a system administrator for any system that is interesting. And it is the case that if you were to have some credentials for an interesting network, and you are being targeted, one of the ways that you will be targeted is that you will be targeted by an analyst who looks for selectors. They look for your federated logins, they look for your centralized logins, they look for all kinds of information, and it is the case that they program this, essentially, into a targeting system. And that targeting system will automatically attack you. It will do man-in-the-middle attacks, it will do man-on-the-side attacks. It appears that they have certificate authority resources so that they can do man-in-the-middle attacks on SSL, which tells us that we really need to work on things like TACK and SSL cert pinning, that's very important.

If, for example, we notice though, crypto changes the way the selector-based surveillance can even function. If, for example, you note that all of my traffic right now is going through the Tor network, and in many cases people, three people at the moment, are connected to the Tor hidden service for this video feed, there is no possibility for selector-based injection. That is, it's just TLS traffic, the network distinguisher is pretty close to normalized across all Debian Tor users that are doing what I'm doing, which I admit, there's probably five of us in the world, but, you know, that changes the fundamental trade-off. That is, doing the selector-based surveillance will have less of a return than it previously did, or will require more targeting. And this is where anonymity comes into play. If you have anonymity, it becomes significantly more difficult for someone to target you. It's not impossible, of course, but it's... Well okay, maybe it's more than five people. I've got a little lag here from the IRC channel, but apparently there are a few Debian users that route their traffic over Tor in the audience, I hope? But if we see the anonymity benefit there, there's of course a downside, you know? When you have these systems, they're laggy sometimes, they're not always really seamless to work together, it takes a lot of specialized knowledge, again.

And this returns me sort of to the key point, which is about usability. So one of the things that I've seen when looking through documents for this research is that usability and security is the worst nightmare of a signals intelligence agency. So for example, TextSecure and RedPhone, by Moxie Marlinspike, those pieces of software really make the life of someone doing these kinds of attacks hell. At the very least, it means that they have to take what was once a passive thing, where they could silently record data on everyone, and they can essentially now no longer do that. They have to either attack his systems, which they probably will at some point, if they haven't already, or they have to attack each end user's system.

So this means that once we start to deploy this widely, for everyone, it changes the balance, where mass surveillance becomes less and less economically useful for these attackers. And that's important, because again, even if you think the NSA is sent from heaven, if you believe in heaven, you have to understand that they're not the only people out there, doing these kinds of things. So every time, for example, an American businessman or -woman goes traveling somewhere, that person is effectively targeted for signals intelligence collection. Now sometimes it's not always targeted in the literal selector sense, but they're using systems that are SIGINT-enabled. Now whether or not the phone companies know is an interesting discussion. Most of them do. Most of them go along willingly, but the key thing is that, on balance, when you travel abroad, you basically are subject to the whims of all the countries where you are traveling, obviously. Those whims may be significantly more harsh, they may have different economic interests, they almost certainly have different political and legal and economic interests than what you might like. If you're a free software developer, and you're not an American citizen, for example, you are a target, almost certainly. If you are a system administrator, you are a target, almost certainly. And it doesn't even matter if you're an American citizen, really, though there is a slight distinction about that. I think that in the near future, we'll learn that that distinction is largely bogus. Sorry for the bad news, but...

I guess it's sort of important to tie these things together. So, free software and free hardware, they can potentially bring some solutions into play, but it's not exactly clear how we get there. So for example, if you're a system administrator, you probably have a SIM card in your phone. This SIM card is, without a doubt, a piece of proprietary software and proprietary hardware that then plugs into another piece of proprietary hardware, almost always, and that runs proprietary software. And those are actually described, some of those cell phone systems are actually described as master and slave systems, where the master CPU is in fact the proprietary one and it enslaves the free software CPU, which is often Android, which is not always free to begin with. But if we were to say that in an ideal world you had the most free cell phone, you'll probably still have a baseband, which is proprietary software, with a SIM card that is proprietary software and proprietary hardware. So as you're doing your job, even if you have all free software for your laptop, for example, you have this unfortunate conundrum where, to do the basic work that you need to do, you are completely surrounded by proprietary hardware and proprietary software. Now, the NSA has toolkits that they can deploy into those SIM cards, and into the basebands, not just as a matter of exploitation but actually in some cases by design. Some SIM cards, for example, allow you to add an app to the SIM card without even having a cryptographic key or even exploiting the device, just sending a well-formed message will actually do this. Carson Noll1, without realizing it, rediscovered this and he showed this at the CCC. It's almost identical to things that we showed in the Der Spiegel reporting from last year that I discussed at the 30C3. I gave a talk there that was called "To Protect and Infect, Part 2," and I don't want to rehash too much of that talk, but I would really encourage all of you to watch it, because I basically talk about the technical details. And the technical details are important, because in a democracy, for us to be able to understand what it is that we're consenting to, we need to have some concept of what it is we're talking about.

Unfortunately the laws are not so great, and from what I can tell, some people have even called for my prosecution as a result of showing and discussing these things. Now, that went through very careful editorial control at Der Spiegel, so I don't feel too threatened by it, but it's an important point that people don't want you to understand how the machines that control you and surveil you work. They don't want you to be able to change the way that those things work, because it is about power. It's about controlling you, it's about controlling your machines, and it's about ensuring that those people stay in control over you and your devices, should they wish it. So, in a democracy I think it's fundamentally important for us to understand how the machines work, to understand how the power works, to understand what the dynamics are, to make sure that these devices, for example, how they're being subverted, that we understand it. For example, when we understand that exploitation often leads to hoarding of bugs, that means we understand that people are letting us stay in a vulnerable state so as to be able to exploit us. But many people may find those bugs and exploit us, so it is not nearly a simple thing where we say we cede some of our autonomy so that people will be able to do their job. You know, the local policeman needs to do their job, cooperate with them, for example. Because it is every local policeman on the planet. It is every intelligence officer on the planet, with enough of a budget or the technical know-how, who will be able to exploit those things.

So on balance, it seems very clear that we want to build secure systems and not make that trade-off, because we can never actually ensure that the people who are doing this are acting in our best interest, are democratically in a position of authority that is legitimate, if, you know, you bear with me here and we say there is some authority that is legitimate... I know it's a free software crowd, so I like to think that some of you there apt-get install anarchism from time to time. But it's very important to understand that that balance is something which is not a part of the discussion, and a big part of ensuring that it's not a part of the discussion is to try to hide the details. And so for me, I think it's very important to bring out the details. It's very important to show that they use continuous wave generators bounced against reflectors that they've installed after stealing your mail. Do we want to live in a world where these people steal our mail? And where they take our laptops when we buy them online and add devices to them? I don't want to live in that world. I like to think that people that work on free software not only don't want to live in that world, they are actively working to ensure that everyone on the planet can choose to live in a different world.

To get back to the point, things like the Milkymist, Novena, coreboot, these are on a spectrum of free hardware to, as we go down the line, free software-enabled proprietary hardware. Now, I use an X60 with coreboot and I removed almost all of the hardware I don't need. I try, for example, to get rid of anything that would need a binary blob. I think I've done that. This laptop, I think, only has one binary blob left, and that's in the embedded controller for the keyboard. I sometimes use an external keyboard that doesn't have that, although obviously that keyboard has some binary firmware device inside of it. Hopefully not remotely flashable. As far as I can tell, that's the case.

Systems like the Novena, made by Bunnie, I think, are the future. That is, he has built almost entirely, as much as is possible at the moment, an open hardware device based on an ARM CPU, where you can fab this device, where you can very easily, if you are a hardware person, modify it, and you can fab this changed device. So that, I think, is critical. There's still a proprietary CPU, but there's a trade-off to be made here. So in this case, the Freescale CPU that he included is pretty fast, it's a quad-core CPU, and it has a hardware random number generator. Who knows if it's SIGINT-enabled, hopefully not. If we see this, we see that it is significantly better than, for example, the ThinkPad that's sitting in front of me where we don't know the designer, we don't know their intentions, we don't know for example if the Intel microcode, if it can be updated remotely by someone who has the key that isn't Intel. Probably, would be my guess, if they understand the format, if they can add backdoors, which, if they understand the way the microcode works and they have the key, then of course they can do all of those things. So there are some architectural changes in the Novena which I think are pretty spectacular for that.

If we go all the way, I see something like the Milkymist, which for a time I used instead of as a video mixer but as a machine for running screen and irssi. And it actually is a FPGA device where the CPU itself is free software. Unfortunately the tools for synthesizing the FPGA, those are not free. We lack free software tools for those things, as well. And we really, really, really need free software tools for all of these things, and we need free hardware platforms to build on top of. Without that, it's very difficult for us to secure our systems. I think that it's critical to do that. And there are some people that are doing that. So lekernel, the guy who's working on the Milkymist, and some of these other free hardware devices, I think he really needs support, and I think it would be great, because his devices are the kinds of devices where you can do forensics on it, but you can also prevent adversarial forensics. That is, you can program your device to self-destruct, but you can also check to see if someone has changed the bootloader, if someone has changed the VHDL output, you can actually verify these things. Novena is much the same. I've been working on making Debian GNU/Linux run really well on it. I actually have a Novena here in front of me, with the little helpful Intel sticker, but that's just as a joke, because there's no intel inside with free hardware and free software, if we do our jobs correctly. So, I'd like to think that this is a good start, but the only thing that makes this useful is, of course, the free software on top of it.

So in this sense, I think that Debian GNU/Linux is very important. I think lots of things, like Trisquel, for example, are very important. And we need to work on making those systems usable. I think the GNOME project has done a very excellent job with that. There are a few things about it that drive me crazy, but mostly just because I've been using computers long enough to have bad habits, so I think that that's a sort of humbling experience. The Tails operating system, which is a derivative of Debian, is set up in such a way that you don't need to understand anything about anonymity, you don't need to understand anything about security and privacy. It comes with a chat client that has Off-the-Record Messaging by default. Everything is configured to work over Tor by default. This is great. This helps us with this paradigm shift of privacy by policy to privacy by design. Now, there are a couple of problems that still exist. Even if you have Tails, even if you find it usable, which it isn't, in my opinion... It's much more usable than all of these things were before Tails, but it's a progression. If we take a step and we go further and further down the line, one of the things we'll note is that there isn't an easy wizard for setting up, for example, a chat account that just works. Where you can just easily send a message. Where it's not hard.

As an example, I installed Jitsi on this laptop, and at the bottom of the screen you see "surveillance_target@jit.si". You can, of course, ask me questions via OTR if you want here, and if you'd like to, you'll note that it is, of course, using OTR, it is also using SSL/TLS... well, it's TLS 1.0, I think, to connect there. You'll note that it's essentially... it's like an email address. That is pretty good, and I think we may have to sit with that for a while, while we work on coming up with different naming systems. Zooko's triangle, if you haven't heard of it, I would recommend you look it up on the Wikipedia, it's definitely the case that we have some hard problems to try to tackle. Skype, for example, is really easy for people to use, and they do use it because they have probably purchased every webcam in the world, and then they've made it work with every piece of software and hardware combination that they could get their hands on. And they also solved the problem of making it as simple as adding, for example, one username and password, and then forever just keeping that identifier. And then you just have to simply say, "Hey, I'm Alice at, I'm Bob at" and it's really easy to use. For example, if Jitsi had a setup wizard, where it automatically generated your OTR keys and it automatically did everything that needed to be done and it added you to their server, but it also allowed you to configure a secondary server, that would be fantastic. It would mean that you could just download Jitsi and it would work. If it forced OTR, even better. Those kinds of things are really simple changes that really would make a world of difference. And since Jitsi works on other platforms other than pure free software platforms, it can even be used as a kind of gateway crypto system. So, I think that kind of stuff becomes very important, because once you have something like Tails and you have something like Jitsi and you put them together, you still have that last step. And that last step is a hard one. But we're really close.

In looking through and understanding some of the things that I've been looking at and studying with regard to the technology, and I said this before, usability and security are absolutely critical. But I also mentioned before the active attackers. So one of the issues that we see is that these active attackers are actually pretty good, right? So if you have a longterm cryptographic key but you don't use it correctly, that is, you encrypt all your traffic in a non-forward secret way, you have a pretty serious problem, which is that these people will break into the computer and actually take the key so that they can decrypt traffic, or to impersonate you. So we do need to come up with some notions about, for example, ratcheting. So TextSecure, which Moxie Marlinspike has been working on for many years now, it has this notion of forward secrecy as well as future secrecy. I think that we need to think about some of the stuff that he's been working on to make this kind of system usable. Adam Langley, who is probably, I would say, one of the great living cypherpunks right now, he has written a system called Pond. I've worked a little bit on the key agreement system PANDA, where we dynamically meet by using a shared secret. That kind of a system is really important. It changes the game from "you're totally surveilled, but maybe you have PGP email if you're lucky," or "you're totally surveilled, but maybe you have OTR if you've had a proper chat client for it." And basically it switches it, so everything goes over Tor, everything goes to a server, that server only sees a delayed set of messages. That kind of system is really a significantly different way of doing communications, and it's not so different from email in some ways. But it's different in the ways that are really important. So that, however, is completely useless without a usable interface, without having it be deployed, without teaching people things.

So as an example, when you teach a journalist something—one of the things that I've found is almost impossible to do is to teach journalists—but if you teach journalists something, you want to teach them one or two things that you absolutely impress on them that they must do. So for example, verifying an OTR fingerprint is one of the things that I impress on every journalist that I work with. And that's actually my litmus test. If they cannot use Tor and OTR and Jabber together, I don't work with them, because they can't maintain confidentiality, authenticity, any kind of integrity in a digital sense. And that, for me, is important. But it's also the case that I spend a lot of time teaching people how to build Tails disks, or building Tails disks for them, or getting a special laptop and then setting that up for them and then adding Tails to that system, or a Debian GNU/Linux setup. And it would be nice if it was as simple as "just use Trisquel." If it was as simple as "buy a laptop from this vendor and it's all free software enabled." Now there are some vendors that do it, and the problem of interdiction, which I mentioned before, where they steal mail and change things, that's a serious problem. So we really need to make sure that these things are available in regular stores where regular people go. That makes targeting, again, much harder.

I have a couple of other things before I start to take questions, but I think that the real key stuff that we need to consider is that we need verifiability. So for example, reproducible builds, things like Gitian, which is what we're doing for the Tor browser, that's very critical because anytime you build software for anyone, you are a target. Especially if the person you build software for is themselves a target. So, for example, we can imagine that with this three- or two-hop-out idea of targeting, we know that it's possible for someone to target a sysadmin because the sysadmin is interesting, but it follows that if the sysadmin is really good, and a lot of people in the free software community are really good with securing their systems, those people will be targeted by targeting their operating system vendor. So, for example, we know that Debian has been compromised in the past. We have to make it so that compromising Debian is not only detected, but that it doesn't make sense to do that. So if, for example, you were to think about this from the Gitian perspective, if we have anonymized builders that are regularly building packages and reporting those results, it will allow us to see if someone has changed a package on the server, it'll allow us to tell if that piece of software has been tampered with. We need to have a kind of binary verification process which, at the moment, we don't really have a binary verification process. Some people use proprietary software like IDA Pro to reverse engineer this, some people try to disassemble or decompile the software to see if it matches what they thought, but that's usually a hand process. We need to automate some of those things, and we need to do it in a way where people are able to report back to the community anything that they see, basically in real time, that is problematic.

So for the Tor browser, I actually have a machine that builds with Gitian, and if it ever builds a hash that doesn't match what I would expect, then it alerts me. And it alerts me in a way such that it just looks like any other person downloading the source code, so it's harder to target, and it is the case that it verifies signatures where I've verified the keys in person. So I think that's pretty good, I think it works relatively well, but that doesn't scale. And right now, it's a one-off. So we really, really need to consider this as a reality.

We need that also for devices. For example, if I buy a hardware device and the NSA adds something to it, what has happened? Right? Well, usually, they've won. That's a really serious problem. And it's not just the NSA, it's anybody who can steal mail, and especially at scale. Right? Computers go through customs. Free software might not always go through customs, but computers do. And that's where the state's advantage is often used against people in a way that they don't understand and certainly wouldn't consent to. And so we need to have a way, really, and this sounds kind of outlandish, but let's go for outlandish. We need a way to be able to X-ray our hardware and compare it with a known good state. And with no binary blobs, it becomes a little bit possible. A little more possible for us to make sure that the systems we're carrying around are not just bugs for an oppressor. They're not just systems to be used against us. Now, I know that that's a tall order, but the GNU project itself is a tall order. And so we need to move towards free systems: free hardware and free software systems for freedom.

Because really, it would be very difficult to maintain freedom and liberty in the future, and even, I think, to keep our democracies in a world of mass surveillance. Especially if all of our devices are the thing that is oppressing us, or that are acting as an oppressor. In the past it was the case that you had a neighbor, and the neighbor maybe received some benefits. Now the changes are different. Now people report on each other as a matter of, you know, fun. For society. With Facebook, for example. Well, what happens when the Philip K. Dick nightmare is not just worrying about every person spying on you, but what if it becomes every thing that's spying on you? Part of the way that I deal with this is I literally remove the physical microphones from my computers, because I know that it is almost impossible to secure machines such that a really powerful, well-funded adversary could enable them again. So that is not really something that scales. But we can think about it when we build free software laptops, we should make sure that there is an LED that if the microphone is powered up, the LED is on. Just the same way with a camera, but not as badly-designed as most cameras.

I guess probably now, we should probably take some questions, given the timing. Is there anybody that wants to ask a question? If so, we're in #libreplanet on OFTC, because OFTC allows me to use Tor to connect to their IRC network, which I'm very thankful for. I would be happy to take some questions, and I know that some of you have contacted me on Jabber. So you can of course, you can of course ask me questions.

The first question says, "What is, in your opinion, the most important technology for journalists to learn?" You know, it depends on what kind of journalist. I tend to think that the key technology for people to understand is not a specific technology, but rather the philosophy of free software. I mean, Richard Stallman, who is one of the most brilliant people to have ever lived, really hit the nail on the head when he talks about free software not as a matter of cost but as a matter of freedom. And I think that when people understand that power dynamic, when they understand the tradeoffs they're really making, they'll change the pieces of software that they use. And hopefully, by the time they make that choice, that software will be usable, so that when you use Jitsi, for example, it does not allow non-OTR conversations. Or when you make a video call, it doesn't allow the user, basically, to make an unencrypted stream without jumping through hoops. By default it is secure. By default is privacy by design. And so, if people are going to learn one specific tool, I feel like we're sort of failing. I think, as Schneier is often quoted as saying, privacy and security is a process, not a product. Maybe he only said that about security, but let's modify it a little bit. Right? Liberty is also a process, it's not a product. So journalists need to learn about the world around them, but that's for every person as well. And so when we want every person to have this, we need to make sure that the devices and the software that we use actually enable that by default.

The next question is, "How useful is a SHA-256 sum for checking software binary zip integrity checking?" My feeling is that hash functions are not going to be the weakest point. But if you would like, you can take the approach that Debian takes. When you upload a package, it has MD5, SHA-1, and SHA-256, and then you do a GnuPG signature over that. Now I use, because of the fact that I'm certain there are people that are trying to attack my systems, I use this, which is unfortunately not completely free. But it is a GnuPG smart card. And that GnuPG smart card, I also don't leave it plugged into my system very often, and I usually use it on an offline machine so someone has to break into my house to be able to even begin to mount an attack on the smart card. But that, I think, is really the way to go about it. It's not just about hashing, it's also about ensuring that you compose those hashes into a system that makes some kind of sense.

But again, in this case with Debian, when you hash the files and you upload them, the binaries I built on my system are the ones that Debian gives out to users. Is that really what we want? I think that's a bad idea. What if my system has been compromised, right? We don't want that binary going out. And there's some work on changing that. But if we think about it just in terms of hash functions, I think we'll rarely find the hash function is the issue. Obviously there are some things, like MD5, that are just hopelessly broken, so we should be moving towards things that are not hopelessly broken, but it's difficult, because a lot of our standardization agencies, they're not very good at their job, in my opinion. Right? When NIST collaborates with the NSA willingly or unwillingly, wittingly or unwittingly, I think we have a problem. So we should look for diversity in this, and not just choose one thing, but choose a few things that make it significantly harder for someone to attack any single thing.

And as far as average users being able to verify software, I think this is a really tough problem. Basically, the real issue is a bootstrapping problem. We need to make sure that operating systems have some notion about actual integrity of packages. And that's a really difficult problem to solve because many people start with a proprietary software platform, like Microsoft Windows or Mac OSX, and those platforms, they do not respect peoples' liberty. And naturally, they don't want to help you to move to a new platform that respects your liberty. So, in a free software world though, we should be able to have packages that do have verifiability in the operating system, as well as in the packages. That is, that are signed, that are hashed properly, that have some notion of the web of trust, or something that replaces it, plus a user interface that makes sense. And that's a really difficult one. Snowden calls it the "Greenwald test". And I think that that's a good test, actually. As someone who actually asked Glenn to use a bash shell on Tails and showed him how to use a bunch of command line tools, allow me to elucidate how important that test is. Jesus Christ, that is a serious test. It's really, really, really hard to get Glenn to use those tools securely. But it shouldn't be. In fact, every time that a user can't figure something out, we should say to ourselves that we have failed. Not seriously, but we should say to ourselves that we have failed, and we should try to succeed where we have failed before.

I have another question here: "What are the bare minimum fundamentals we should teach the general public when advocating privacy?" Well, I tend to think that the bare fundamental is that we're not talking about privacy, actually. We're talking about autonomy, we're talking about dignity, and we're talking about our liberty. Privacy is merely one of the manifestations of this. So, for example, when people say that they don't have anything to hide, it's not about hiding things. It's about having a private sphere in which to think about things before you reveal what you have decided. Where you don't have to reveal the process by which you make a decision. But also, where you get to choose. It's not that, for example, I have nothing to hide underneath this great Cyberpeace t-shirt, but it should be me that actually chooses if I should take it off. And so, for example, right now since I understand there are sixty people watching this webstream, I'm going to keep it on. And that is, when we talk about privacy, in a sense when we say that privacy is dead, what we're hearing is our modern generation saying that liberty is dead. And I don't like that. So I refuse to say that privacy is dead when people really mean liberty. And I think it is important that we reject that notion, and we should talk about how we should have a right to autonomy, we should have a right to express solidarity, we should have the ability to be able to, in the case of free software and free hardware, have devices that actually empower us and that we understand how they work. And we should be able to be secure, end-to-end secure.

So, I've got a couple other ones. Oh, wow, joeyh! One of my favorite Debian developers of all time, that's incredible. I feel honored that you're asking me a question. "Should Debian work towards integrating Tor more?" Yes. So I'm a new Debian developer, it took me ten years, because I'm slow at becoming a Debian developer. But I'm error@debian.org, and I'm super happy to help anybody to be able to integrate Tor and anonymity software by default into Debian. As an example, I have a transparent Tor network that I use to be able to ensure that I can install Debian on new machines without my Internet service provider being targeted by the NSA or other people. I also run a Tor mirror on the Tor hidden service, as well, and a Debian mirror on the Tor hidden service as well, so that I can install packages on these systems without having to worry about basically being attacked. Even if a Debian developer FTP Master's key is compromised. So that a targeted attack is significantly harder. As you can imagine, that's not very usable for regular people, and as you can also imagine, it probably doesn't work very well. So yeah, we should make it so that a Debian user can say, "help! I'm a target of surveillance, and I'd like to be able to use free software without being tampered with." And that would be great if we could make Debian, if we could make Debian more friendly to that. Because basically we, for a long time, have lived in a world of privilege, where we thought we were exempt from the power dynamics of the world. And I think one of the things we will learn, especially with Debian, is that that isn't the case. And the more international a team is, the more the legal authorities of intelligence agencies suggest that they are fair targets. So, that also extends to the users. So I'd love to make that happen. And, yeah, wow. It's incredible to be able to talk to you guys here. Probably the only group of people that really can make these changes, right? I mean, there are other free software people around the world other than the ones in this room, but it's really critical to understand the role that you guys play. And that all of us play, together.

I have a couple of other questions here. Helican1 asks, "Do you think the time is right for a free hardware FreedomBox with Tor built in, and do you have any news on the FreedomBox front?" Well, I was very depressed about some of the discussions around FreedomBox for a while, where I felt like people were taking anonymity as a sort of, like a luxury good. And they felt like we didn't need Tor, or something like that. Now, obviously, I work on Tor and I'm paid to work on Tor, so I feel like it's a conflict of interest for me to say this, but yeah, I think we of course need to do that. At the same time, the reason that I work on Tor is because I really believe it. I think Tor has probably saved my life a couple of times every month for the last several years, from military dictatorships to other places where I've traveled. So I think it's critical to make that possible. So as an example, this device I held up here, this Novena board, well, as you can tell, the basic idea is to have a device that is free hardware and free software that, you know, gives you exactly what you've just asked for. And that's in fact what I have. That device, when I plug it in, it sets up a wireless network that transparently routes people through Tor, and it also sets up a Tor relay so that it will relay traffic for the rest of the network. And it's entirely powered by free software with no proprietary software at all. I think that that is, yeah, I think that's a good thing to do, and that's what I've been spending my time doing lately. If you want to help with that, it would be great. And if we can get the FreedomBox to adopt the Novena board, I think that that is great. I think we should try to raise a million dollars for Bunnie so that we have a free hardware solution, or open hardware solution, that actually is usable, that's fast, that doesn't support a company that doesn't care about our liberty, but instead supports a developer who really does care about our liberty and about our freedom.

The next question from Malapart1 is, "Is .onion today what SSL was in the mid-90s?" I really hope not for a whole bunch of reasons.

So the next question... ah. So someone in the audience wants me to elucidate on the link between anarchism and free software ideology and goals. I think that it's important to not focus too much on that, in particular because I think that sometimes talking about anarchist philosophy alienates people, because they think that anarchism is the same as complete chaos, or synonymous with violence. And so I'd like to sidestep that and say if you apt-get install anarchism, literally, that's the Debian package, you can read about the philosophical texts of anarchism. But the basic idea of anarchism is about mutual aid, it's about solidarity, it's about respect for human rights, it's about the same things that the free software movement are about. But there's a lot of propaganda out there about the notion of democracy, in fact, in the form of anarchy. And that's unfortunate, actually. And I think we can change that. One of the ways that we can change that is to actually have propaganda of the deed. In this case, making free software and free software available to everyone means that people understand the fundamental tenets of anarchist philosophy in their everyday life, and they don't have to learn about the philosophy too much to be enabled by it. And if they want to, they can learn about it. I think that's a very powerful way to make that happen, because it's very easy, for example, to talk about it philosophically, but until you have a tangible thing, it's not really clear. That is, when we didn't have an anonymity network, and people said, "Well, do you really need anonymity?" you would make a different choice than if you have an anonymity network and it will be taken away from you if someone says "do you need anonymity?" and you say no. So when you say yes, and it's there, and it's tangible, it changes it. And the same is true for anarchist philosophy, and the same is true, I think, for free software and free hardware. And especially when these things work together, they actually help us to build autonomous communities, they help us to build secure systems across hostile networks. I think that that's very powerful, and I think that the way to get people to care about that is actually to show them that. People care a lot more about connecting now that it's easy to do. It wouldn't have been a relevant question thirty years ago in the way that it is a relevant question now. So the freedom to connect, the freedom of free hardware and free software, these tie fundamentally into anarchist goals. I'd really encourage people to look up the works of Emma Goldman, for example. I think she's one of the greatest feminists to have ever lived, though most people don't know who she is. And if you looked at the Wikipedia page, for a time there was no mention of her on "Important feminists of the twentieth century"2, I think is the page I was looking at. I think that is a bit of a shame, but I also think that that's, you know, an easy thing to reconcile with reality if people go and they look it up and they study about it. So hopefully people will install that Debian package and otherwise learn about that if they are interested in it, but really, I think, getting people to have the values that are embodied in that is just as important if not more important.

The next question is "What about those countries that are completely blocking projects like Tor? What should we do to help people in those countries?" I tend to think not in terms of charity, or helping people, but rather in terms of solidarity. So, the Tor project, for example, is not having a war with China, right? China often does not respect its citizens' autonomy, and blocks its access to the Tor network. One thing that would be helpful would be to make it so that, by default, a lot of applications use Tor so that the so-called collateral damage, though I'm loathe to use that term, becomes higher and higher, becomes more difficult. And also, such that people start to use what are called pluggable transports, like the obfuscated proxy obfs3. The proxy actually, right now, is not blocked in China. It becomes a sort of cat and mouse game, but it may be the case that as we build more difficult-to-classify protocols, as applications understand that sometimes the Internet does not respect your autonomy and wants to tamper with it, that will be something that will change the dynamic about how that blocking and that arms race works. We have another project, ooni.torproject.org, that's the Open Observatory of Network Interference, or we used to call it Open Open Net. That is a free software tool that we've been working on for several years now, to be able to look at censorship and surveillance. You know, censorship is a second-order effect of surveillance, so this tool, while at the moment not the most usable tool, it actually allows you to diagnose, understand, and share the data. So one thing that can really make a big difference in these topics is to actually share the data. Right? Once we start to study and understand these things, especially the techniques of censorship and surveillance, it allows us to change not only how the networks work, but how societies work around those networks. So, I think studying that can be useful. If you want to be a Google Summer of Code, I know that's kind of ironic, but if you want to be a Google Summer of Code student to work on one of these projects with the Tor project, I think we still have another day to apply for that, though I'm not totally sure about that deadline. If you just want to come hack on free software with us at the Tor project, we have a bunch of projects that are like that.

Some other questions. It looks like... Yes, the Novena router does not have AMT, that's right, so it does not have a built-in backdoor, which is, I think, nice. At least not one that we know about.

Any other questions here? Ah, I see, there's like twenty. Alright. "Could you please elaborate on the idea or concept that it is required a critical mass of privacy-minded users to create enough obfuscation for making discovery schemes like Tor network node spying to be unlikely?" I don't think we're going to make it unlikely for spying to take place. What we need to do is change the economic balance, and that may allow us to move into a world where mass surveillance of our intentional communications are much too expensive to do for everyone. And this is important, because if you have a phone, and you make a phone call, people think of surveillance and conceptualize surveillance as surveilling your call, but there's all the unintentional data that you leave behind. All the towers you visit, and so on. That stuff is also, unfortunately, a huge target of mass surveillance. So even if everyone's using something like RedPhone, we still have these little spy devices in our pockets. Even if it's free software enabled, the networks themselves are harmful to privacy. That said, we do need a lot of people using this stuff, because the more people that are using it, the more likely it is that it will stick around. If it's just people like Edward Snowden, Julian Assange, Glenn Greenwald, Laura Poitras, or myself using it, yeah, I mean, that's a problem. Not only do we stick out on the network, we have, well, basically it makes it much easier to target, and it also makes it possible for people to try to ban that technology, whether that's by DPI or by legal methods. That is a really serious problem.

Okay, I think that we should wrap up here soon, it sounds like. If there are any other questions, I will take them, but otherwise I think I'm going to end the stream. If you'd like to hack on free software with us, and anonymity-related stuff, I'd be happy to talk with you. I will never use this Jabber address again, so that your social graph is not tainted by mine. If anybody wants to send me an email, you can send a mail to jacob@torproject.org, or if you'd like to, for example, chat with me on IRC, I'm ioerror in #libreplanet for now. Thank you so much for the honor and privilege of speaking with you, I'm really sorry that I cannot set foot in my own country right now, to be able to speak with you in person. But thanks to free software, I am able to speak with you. So, thank you so much for making that possible, especially to all of the free software developers in the room that actually made that possible.

Thank you very much for your time, and I hope to meet some of you again, someday, in real life. And remember: if not, it was murder.


1. I did my best to verify spelling, but I was unable to check the spelling of this.

2. I'm guessing that he was referring to the article "List of feminists," which has a "Mid and late 20th-century and all 21st-century feminists" section.

comments powered by Disqus