On Being a Citizen Hacker

something that i call being a citizen hacker and it's kind of a term i invented last night um it's a play on um uh the term citizen scientist and i i kind of despise the term citizen scientists because it kind of implies that people who are citizens can't be scientists and it's usually used to refer to things like having school children go and do experiments and it's kind of like you're being this consolation-prized version of a scientist because you're not an official scientist you're just a citizen scientist but in fact you know as soon as you start doing science you're a scientist and i'm an official scientist so i get to say that and in fact i get to declare

who is a scientist and that's everybody who practices science but what i think about when i think about a citizen hacker i want to kind of invert that we're all hackers everybody in this room uh is a hacker i mean we're a little suspect for being in a room before noon but uh other than that um you know we're you know we're definitely hackers but i want us to also think of ourselves and the hacking that we do as being in the larger context of our responsibility as as citizens as you know however you may define it and what i'm going to talk about in this talk is the way i've defined that for myself

and hopefully to encourage you to think about some of the same things and and maybe join me or or do something similar um on your own so a little bit about my background um so in 1993 i finished uh my degree i you know finally did everything you could do with school and i landed at um what i i would have to say is the best job ever which was at bell labs and bell labs was a magical place bellev's research in 1993 and that era because essentially it was the research arm of the telephone company and uh it's the telephone company basically didn't you know had a lot of extra money because it was a big monopoly

and uh a lot of the way it budgeted for things was based on well we budget a percentage of what our expenses are so the higher our expenses the more we can make and so they said well let's build a big cost center and they built a research lab with um you know great people who were basically given the freedom to do um what they wanted um that unfortunate and if i could go back to any job that i've ever had or could imagine it would probably be you know 1993 at bell labs um in 2004 after uh you know at t kind of came to its senses about research i joined the faculty at the

university of pennsylvania in the computer science department uh and much to my surprise that was about half the salary and twice the work and also uh the best job ever um it was is one of the nice things about being a professor is you're officially defining your own destiny and you nobody can really tell you what to work on and that's an amazing freedom and an amazing responsibility to have uh and then in january of this year 15 years after i became an academic again i got the another best job ever but also one of the strangest jobs ever georgetown university which is down the road in in washington dc um basically said we want to give you a really weird job

we want you to join our computer science department and we also at the same time want you to be a law professor in the law school and they noticed that about half of my recent papers and publications have been in law reviews rather than just academic computer science oriented journals and they wanted somebody who in the law school who could really speak to the detailed issues of technology and its larger impact on law and public policy and related things so i managed to you know kind of hack my way into being a law professor without ever actually having gone to law school um and you know and i mentioned to people hey i'm you know i'm joining uh

the georgetown law and they said oh well that that's great you know let me give you some advice on your first year classes and uh no no no i'm i'm gonna be a law professor and they said oh really um that's uh um how'd you pull that off and i'm still wondering um and uh so i'm now officially um i think my job is now a citizen hacker i i want to understand really as my day job of the larger impact of the work that i'm doing rather than as something that i have to kind of uh hide from my uh employer and so i'm going to spend a little bit of time talking about some examples of

this and then i'll do a deep dive into kind of a current uh an ongoing uh project that i think you should um be thinking about your own uh devoting your own talents uh too so uh just a few highlights of of some citizen hacker work that i've done in 1994 uh a year into my a little less than a year into my job at uh bell labs at t uh the u.s government almost immediately after i started there proposed a system uh for encryption to replace the data encryption standard which was the uh 56-bit uh secret key algorithm uh that was a standard for the us government as well as the de facto standard for encryption

used uh in the private sector and you know it was pretty widely recognized by then that des was showing its age 56 bits is not a big enough key space in practice it would soon be vulnerable to exhaustive search and so the people were clamoring for a stronger secret key cipher algorithm and at the same time parts of the government particularly law enforcement although the intelligence community probably carried most of the public water for it but it was probably really driven by law enforcement uh was worried that uh encryption if it proliferated and they used the same language used to talk about nuclear weapons when they talked about encryption technology that if encryption proliferated it would

shut down law enforcement's ability to do things like wiretaps and um this manifested itself in all sorts of ways that were very very harmful to our security we're kind of still paying the price for them now in particular the you official us government policy was to discourage the use of encryption in the private sector and there was no actual direct law saying you can't use encryption but what they did was a kind of interesting backdoor around that which was if you built a product using encryption you couldn't export it without an export license that they basically wouldn't give you if your encryption was strong and so that ended up being a de facto um control on domestic use of encryption

um because you know nobody wants to build a product that you could only sell in in one country even if it's a large country like the us um and so this made kind of everybody unhappy and in particular it was happened at a really terrible time because the early 1990s were basically when we built all of the standards and all of the infrastructure that we're now using today for things like the web and for almost every part of our infrastructure all of the standards and all of the basic platform assumptions can be traced pretty directly back to that era and in that era adding encryption to something was not something you would just do naturally

because of course you would do that it was something you really had to think about because it had all of these implications about exports and so we ended up having things like you know the cryptography option and you could turn crypto off or downgrade to weak cryptography and if you look at attacks that have been found in systems that use cryptography it's very rare that it's the algorithms that are uh at fault but rather you do things like downgrade into the non-encrypted option and we're still paying the price for that today so in 1993 the government proposed a new crypto system to resolve this and it was unbelievably terrible in almost every aspect of its assumptions um it

used a new cipher or new publicly new cipher that nsa had apparently had sitting around for a decade or two uh called skip jack and this had an 80 bit key much better than 56 bits um so that was probably you know that would probably be good enough to get us to roughly you know around now um so you know a couple decades worth of of resistance to exhaustive search that's great and then it just went off the rails it's a first of all could they wouldn't tell you what the cipher was it was classified it was classified at the secret level so you could only get it in tamper resistant hardware which means that

the cost of including encryption in something included buying special purpose hardware uh in this chip called the clipper chip and that wasn't the worst part um the worst part was that this trip had an extra feature in it which is that as part of the initialization vector string uh instead of being 64 bits long it was actually 192 bits long and the in these extra 128 bits it would embed an encrypted copy of the session key uh encrypted with keys that would be held by the government so that if they wanted to decrypt your traffic they would just intercept the initialization vector and then um they could decrypt the session key and then decrypt the traffic

and that naturally made everybody really happy um yeah um you know because the uh you know all you have to do is completely trust the united states government and not not just trust them to um not just trust them to have honest intentions but to trust them to actually safeguard these keys and we all know that data breaches never happen on any kind of a large scale um so you know whether you trusted the intentions of the us government or not you also had to trust the ability of the government to maintain a secret database of these keys it was it was just a horrible mess from a policy point of view but to the government's great credit

none of this was hidden none of this was done surreptitiously they were out in the open about this and at t um the part of att i didn't work for the part of 18t that tried to make money rather than spend money saw this as an opportunity to build a product around this and they built a device called the att telephone security device model 3600 um which you know 18t was great at marketing names um and this was essentially a civilian version of the stu-3 encrypted telephone it was actually made by the same people who made the sku3 at at t and it would incorporate the clipper chip this this this key escrow chip and att

you know said we want to be first we want to we want to be ahead of the curve on this and they were so far ahead of the curve that they were actually the only company that built a product around this device and they kind of hurried this out to market and as a result of att's involvement and things like that nsa who designed the system uh came up to at t research att labs to give a presentation on it and so i naturally said oh i want to make sure to be at that i'm really interested in this and um you know so they described how it worked and you know casually at the end of the

meeting you know i i went up to one of them and i said yeah you know if there's any way to get a sample of this i'd love to you know get get one thinking they're just going to blow me off and they said sure come down to fort meade and uh um so i showed up a couple weeks later at fort meade and i spent the day in a in a secure room with them uh lights would go off and flash anytime i would walk in the room an unclean visitor in the room you know hide all your good stuff um and people would be erasing their whiteboards and then i walked out with a couple of uh of pcmcia

card form factor uh clipper chips uh on it uh and by the way getting hardware out of the building at nsa um isn't supposed to be that easy and you know so they kind of realized this as i was leaving oh you're gonna have trouble getting that out the door here stick it in this bag um and uh um so you know i went back on my on my way to new jersey with my my clipper trips and and you know i thought okay how do you how do you use this uh and so i said well well first let me see how this key escrow feature works how does this how does it enforce

this uh you know could you defeat it could you use this better cipher without exposing yourself to the uh law enforcement access feature and what i discovered was that the obvious ways the most obvious ways to defeat this not send it didn't work it had built-in safeguards that would check to see whether you were sending a valid um field um before it would uh allow the receiving chip to actually operate but then i discovered well how does it actually do that and i discovered basically the second most obvious way of doing this did actually defeat this and that was to exhaustively search on possible valid law enforcement access fields until you'd find one that matched the

built-in checksum that's under this encrypted field because it was a little longer than 80 bits and so basically with with 2 to the 16th effort you could find a forged leaf field that would be happily accepted by the person you're talking to um and you'd be able to encrypt away but it would actually be useless because it was just randomly generated for decrypting your your traffic and then at that point you know here i am i discovered this thing i'm a scientist type person and my natural idea is to you know publish a paper um and so i wrote this paper up and i circulated it to some colleagues and you know eventually it made its way

to the part of at t that made the product around this and that valued its relationship with the government and they were um and they were uh less happy than a a lot of my other colleagues were you know and they pointed out quite correctly you know that uh you know if you publish this we will cut you um you know and uh um and to the great credit of research arno penzias the nobel laureate who ran at t labs bellev's research uh said no no we're research um we're the research part of att our job is to tell the truth and uh the the truth serves us well and he really meant that that wasn't

just rhetoric that was what he was saying to the people in the company that we're trying to to stop um publishing from happening this is what the culture of of this research lab was like in the early 1990s and that stuck with me ever ever since that i wanted to make sure that no matter where my life took me i wanted to make sure to have a job where fundamentally i could say my job is simply to tell the truth and that's a great privilege and a great luxury if you can pull off having a job like that most jobs have other considerations in them too but if you can have a job where your

job is to tell the truth um you really have a strong responsibility to to do that to to find out things and and and and speak the truth as you understand it uh and many of us you know not all of us but many of us have that privilege and i think uh one of the things we need to remind ourselves of is that if we are privileged to do that we have also have a responsibility to do that and um that's certainly a way that uh i've tried to guide my own career and it's one of the reasons that i went from 18t to a university at the time i did so after landing at penn um i um

uh started to look at uh i i tried to to kind of look at non-internet security and one of the things that i did kind of just before i left at t although it wasn't connected to my leaving 18t i should say is i said well you know this kind of place that i'm in it's kind of burning down what should i do let me look at other kinds of security and lessons for that and i decided to look at mechanical locks now in our community at this point locks and lock picking and understanding how locks work is part of our an ingrained part of our culture at the time it wasn't as big a deal in

our culture then it wasn't something that was part of the sort of background of every every hacker skill includes understanding and manipulating mechanical locks and i was interested in a question of is there a relationship between cryptography and mechanical security devices like locks and i found that as soon as you start to look at mechanical locks in those terms some really interesting things pop out of that and one of them is that masterpied mechanical locks which is the locks that um are used to open uh all the doors in a building where the janitor has a key that opens all the doors you have a key that opens only your door um what i found is that there is a

a privilege escalation attack against locks that's very efficient you can think of the lock to your door as an oracle and with making you know a small number of test keys um basically quadratic trials in a linear in the number of pins you can convert the information about your key into the key that opens all the doors in the building and i thought well that's interesting let me you know let me publish this and you know i thought well you know look my last experience with publishing this this sort of stuff was the nsa and they actually had a really good sense of humor about the whole thing um uh so you know this is this is

nothing compared to you know the breaking a government encryption system uh you know uh let me publish this and i didn't think much of it and then i like immediately the day that this got published started getting hate mail from locksmiths um and uh saying basically you know you have ruined everything um how dare you uh this is going to uh completely destroy security for everybody uh so locksmiths turned out to have no sense of humor about this sort of thing and we're very much like computer companies in the 90s a a a thing that will keep coming back over and over in 2005 i looked at surveillance systems again with my my grad students micah

gaurav shah sandy clark and we looked at how do law enforcement wiretaps actually work that is what are the mechanics of tapping a telephone and what we found was that there's in-band signaling in wiretaps uh which is to say there's an idle tone on a wiretap when you're when your phone is being tapped and if you put the idle tone on your um on your phone line while you're talking at a lower level it turns off a wiretap and we discovered that by basically buying a bunch of wiretap systems that had been decommissioned on ebay and experimenting with them and then we found that there are actual standards for how this works because they can interoperate

with each other uh and we basically found out a way of turning off law enforcement wiretaps and so we published that and um what we discovered was that the fbi has more of a sense of humor than locksmiths do um they were actually very adult about the whole thing um at least as far as we know um uh so um uh uh and i don't know if it still works um and then in in 2013 what i discovered was that i decided to look at um more non-internet security and looked at the protocols used to encrypt the two-way radios used by the fbi and secret service and so on and we discovered you know the

this is an example of this these are basically more expensive versions of the walkie-talkies that everybody is uh carrying around and they can do aes encryption uh on them and they're used for you know uh highly sensitive law enforcement and national security operations and so i got a bunch of these radios and kind of reverse engineered the standards called project 25 and what we discovered was that there were all sorts of um uh subtle protocol errors um that you know if as soon as you start looking for them you find them um you can do denial of service you can do things like erase keys remotely um you know we found a number of ways that you can actually

ping radios without any encryption keys and they'll tell you their location because they have gps built into them uh which is you know somewhat useful if you're worried about being surveilled by the the government you can kind of get the marauders map of where all the watchers are and uh you know the standard was just kind of a a complete mess and then you know actually wondered you know is anyone using these attacks and then we discovered none of these attacks matter because about half of the encrypted transmissions that the government was using at least in the cities that we sampled this in were going out in the clear even though clearly they were thought they

were encrypted they were like talking to someone where half the conversation would be encrypted and the other half wouldn't be um so we discovered there were terrible usability flaws in these radios that completely um dominated any of the clever attacks that you know we thought were clever attacks that were found all you had to do was what robert morris uh the bob morris uh senior who was at nsa uh he gave a talk at crypto in 1994 in which he gave the nsa's first rule of cryptanalysis and we were all very interested in what the nsa's first rule of cryptanalysis is and everybody got out their pens and started taking taking notes and he said rule one of

cryptanalysis first look for clear text and in fact uh that that works as it turns out um and in fact it even works for the us government as far as i know uh the these still have a problem so one of the things that all of this stuff has in common is that the um half of the difficult part half of the interesting part and half the important part wasn't finding these things out they were pretty easy to discover using the standard toolkits of computer security and hacking and and computer science these were all none of these are great intellectual achievements but all of them required quite a bit of effort to actually have impact and to get fixed

and in fact some of these things still aren't fixed half of the work is explaining it in ways that will explain why this is important why this is worth worrying about why should why is this worth fixing and sometimes even the best of efforts won't do it in after uh the clipper chip uh work got published i got invited for the first time to do something that i thought wow this is a great honor i got invited to come down and testify in front of congress at a hearing on this and i you know i'm i'm trained as an engineer and you know as an engineer your idea is well if you're if you're right about something

you just explain it and then they'll fix it uh and it turns out that's not actually how it works in washington um you know the being right isn't isn't sufficient you have to kind of make people care and it really took about um six or seven years for the u.s posture on encryption to change from being hostile to being encouraging um and that's because you know people in uh in industry took the time people in the crypto community took the time to explain why it's really critical that we include um strong cryptography in our in our infrastructure so knowing it isn't isn't enough we have to actually learn how to explain it so that brings me to

a project that i've been working on for a little bit north of 15 years in one way or another along with many very very talented colleagues around the world and in our community and that's the computing systems that support our democracy itself the voting systems and election systems that we have so i want to spend a little bit of time making the pitch for you that this is important to work on and what some of the parameters of this are and how the technical aspects of this interact with the larger social and democratic institutions they support and these are intertwined in ways that really make this i think secure voting and high integrity elections in the you know

25 or 30 years that i've been working in security this is by far um you know without doubt the hardest problem i've ever encountered um so um this is an interesting history um elections in the u.s at the beginning of of our democracy were pretty non-technical they basically involved people showing up at the town hall and raising their hands um and voting and there'd be kind of consensus in the room about who the winner of the local election was and that's kind of how it worked um that has the disad a couple of disadvantages the most obvious in retrospect is that's not a secret ballot and we kind of understand a secret ballot as being a fundamental

requirement for elections but that's something that only evolved over time as technology for secret ballots started to be implemented we only later started to understand that this is actually a requirement for elections the reason we stopped using everyone show up and raise your hand is that it really scales very poorly as your country starts growing and uh you know this may work for you know electing the mayor of your very small town but it it um stops it starts being very cumbersome for any kind of larger thing and where the number of voters exceeds the number of people that fit in the room um so voting mechanisms and voting technology have started to become important very quickly

as the united states was growing um starting with the very simple technology of marking a paper ballot and putting it into a ballot box in the 20th century really the mostly the second half of the 20th century more mechanical and electronic voting mechanisms have started to come into play machine counted ballots um uh voting lever voting machines that you um clip flip a little lever and then pull a big lever and that opens the curtain and you hear a big chunk and casts your vote machine counted paper ballots that are read by an optical scanner punch card ballots in which a little hole is put through a ballad and those were pretty famous after the 2000 election i

have an example of one right here and then finally in the 21st century we started to see outright voting computers where you'd vote on a terminal um and you know one of the consequences of this is that our confidence in the validity and legitimacy of the outcome of the election now depends not just on you know did we all get to vote but it also depends on the question of is this technology that we voted with actually of high integrity should we trust the technology itself as software people our answer is almost certainly are you kidding this is this is made of software now let me just point out that some of the requirements that we

now have for voting fundamentally contradict each other um one is that we have a requirement for secrecy and we also have a requirement for transparency we have a requirement that you no one can find out how you voted in fact you shouldn't even be able to prove how you voted because that would enable someone to coerce you into voting a particular way but on the other hand you want to be able to be sure that your vote got counted and you want to really be sure that everybody's vote got counted how do you do that how do you meet both of those requirements at once it sounds either hard or impossible and another difficult problem is that most of the

technology that we know how to do as computer scientists with things like cryptography are aimed at detecting irregularities they don't actually fix irregularities so one of the problems with elections is that it is virtually impossible the ninth district in north carolina this year is proving to be an interesting and very painful exception to do an election over if you discover an irregularity you can't just hold a new election you have to just kind of do the best you can in general with the results that you have so in fact a technology that says oh this this cryptographic hash doesn't match is almost worse than not having it because what do you do at that point and so

one of the really hard problems is remedies for irregularities not just detecting irregularities in the u.s there are high stakes in elections we care about who wins them there's also a long history of various kinds of fraud mostly in local races not um presidential races presidential races get all the attention but you know if you want to buy an office it's uh historically you know mayor or sheriff or dog catcher are the offices that uh have a lot of the shenanigans going uh on we also now have as we discovered after a recent presidential election that the threat is not just people trying to steal an election because they personally want to win it or they have a particular

candidate that they support but is state actors who might be satisfied with simply disrupting an election um or casting doubt on the integrity of its outcome and so we now have uh essentially local county clerks up against you know the fsb as um and and that's uh not a particularly great uh situation so uh voting in the united states is decentralized but high but also hierarchical in that the um federal government has a very limited role uh it sets broad standards uh things like you know if you're a citizen you're entitled to vote um if you're above a certain age and so forth um uh but each state has the laws that govern the details of elections

within that state uh in most states elections are run by counties there are about 3 000 counties in the united states which means there are actually well over 3 000 individual election administrators in the united states all of whom have to basically procure voting equipment and manage all the logistics of an election and report results out on election day and very often these same people have a different job the rest of the year so very often the i.t department is you know also supporting the roads department and the budget for managing elections is often competing with the budget for things like uh you know fixing potholes and building fire stations so if you ask you know which would you prefer a new

firehouse um uh or you know shiny new voting equipment that's a pretty tough sell for the shiny new voting equipment or for securing the um voting system that you've got because that's the simple reality of how we we fund these things um we also have uh probably the most complex elections in the world uh and by complex what i mean is the number of different races the number of choices that voters have to make in the u.s is larger than almost anywhere else on earth and we have and even an individual precinct an individual polling place might have multiple ballots depending on where a voter lives or which party they're registered in give you an idea of the scale of this

there are about 117 000 physical polling places in the united states there were in the 2016 election 138 million ballots cast some of them were cast by mail some of them were cast ahead of the election and that basically leaves 82 million americans who voted on in person on election day at a polling place that's a pretty big logistical thing managed in a totally decentralized way so the 2000 general election was an interesting thing who recognizes this picture uh this was the recount of the 2000 general election and this picture was on every newspaper a new news report in the country and it basically was look how silly those people in florida are they're using this antiquated punch card

technology we need to get rid of that and i'd like to look at that punch card technology because um it's kind of interesting this is an example of a voting machine of the type used in florida though this particular one was from i believe uh michigan and what you would do is take this little stylus you put your card in the top of the machine and you take this little stylus and you punch a hole through um where your um uh candidate was and the ballot was very complicated there was this thing called the butterfly ballot where it was printed on both sides of this book but that's one of the interesting things that would happen was that normally

you'd punch a ballot and you'll notice number hole number 68 is punched through um there's basically a little piece of cardboard that was on the ballot card that disappears after you punch a hole through it um now you i'll notice this is a totally mechanical device the only electricity involved is for the light um so that you can see what you're doing but uh this has an interesting property that little piece of cardboard because of you know conservation of matter goes somewhere um and uh where does it go well it goes right behind where the number 68 is in this particular case and so what that means is that on a normal day everything's fine but if an unusually

large number of people turn out to vote the cardboard starts to back up and it becomes physically harder to um vote for the most popular candidates because what eventually happens is that you're only able to kind of dimple the piece of cardboard and not actually remove it and even though you're making a little mark there when the card is read by machine it won't actually show a vote in that position and so this is really interesting because i i think this is an example of a completely mechanical system with what you could argue is a buffer overflow in it so what happened was that congress after the 2000 election totally divided country disagreement on who should have won

widespread agreement that we need to replace these antiquated voting systems and so congress passed something called the help america vote act which basically provided federal funding to switch to accessible voting technology states could now buy new shiny voting equipment the problem was that that new shiny voting equipment didn't actually exist at the time the funding to buy it um was provided and so into an industry kind of popped up out of nowhere to produce voting machines and the most popular is called the direct recording electronic voting machine which is basically a touch screen computer that records your ballot selections in its internal memory so in other words we've now created voting systems based on software

i will observe this is computerized voting and that it depends on software and also i will make an observation that software sucks um see all the other talks in this conference uh for a reference um so um one question is can we make voting systems secure out of hardware that we don't actually know how to secure out of hardware and software that we don't know how to secure and i will make a humble observation as a cryptographer which is that cryptography doesn't actually help with most of this because most of what cryptography can do is detect errors it can't correct them so this goes back to the problem that knowing that your election was tampered

with is incredibly unsatisfying if you can't produce the correct outcome i'm not going to go over all of the details of this i will say that in theory this should be a matter of looking for a line in the code that says you know votes for candidate equals votes for candidate plus one and if that line is present your voting system is probably good if not then there's a problem but in practice it's much harder than that because it's dependent on the entire platform and in fact when we looked at voting systems at the university of pennsylvania i led a team in 2007 other teams looked at other manufacturers i looked at the system my team looked at the system made

by esns what we basically discovered was that every single component of the esns system had unbelievably terrible flaws in the software and hardware that would allow you if you could touch any single component you could compromise kind of the entire election in that county and um those results turned out to be typical anybody who looked at any vendors voting system pretty much came to the same um pretty much came to the same uh result every current voting system that's been examined critically has turned out to be terrible so what can we do about this and i'm now going to go back to the citizen hacker aspect of this so one of my pr i think the

my proudest achievement was my participation in getting an exemption passed about four years ago to the digital millennium copyright act everybody's least favorite law um this is the law that basically makes it illegal to publish vulnerabilities and look for vulnerabilities in systems that protect intellectual property which is almost everything well we managed to get an exemption to this that allows for good faith testing of consumer products for security flaws and this is defined sufficiently broadly that it includes voting machines because consumers use voting machines and that's kind of a stretch but they specifically said this definitely includes voting machines and this was actually just renewed it has to be renewed every three years and if the effect is it makes it now

legal for anyone not just the people who got the exemption to go out and um uh get voting machines and look for weaknesses in them and publish them and in fact every year at defcon we now have a thing called the voting machine hacking village where we get a bunch of voting machines and invite people over the weekend to come in and do their worst with them and replicate some of the attacks that have been found and find new ones and you know essentially what we find is that wow we weren't that smart because other people the things we took weeks to find other people are able to replicate in like three hours um so um you know what should we do is now

the question as citizen hackers so um the internet is almost unanimous if you ask random people on the internet the solution is either no software should be anywhere near elections you should hand count everything or more software should be used in elections um that is you know things like vote on the blockchain so i'd like to to look at both of those extremes the no software sounds kind of you know i was originally in that camp when i first started thinking about this and i know how software is uh and i'm thinking i don't want software anywhere near elections any more than i want software near my you know my surgery or something like that

but unfortunately u.s elections are really complex and computers actually solve problems in vote counting and vote tallying and ballot creation that election officials actually have and one of the most daunting of these problems is accessibility of the ballot not everybody can read or hand mark a paper ballot and those people are as entitled to vote as you are and so how do we complicate how do we accommodate very very complex elections voters who speak multiple languages and have different physical capabilities um and so on computers have a role here and it you know saying that computers shouldn't have any role is not actually a practical solution uh the more software solution is use the blockchain

um you know and it sounds almost perfect it's an immutable decentralized ledger but in fact it doesn't actually solve any problems that elections have but it makes a lot of them worse elections it turns out are not decentralized consensus exercises even though they sound like it in fact elections have officials who are empowered to declare who the winner is and are empowered to run the election itself a blockchain doesn't actually help solve that problem when you have a single authority whose job is to basically decide this is who the winner was and who operates the election itself and every blockchain based system that's been proposed has has had the property of both having terrible security problems

and not actually solving any problems so please just stop with this um so uh the probably the most important document that comes out and that i'll um that i'll um urge you to to look at is the national academies of science engineering and medicine produced a consensus report last year called securing the vote it is a remarkable document it's just a little over a hundred pages of a overview of how voting works what the problems are and recommendations of solutions and it is really one of the first things to convert the problem of voting into a real engineering discipline um and and i if you're interested in this i have two big recommendations first read the

national academies report you can download it for free on the internet um the um if you google securing the vote national academies you'll find it uh quite quickly it is well worth your time if you're even remotely interested in this and if you sat through this talk you should be remotely interested in this you've already spent uh you've already invested time that's a sunk cost we all know after a sunk cost you should uh continue on right that's the only way to go at least that's how bitcoin works um the uh um the uh um the second is become a poll worker um almost everywhere needs people to help on election day and that will

teach you how elections work on the ground and get your relationships with the people who run um elections in your county and who could really use your infosec expertise um there have been two real important advances in the um engineering of elections one is due to um uh rivest and whack the same rivest who's the r in rsa uh he came up with this concept it's badly named it sounds like it means something other than it means but it's a really important idea says that yeah you can use software in elections but your election system should architecturally be designed to be software independent and what does software independent mean well it means a voting system is

software independent if an undetected change or error in the software can't cause an undetectable change or error in the election outcome now it sounds like if you think about that requirement it sounds like that's just the same as saying don't use software and that leads us to the second important contribution of elections which is a due to phil stark at berkeley a statistician who came up with a scheme called risk limiting audits and the idea of a risk-limiting audit is it's a statistically rigorous method of sampling um precinct counted optical scan ballots which are counted by computer in a way that allows you to sample a small number of them verify that they were counted correctly

and give you calculable confidence that the correct outcome of the election has actually been reported and if you define discrepancies you have to count more of the ballots until ultimately if the software was failed you end up hand counting all of them but if the software was working with a small amount of sampling you can you can only do a small amount of sampling and and take advantage of computers for doing uh the majority of the tallying it's a huge advance because you can use existing equipment optical scan paper ballot readers in the precinct in order to achieve this um so where do we go from here well we have the answer but having the answer is only part of it uh

it's the easy part um the hard part is now we have to actually get our elections more secure and you know the national academies report carries real weight that was a very helpful thing but there are over 3 000 counties in the united states there are about 5 000 individual election administrators all of them need to understand the risks of this also there's more to the elections than just the voting machines there's back-end infrastructure for managing voter registration databases that was actually targeted by foreign adversaries in 2016. the mueller report actually talks about successful penetration of some of these systems in other words those three thousand five thousand counties they need our help please help them so

thanks very much

i think i i think i have about 45 seconds for questions okay okay great oh yes yes sir yep what about estonia estonia is a different country than the us they it's a parliamentary democracy um estonia is a land of many contrasts uh the uh so estonia is uh uses online voting they also have in-person voting um the um they've had interesting successes with it there's also been questions raised about the security of their system their elections are much simpler than those in the united states okay thanks very much