 This project is hosted by Karuna, which is a collection of independent critiques of various aspects of digital India. My name is Shankarshan. I'm a member of Karuna and I'm the curator and promoter of this project. The project is also supported by Hasgate.com. For most on this session, Hasgate is very familiar, but just in case you do not know or you're not familiar with Hasgate.com. It's a platform for collaborations across practices surrounding technology, design, law, policy, systems, data, and all sorts of things. The collaborations are through user-generated content that are shared by practitioners and experts. This leads to discovery and elevation of ideas. Hasgate.com and Hasgate Media Division provide the underlying infrastructure, tools, services to facilitate these collaborations. The reason we are having all these masterclasses are to enable all the participants to acquire some foundational knowledge and perspectives, which are required to evaluate the intended and unintended consequences of technology interventions. And we hope to achieve this through a diverse set of perspectives, which focus on identity, equity, privacy, security, rights, agency, and the socio-economic impact of such proposals. I'm absolutely stoked to introduce today's moderator because Smari has been at the forefront of some of the things we're going to talk about. He is a politician, a software developer, an innovator, and an information activist, known for his work relating to direct democracy, transparency, and privacy. Smari is a member of the Icelandic Parliament since 2016. He chairs the future committee and the European Free Trade Association Parliamentary Committee. He is a regular speaker on topics like director of electronic democracy, rest freedoms, critique of industrialization as a centralizing force, and the culture of the internet. With that, I'm going to hand it over to Smari Makati and Olios. Take it away. Thank you, Sankarshan. It's an honor to be moderating this event. We're going to try to dive a little bit into the various topics of electronic voting, psychology, technological mechanisms for supporting democracy. I hope everybody can hear me okay. But there's of course been a lot of discussion about whether, how, and to which degree these technologies should be employed in our democracies. And on the one side, many people see the potential to expand democratic participation and make conducting elections cheaper and more secure and offering more nuanced possibilities for participation. For example, here in Reykjavík, Iceland, the public have for many years been able to participate in collaborative project selection and budgeting to improve their neighborhoods. But on the other hand, many have criticized the using these technological methods for lack of transparency for complexity for the need for expertise. And in many cases, elitism, as well as showing many ways in which these technologies can be tricked and abused, opening up a number of very valid questions about the integrity of elections. So to lead us through this complicated topic today, we have a presentation from Professor Sopacius Banerjee. Professor Banerjee teaches at the Department of Computer Science and Engineering at the Indian Institute of Technology in Delhi. He has been deeply involved in writing, educating and commenting on the intersection of technology, society and politics. And his recent work includes commenting on a range of topics such as contact tracing applications as well as electronic voting and I've had the pleasure of reading a little bit of his writing recently so it will be very interesting for me to hear his presentation. So without further ado, I'll hand it over to Professor Banerjee and then afterwards we'll take questions and comments and hopefully get a lively discussion. So, Professor, the floor is yours. So thanks, thanks Hasgit for inviting me. So I'll talk on electronic voting today and amazingly the last year, the whole year of COVID I've been involved with electronic voting because of this, my participation in the Citizens Commission of Election. So this was formed just before the pandemic started around March 2020 by Mr. Davis I am who is who is an ex-armyman retired as a secretary to the government of Ariana and one of the ardent anti electronic voting activists that you can come across so he got us together. My thought was to have a public tribunal of sorts, which was not possible because of the pandemic, but we got a whole lot of depositions from people. And we came up with two volumes. The first volume, which is what I'll try to cover today is about the EVM and the VVPAC and the compliance with democratic principles. The second volume probably is more interesting which I will not cover today has chapters on electoral roles. That one is by Harsh Mandar and Veer Amani. Money Power and Criminalization by Anjali Paragraj. So it took a lot of work to get the two volumes out and probably tens of hours of Zoom discussion getting into midnight and I'll try to cover some of that today. So what I'll cover is based on depositions by these ladies and gentlemen. So, and so I'll pass on the slides and the depositions are linked from the slide. So the notable part is there are two notable points. One is that ECI did not depose the election commission of India or their technical expert committee. There was no response forthcoming from them, despite several reminders. And second is that all these ladies and gentlemen, the second part of the thing, they readily agreed to depose but no deposition was forthcoming from them either despite repeated reminder. So my erstwhile colleague would be Vora volunteered to get in touch with them over phone, email, and she gathered thoughts from everybody compiled into a massive report and got them to sign off right so each of them. So this is a combined deposition by, by all these people and some of them are pretty well known, I think they represent the electronic voting community computer science voting community worldwide. There are others who did not respond to the deposition. So why electronic voting at all. You know, so when we talk to a lot of people, they thought that it's primarily for efficiency, but computer scientist and election commission of India they believe that it is more because of correct. The people in the commission they have conducted large elections and they think that manual counting of paper ballots is not that difficult even in India can be done in a day or two very easily. But as I was telling smarty, you know, I grew up in Bengal where it isn't the ballot stuffing. So by, by using dandas is a is an industry and I think young Indians may not know it may not have seen it the way I have. And it is generally believed in this country that electronic voting has prevented some of that. So it has been successful to prevent some of that and overall, I think that we found that people think that electronic voting has helped. However, this correctness we found is very ambiguously defined. What is correctness of voting and that concept both in both in the minds of electorate in the minds of VCI, and in the minds of the designers of the system are not. And that is something that we will try to address in this talk a little bit today. So when you talk about correctness of an election, the public trust in elections come from verifiability. And that how very file is the election outcome and verifiability traditionally has this connotation that is administrative, it is verified administratively. It is the election Commission of India, which on behalf of all of us verifies the correctness of the election and it's a very complex process that starts with certification of equipment. You know, equipment manufactured by BL and ECIL and they're tested and certified audited. There's a trustworthiness of custody chain. There's a process that defines the trustworthiness of custody chain of election equipment from the time of manufacturing to the time time of deployment. It's a three stage randomization process by which the voting equipment reaches the polling booth. And then post election they're collected, they're sealed, and the polling agents sign them off, and they are put in GPS track boxes to strong rooms so this is a custody chain of a very complex custody chain of processes. There is a trustworthiness of custody chain of VVPAT. VVPAT are the voter verify paper audit trails. So this is a paper audit trail thing that every vote that is cast is also recorded in paper. And this was introduced in India from 2014 parliamentary election but the 2019 election every vote had a VVPAT step. And the trustworthiness of the custody chain of VVPAT is also defined through a very complex process. And finally there is a VVPAT best audit of electronic cards. So this is what is called administrative verifiable. The general public like for example I do not have too much of a role in it. You know, we have to put our trust on the election commission of India or the officers who are conducting the polls for the verifiable. So this process can work. This is the way it works in most countries, not only in India. But you know, there are several lacunes in our processes that I'll try to bring up that we found out that I'll try to bring up. Now there can be much stronger verifiability that is possible by using computer science. If you use computer science, computer science has looked at electronic voting has a rich history of studying electronic voting for over 40 years. And you can give mathematical guarantees that every vote is cast as intended is recorded as cast recorded at cast means that you can show that it is recorded correctly just before you start counting and counted as recorded. So you can give a proof to every individual as a theorem that all these guarantees are met and there are no spurious vote injection. So this is possible. I really use those and the reason is that this will require you to use some non trivial cryptography and whether it is proper to use cryptography in large public elections when the electorate has no understanding of it, there are some ethical questions. So, which also I'll cover a little bit. So, so public verifiable publicly verifiable elections to the best of my knowledge and not not not not common anywhere in the world, they've been tried in the United States and a few places like Estonia. They have pilots and there are no large scale deployments so we have to rely on administrative verifiable primarily for correctness of elections. Then there is a notion of internet voting. You know whether you can do voting on the internet at all. This is also particularly relevant for India because the percentage of the Indian electorate about 200 million people are migrants. They're not at home, then voting happens and traditionally over the last 7075 years they have not voted. People who have traveled working laborers are a small number of non resident Indians who are abroad. 100 million people can do not get to work and that's that's that's an incredible statistics and there must be a way to technologically or otherwise to enable them to work clearly and lots of people are thinking of internet voting. Internet voting in computer sciences. The current consensus is that it's almost impossible because of two problems. One is that in the last talk alluded to the second one that it's impossible to make it coercion free. I think she said that women in the family are coerced by the men in the family to cast vote in certain ways. So, if you want to vote for other party and your husband thinks that you should vote for another party. So it is impossible to make it coercion free, especially in a in somebody's home. You know, even even if you can discount the neighborhood goondah telling you that please share your screen before I vote before you vote and show you who you're voting to, even between husband and wife. So it only booth is probably the only option. There is also this extremely hard problem of that reverse pointed out called the secure platform problem that you may own a laptop or you may own a mobile phone. But you very few people really understand how they work. They don't they don't control the operating system of the hardware for a very little understanding of the operating system of the hardware. So to trust an apple to do encryption for you correctly or do cryptographic execute cryptographic functions correctly for you and not leak your vote to others is a concern and reverse says that we cannot have democracies being controlled by apples and apples of the world. So, so there is a concern and there is no known method to actually that's not true. There has been a lot of research on both the secure platform problem and on the question free problem to enable internet voting and computer science, but those techniques are extremely You know, most computer scientists will find those from those those protocols hard to execute from home. So they're not practical. So there are there are theoretical systems that can demonstrate that both these problems can be solved using cryptography of complex kind, but they're not very popular they're not very easy to deploy. So the current state of affairs is almost impossible. And there's a joint report of the National Academies of Sciences Engineering and medicine that came out in 2018 is a 200 page report. And it was recommended that the US presidential election of 2020 we conducted in polling booth, so they should be strictly pulling book for the world, there should be no internet involved. And they can be electronic voting but they should always be accompanied by human relations for. So that was a national recommendation of the committee. And many of my computer science scientists friends tell me that the US elections, the last US elections got saved, only because of this, this report and its recommendation, which was followed by most of the states and Okay, so let's first consider why the US National Academies made such a recommendation and let's consider an EVM only solution and even being an electronic voting machine. So no VVPAT, no voter verified paper audit trail, you go press a button and your vote is recorded. Now of course a voter cannot see what has been recorded so you press to and you have to trust the machine that it is recorded to and not recorded to right so this proof is impossible. So there is no way to resolve a dispute. So what transpires between a machine and a man in the privacy of a polling booth without a witness. And if there's a dispute in that there's no way to resolve it. Even if there are two human beings who have agreed upon something agreed upon a number in private and they come out and raise a dispute there is no way to resolve that either. There's no way to figure out who is lying but then there are traditional methods like swords and pistols and fist fights but you can't you can't even bash up a machine. So, so this dispute resolution is theoretically impossible in an EVM only solution, there's no way to resolve that where the man is lying or the machine is lying is will forever be a mystery. So there is this classical paper by Rebecca Mercury actually a PhD thesis from 1992, where she showed that a machine as complicated as an EVM which means that it doesn't memory. It has got few registers. It records a vote in a, in a, in an e-prom kind of a device, and it has push buttons so a system as complicated as that cannot be proved to be correct. So proving it correct is at least NP hard may even be undecided. That's what she proved the theorem out there that nobody can ever prove that with bounded resources that an EVM is correct. So that's that's not possible. That's an impossible result. So if you cannot ever prove that it is correct there is no way to predict that it will whether it will ever reach. You know, it has an EVM will have a will can be in a million of states right so whether any of these states will violate democratic principles or not. It's something that is impossible to to to ascertain. So, whether an EVM can be hacked or not is actually an irrelevant question. It's irrelevant because that question can never be answered. If it, if it, if it's hacked it is hacked. If it is not hacked, then you don't know whether it can be hacked on. So, that's not the way to look at the problem. So, so that an EVM has not been hacked is no guarantee that it cannot be hacked. Also testing right our election commission of India seems to be saying that the EVM is safe because they've tested it adequately. You can only test for malfunction. You cannot test for a hidden program. That's that's known to be an impossible problem in computer science so testing does not constitute a proof. So you can test and detect that the display is not working or some button is not working but what is hidden inside a machine cannot be found out by testing you cannot. You cannot conduct infinite number of tests potentially and the state space is infinite exponentially large. So you cannot ever test for all possibilities right and to say that an EVM is not networked is a extraordinary nice statement because there can be no machine. That is cut off from the environment so it is constantly interacting to the environment if not through Wi-Fi, but through sound heat. So, so, so there can be communication channels that are open that are not obvious like YouTube or Wi-Fi or so on so you know there may be triggers like with the sound they can be triggers with infrared, then can be with an ultrasound from outside the booth that can be all kinds of triggers so it can be all kinds of antennas hidden and a HAC system can be easily made to behave correctly under test conditions so and it cannot be detected by testing. So none of these one-time programmability quality assurance etc they do not they do not lead to verifiability. So that line of argument is a false line of argument. So you cannot ever satisfy a voter with an EVM that these guarantees of an election can be met by any EVM design whether it is by ECI or anybody else. So there is a great deal of problem with the nature of discourse about EVM hacking in India so the owners says on the election commission of India to demonstrate correctness right not on, not on a general general public to show that it can be a hack. So there is this public invitation to come and hack you know hackathon and so on so forth. These are like a little bit like theater of the absurd this these things don't make any sense at all. So there has to be robust system to demonstrate that in spite of the EVM the election outcome is correct. This may not be a mathematical proof using cryptography but this you know it should be so the due diligence should be such that there is public trust ultimately you know that's that's the conventional system. So, so the legal examination of this issue is by was by the German constitutional court. I think that on 2009 and it gave the following ruling, it said that use of voting machines that electronically record votes meet constitutional requirements only the results can be examined reliably without any specialist knowledge of the subject. So, you know, some IIT professors declaring it safe will be unconstitutional or some election commission experts saying that I declare that the EVM is safe does not meet me constitutional provisions and this was a ruling of the German constitutional There was a friend of mine called Professor Court Milan, a computer scientist of some repute, he deposed before this court in favor of electronic voting and got overruled by the court. So I have interesting conversations with him about the German constitutional course rule. The court however said that the legislature is not prevented from using electronic voting machines in election provided the votes are recorded in some other way beside electronically. So this ruling is not applicable on other jurisdictions but somehow this is respected by most jurisdictions in Europe. So this ruling has led to Holland, England, Ireland, France. Good moving away from electronic voting and to paper based. Iceland I was discussing with smarty always had. Amazingly, most of the states in the United States also follow and side this German constitutional course. So this, this is a constitutional course ruling that defines elections at most places worldwide and in India we don't. And I agree that, you know, to declare that a paper based voting is correct. It is understandable by public. That statement is also not true. So so so if you if you do a completely paper based old fashioned ballot style voting. There are a lot of administrative processes which are defined by the election commission and a general public may not have any understanding of that either that that process is also fairly complex and correctness is difficult. So this constitutional course ruling is not. Not that it is not controversial or not that it cannot be questioned but it stands as the major ruling today. And, and pretty much use of cryptography in electronic voting has been ruled out because of this, this, this, this, this particular computer scientists are more or less backed off. So what to do then then you need baby packs, you know what a verified paper audit trail I think there's a German constitution report pretty much said that both servers must be recorded in another way. And I think that Rebecca Mercury the same person who proved that an EVM cannot be proved correct in the same case if this is recommended and actually she made very strong statement, she said it's impossible to conduct elections without whatever So she, she, she laid down the properties of what a baby pack system should be like and recommended in 1992 that all election should be conducted, all electronic induction election should be conducted with accompanying baby pack systems. So it is now commonly understood that there is definitely a need to move away from certifying motion machine as hack free or hackable and so on so forth so these are irrelevant question to establish in that the outcome is correct. You know do whatever use an EVM don't use an EVM, but establish that your outcome is correct and to gain public trust in large election. The only democracy principle is that that the public should believe that your election is current and this constant talk about elections being hacked this that or the other, it does not do any good to democracy. So, one has to design elections, assuming that voting machines can be hacked. That's the conventional understanding. So when a computer scientist designs an election system, he does not depend on trustworthiness of an electronic voting machine. I think that that would be called incorrect computer science. But that seems to be the election commission of India's approach and we are recommending that we should move away from this. So if you cannot trust an electronic voting machine and yet do electronic voting. How do you, how do you ensure that the outcome is correct. There are actually two only two ways of doing this right. One is you have end to end verifiable cryptographically secure voting system. I won't cover it today but in the next talk as an illustration I'll try to cover what such a system may look like just an illustration. This, however, because of the ruling of the German Supreme Court is a no. You cannot allow computer scientists to learn the election right people like me to run the election we are proven incompetent so. So that would be undemocratic. The other option that you have is post election risk limiting ordering using we be back. So that is what is commonly followed all over the world. And there the de facto standard is is this classical paper by by Philip Stark and Wagner. That came out in IEEE security and privacy called evidence based elections, where they showed how to do this, and this is such a great paper. And, and, and Philip Stark I got him to the post to our commission. So coaxed him to the post to our commission, but actually it was not necessary because it's this public paper clearly outlined how to do this audit and I'll talk about this is a little bit more. So there are these two methods, you know, there can be ways to couple them also, but the conventional wisdom the accepted wisdom. Both in computer science and in election author it is is that the second is a preferred option is a more preferred option that you do this team meeting audit with the facts. So let us examine that a little bit. So what are the essential requirement there are two major desired order. If I may say, right. The first one is that that election system should be software independent. This was pointed out by rivest in 2008 and reversed notion of software independence is that that if there is an undetected change in the hardware or software. For example, by due to hacking or by any mistake. If, if a hardware device of the election software functions in a way it is not supposed to. That should not ever cause an undetectable change in the election. So it is to say that there should be an audit procedure that should be able to catch it. If that is not the case that election protocol is incorrect. And this was Ron rivers definition that suppose you're using an election hardware. And if the election hardware has gone wrong and done something wrong, it should be always possible to do. Either you detect that using audit or not as a material but there should be the possibility should exist that there should be always be possible to detect that. And the second desire does dispute resolution. So support a voter is a dispute that my vote was not recorded correctly. So I voted for BJP and it recorded Congress for example, there has to be a clear determination either in fair favor of the voter or in favor of the election. It should be clearly determined that who is which of the two has made a mistake. The second one for a direct record electronic voting system is as we have discussed as impossible. If it happens in the privacy of polling booth. And if I press two and the machine says that you press three. Then it is impossible to determine who is correct. So, so, so the process has to be fair to the voter. You cannot possibly be fair to the machine. So you have to design a system that has to be fair to the vote. So ultimately there are just these two requirements of the voting system. It has to suffer independent. And there has to be a clear protocol for dispute resolution anytime a dispute happens. The Indian election system. Unfortunately, despite its success fails on both these kinds. It is neither software independent nor is it fair to the voter. And these are not minor points and as I'll illustrate as I as I move on. So we be packed. What are the requirements they are required in law by most jurisdiction to make the polling process understandable to the common man. In India they were brought into law because of a public interest litigation filed by Subramanian Swami, which was accepted by the Supreme Court and the Supreme Court directed the election commission of India to conduct all future elections with VV packs. So it's a law in India. It has got into the representation of the people's act, but in a slightly vague way, and I'll say why, you know, so the major problem with the Indian voting system is there's no clear definition of which is the This is the electronic thingy that you record inside an EVM that is the word, or is it the VV pack slip that comes out. And that two are not in one to one correspondence because the electronic vote just increments a counter for BJP or Congress or you know political party a political party B. So when you press up when you cast a vote, some count goes up by one for some for some candidate. So there is no vote explicitly it's just a count some number 37 becomes 38. And a VV pack slip it falls into a sack from a printer gets detached from a printer and falls into a sack. And it is impossible to know which one was your work. So it's all gets mixed up inside a sack and become indistinguishable. The Indian electorate Indian election system in the representation of people's act have not clearly defined which of the two is a vote. This is important because suppose there's a mismatch between the two. As they're almost always is. So if you manually count the VV packets extremely unlikely that will match the electronic one exactly. Which one should be counted, which of the two should be counted, you know, in a, in a, in a nutshell. So this question came up in the Bihar assembly elections. Very, very strongly many of the candidates protest there were a huge number of complaints to the CI also to the Supreme Court, but this has not been satisfactory result the representation of the people said does not answer the question at all. The election commission of India behaves as if the electronic vote is the correct one. Whereas common sense demands that it should be the VV pack that should be the correct definition of the vote and most people I have talked to, or most people that deposed to the commission thought that the VV pack should be the correct definition of the vote. Now also a voter should have a full agency to cancel a vote. If not satisfied for example if I have voted for BJP. And the VV pack slip shows Congress right. What should I do. In India the way it is the system has been designed is that I can not touch the VV pack slip I can only see through a glass window for seven seconds, and then it falls into a sack gets that she falls into a sack. So the seven second visibility. Is that enough. Now if it shows that I made a mistake or it has recorded it wrong what am I supposed to do I cannot stop it so this is not really a voter verified system. So my only option at that stage if I don't agree with what has been done is to raise a dispute. And that will require me to go out stop the election process stop the next voter from entering the booth and ask the polling tell the polling officer look at record in my vote wrong and the polling officer. Penalty of 5000 rupees for frivolous complaint now how am I going to prove that I'm innocent you know what I'm speaking is the truth, because the VV pack slip is no longer there it is fallen there. So what the polling officer does in India is that he comes and does a mock poll. One mock test and if that is correct. He declares that the voter has lied. Depending on the situation either finds or not finds the water and as a result. There's been almost zero cases of water complaint. So many voters have come out and said that you know that guy recorded my vote wrong, but an official voter complaint is unheard of and this is a completely incorrect design. So, in our report we recommended that should this should be changed promptly. And so the Rebecca mercury the way should define baby pad is that that the voter should have full agency to cancel her vote, if you think that it is wrong. And this process of canceling should be simple and should not require her to interact with anybody any polling officer she should be able to cancel and recast anytime she she wants. The only real way to do it is that that this baby pack slip must come in the water sand and the water must drop it in the box with with her own agency and then it's a proper proper voting system. So, the commission's recommendation was that we designed the baby pad system in India to make it truly water verify properly defined which is the vote. And amend the representation of people's act, which makes baby pad mandatory for voting. Okay, so. So that may be trustworthy during voting right so they while the voting happened. Suppose they're trustworthy suppose they design a system where the baby pad slip comes in the voters hand and the water inspects it and then puts it in a box. The question is that that do they remain trustworthy during audit, like in Bengal elections that were happened in seven phases over a month. So, in this one month the baby pad slip was in some strong room somewhere was transported in some suitcases, some, some buses somewhere. And the question is out there trustworthy, how do you ensure that they're trustworthy. So this is a process which in computer science we call compliance or an again start environment says that you know what are the kind of properties that can be defined for this administrative process. And how do you ensure that this administrative process is completely trustworthy I think that what we follow in India not only in India in most countries is some kind of an ad hoc process whose properties are not clearly initiated. So this process is, you know, the custody chain. You know, it can never be foolproof, but it can be made as near foolproof as possible by by by following certain processes and these processes are well defined in this classical paper but almost never fall. We also recommend that this be followed by the Indian system. Then comes the final question that how many of them should be audit right so how many ebm, you know, so the ebms are the electronic counting can happen you know you just shove in the ebms into a port and the electronic counts come out each ebm holds about 2000 words 1200 words 2000 is the maximum capacity. So you can count in a few hours just by the electronic count and a few of these ebms should be cross chipped by full manual count the question is how many you know that is a baby pad audit. That is called the risk limiting baby pad audit. And this process is very well understood. And there are also these issues of suppose they don't match suppose the ebm count and the baby pad count don't match what do you do. Do you declare the entire election for the entire population null and void and we will announce a reelection. Or do you hand count the paper slips for the entire population and say and discard the electronic votes. So this process is again not defined very clearly in the representation of the piece at its ambiguous as ambiguous as it can get. In fact, this, the answer to these questions to questions are completely undefined amazingly and and that makes it extremely unsound right so what is the audit process and what should be done if there's a mismatch. And this is related to which is the which is the current board, right, and the total number of complaints are both to election Commission of India and Supreme Court are incredibly high there 23 cases on this issue pending in the Supreme Court of India, which is not heard for very, very long time. Commission found that completely, completely amazing. So, what is the, what is the, what is the cephalogy technique that you should use. There is this statistical probability distribution called the hyper geometry distribution. And this is I'm quoting state from stocks paper. This paper of risk limiting audit. Start and Wagner. This describes the probability of finding a defective items in n rows without replacement. So the binomial distribution is with replacement. And the only difference between the binomial and the hyper geometric distribution is with replacement and without replacement. So this is the position size of capital and and you assume that there are exactly k defective object. So, in this case k hack Williams, capital K hack. So the rate of hack TVM suppose is known, and I'll tell you where that figure will come. So suppose you know that they're k hack TVM then what is your probability of finding small k hack TVM with small n rows. That's what the hyper geometric distribution is it's second year statistics. It is taught to secondary undergraduate students. But it looks like not to Supreme Court of India not to election commissioner. So, amazingly the stark and Wagner paper was completely correctly interpreted and reported by an IS officer called Ashok Varadhan Shetty and he wrote an op ed article in the Hindu business Hindu Hindu business line in 2018 and saying how to do the He didn't propose to our commission but we took some of the cognizance of his op ed, and he was completely correct. He used the hyper geometric distribution and he showed that if you assume that you have one person for TVM hack TVM then to detect at least one 99% probability. You need to sample. So many years, depending on the population cells. Okay. So if you take the country as a single population the entire country has one population so India deploys about 10 lakh Indians, all over the country for 543 constituencies, a little more than 10 lakh Indians. You only need to sample 459 years. Right, if 1% of Indians are faulty and to detect at least one with the probability of 99 you only need to sample 459 years. But if you take a parliamentary constituency, parliamentary constituency and India can be of a size ranging from here to there. UP will be somewhere out here UP and be hard large states will be somewhere out here. So we have this range for a parliamentary constituency. For a for an assembly constituency. We have about this range. Right. So as you can see that the number of museums that you need to sample the taper off. Right. So for us. So percentage wise if what if you're doing for an assembly constituency, you may have to sample more 50% of unions, but if you're looking at a country of the whole you need to sample as a very small set of. So the question is that what is the correct definition of the population. So this question came to the Supreme Court of India, and the Supreme Court of India. Ask the ECI to file an affidavit to show that what is the correct thing. So, the election commission of India said one per assembly constituency is enough. That was their affidavit that if they do one per assembly constituency one even personally constituency that should be enough. And as supporting evidence that filed an affidavit by two professors from Indian Statistical Institute. Very, very well known statisticians, Professor Abhay Bhatt and Professor Rajiv Karanthika, two of the most eminent statisticians from the country. And they said that one is sufficient, one per assembly constituency sufficient. You know the Supreme Court. You know, despite their lack of facility with statistics found something fishy. They refused to write down an order. But they verbally said no one sounds so small do five. And that is the current practice that you do five. And this case is pending in the Supreme Court, the Supreme Court has not given a decision on the matter. And Election Commission of India does audit of five evils per assembly constituency to declare the election correct. Now, this is problematic for the following reason that you know so where did I have heard the two statisticians come from. They took this figure. They said 478 are sufficient so clearly they use the hyper geometric distribution. They took the actual number of idioms deployed which is a little more than 10 lakh. And did the hyper geometric distribution and came up with the number 478. Now that assumes that you have to have a homogeneous hacking strategy for the whole country. Now the whole country doesn't even have the same political opponents right so to assume that there's an entire homogeneous. So the population is the entire country and your hacking strategy in Kerala is same as hacking strategy in West Bengal is extraordinarily nice right. So this was, this was completely amazing. And that's the star size of our population size, you know, an assembly constituency deploys 300 to 300 ABM and parliamentary constituency deploys so many evils, a state deploys 150,000 to typically 10,000 evils and India as a whole is slightly like more than 10 lakh. So, so what is the consequence of assist prescription actually it's not Supreme Court's prescription it's actually election commission of India's prescription which they managed to get the Supreme Court to sort of agree to write though the Supreme Court thinks that something is fishy because they never pass in order. So cross checking five per assembly constituency or approximately 38 per parliamentary constituency. If there are 25% ABMs that are wrong, that are hacked, right, you will catch it with the virtual certainty. But if there are only 1% of the ABM that are hacked, your probability of catching it is only 0.33. Right, so with the probability and with an overwhelming probability of 0.67, you will miss the fact that EVM has been hacked. Right, so this this this sampling is incorrect. If my second year student gave the sampling strategy I would have given that given the student a zero in the exam, but but it stands. So, the other thing is that of course the margins are important, you cannot come up with a number of how many ABMs to audit independent of the margin so common sense tells you that if the margin is a single vote. Right, if party A has defeated party B by a single vote. Then the audit will require manual counting of everything without manually counting everything you can't say but if the word margin is 10 lakh words. Then you need to audit only a few things right if the margin is overwhelmingly large that party A has defeated party B by 10 lakh words, then counting a small number of words is correct and the hyper geometric distribution tells you that. So the margin is 10% then for a prediction with probability 0.98 of, you know, that that 10% swing has been manipulated by EVM hacking. So let's check only three events in an assembly constituency. On the other hand, if the margin is 1% or 0.8% is a closely fought election. Then you need to sample many, many more. Many, many more EVMs and count them. So this is. So, so the, the EVM audit process the risk limiting audit process has to be much more principle and not ad hoc. So a lot of complaints that are pending before the ECI and the Supreme Court regarding how many BVPAT slips to count are mind bogglingly large over the over the last five or 10 years and I think this is not healthy for for a democracy. So let's take five more minutes and stop with this question. The team that Shankarachan has given me that can blockchain provide verifiability. That's that's that's a question. So blockchain that typically blockchains come with internet voting. So as I have already discussed that internet voting is an impossibility right now because you cannot give a device a secure device to every person and you cannot ensure that the person is not forced at the time of voting. So blockchain can be used. You know, so what's a blockchain a blockchain is a is a ledger is a is a sequential block of pages where data gets into a block. And there are peers. So there are multiple participants in a blockchain. So every peer has a partial copy of the entire blockchain. So if I and Smari and Zainab we are peers in a blockchain. So typically all three of us will have copies of the entire laser or own copies of the entire laser. So to say that it is a distributed databases wrong. It is three centralized databases with three peers right so each one of us is has the complete access to the entire data. So the distributed aspect is only in the control. So what gets onto the blockchain is determined via consensus. So, so we three have to execute a very complex protocol. You know, everybody must have heard about the Byzantine consensus protocol or something similar to decide that what gets added on to the legend. So all three of us agree and electric leader and the leader gets to decide that what gets added on to the legend. So, you know, we cannot put stuff arbitrarily on to the legend. And then the second component of a blockchain are timestamped hash chains. So the hash is a is an electronic signature. So keep computing the electronic signatures of what has gone on the blockchain. In terms of hashes, and this gives the gives the blockchain a traceability and an unalterability so. So which and non repeatability so changing what has gone into the blockchain almost impossible, it's believed to be almost impossible, right, because it's publicly visible it's a publicly visible laser. So I have to tamper this blockchain I have to read compute the hash chain, you know, everything that has gone out there, and generate a fake proof trail, which is impossible to do. Unless I have infinite computing resources so that gives a blockchain and untamperability. So what is ECIS proposal, the ECIS proposal is great wine, right. There is no official official record of an ECIS proposal. I have had some personal communication with ECIS proposal. You know, they, they wanted me to design an election system for them but I didn't like what I said so that conversation fell through. But there are these, you know, I have I have linked it here so I won't do it now, but there are these articles of public articles, you know, in the next person so on so forth that seem to say that what is the ECIS, what does ECIS have in mind and I would say that I have never seen greater confusion than what is reported in those articles, you know, and people saying that why blockchain should be used for design. I, those are completely amazing and hilarious weeks. So, so, so from the looks of it, they are thinking of pulling book protocol not internet protocol. They are putting encrypted EVM data on to on to blockchain so that they become publicly visible, they encrypted and publicly visible. And probably they do homomorphic tallying which means that they do the vote addition in the encrypted domain itself. And there is some big talk on putting the electoral roles on blockchains. So if you read our commission's report on electoral roles, electoral roles in India are dubious, there are huge problems of exclusion. There are huge problems of disenfranchisement, especially of vulnerable communities. Tribals, marginal communities like LGBTQ, there are, so these are documented problems of electoral roles and they seem to think that putting them on blockchain, those problems will go away. And this without getting into too much detail, you know, blockchain derives the security from distributed control of multiple mutually adversarial entities. So it is highly insecure when there's only one authority. There is only election authority of India that is running a blockchain that is that gives zero security, none whatsoever. So they already get to decide what has happened in the EVM, what is the point in putting it up on a blockchain. So where is the mutual adversarial protocol on what should go on a blockchain that has already been decided by the EVM. So, you know, so unless you ask each political party to be an adversarial peer in a blockchain, I don't see how a blockchain can help an election at all of this type. And it is not clear that what aspect of the variability problem does it solve and how. You know, it does not really solve the EVM dispute resolution problem it does not really solve the software independence problem. So which part of the problem does it exactly solve you know that is not clearly articulated. And it is unsuitable for verifiability of electoral roles, exactly for the same reason that you know the electoral role problem defined on defining ground level processes by the electoral registration officers to ensure they don't miss out somebody. So, yes, you need transparency in the process, you need public accountability, but that cannot come from our data structure that requires multiple adversarial authorities for its correctness right so that where is that coming from is So what they may require is a cryptographically secure public bulletin board, you know that may address the electoral role problem somewhat and also working problem. So I plan to touch up on it a little bit. Next Saturday, I'm getting to cryptography a little bit. I will get into a cryptographic protocol not to advocate that it be used in election, but just as an illustration of what verifiability means know what what kind of variability will be achieved if you allow Max to get into it. So I'll stop here. I'll just summarize. So the question of whether an EVM is hackable or not is not a meaningful question at all. It cannot even be effectively answered, unless it gets active it gets acted is that otherwise cannot comment on it. You definitely cannot challenge people to hack it. It's neither here nor there configuration changes like one time programmability, we'll be at control unit you know these these things are immaterial they don't make a voting system very viable at all. You cannot say that it is verifiable because I've got a strong certification and a testing protocol that's fallacious argument like testing can never determine whether machine has been hacked or not. There are bigger problems with the baby pack system that the baby pack system as I said is not water verify. It doesn't give the user the agency to cancel the vote. And that is something which is sort of a low hanging fruit and should be corrected straight away. In the US system everywhere it was a paper ballot that the voter herself took to an scanning machine or put it in a box. The system was used and that that's obviously the correct way to do it. If you wanted by crossing them with manual baby pack should not be based on ad hoc strategies. And if Supreme Court cannot do proper statistics, they should be told so in no uncertain term to register for statistics courses statistics 101. Maybe some statistics professors also need that. They rely on only on well established professors on this limiting audit. And there are established literature on how to do all this. There are great election papers coming from computer science and statistics that tell you how to do it, and they should become the factor standard. You know, the design, if you if blockchain has to be used in election at all. It should not be a closed door design. You know how the first requirement is make the design public, let the public comment on it. Let hundreds of people pick holes in what you're trying to do articulate clearly what are the objectives, what problem in the election conundrum does it exactly address. Why would it make the electoral roles more secure. And apparently, on the surface of it, it does not really seem to address anything at all. So I'll stop there and the questions. Yeah, thank you. For really interesting talk. I agree with all your points. It's interesting just how how much of the complexity of voting gets glossed over very easily by people who purport to be experts and I think you attack the number of those quite succinctly so that was quite good and I have a number of questions but I think we should probably start by trying to take questions from the audience if we can. Yeah, so if anybody has any questions or comments or anything like that, then feel free to let us know raise your hand or put them in the chat and we'll try to get through as many as possible. But if, well, I haven't seen any questions yet. Oh, actually, there we go. So, Rupaksh. Hi, Professor Banerjee. So I had the privilege of reading your paper. And somewhere in the paper you speak about how if it's an electronic voting machine with a completely with a pure electronic process. When when they use when when a person casts their vote, they given a commitment. Now, to me the commitment sounded a lot like a unique identifier of some kind. That's a cryptographic protocol right. Oh, that's a that's a paper I wrote in a don't take my paper seriously. That's, that's only for getting a good publication in computer science journal. But what I'll do with the structures that maybe I'll take this question on next Saturday after I discuss a cryptographic. Absolutely. Thank you. Great. There's another question from an end in the chat. I think, unless someone wants to say it himself. I can just read it out. So, or it's more of a comment, I guess, a blockchain solutions, even if they have any point or only try to help avoid tampering after the vote is recorded. There's no mechanism to protect from the risk of malfunction of EVM that isn't that going to give a false sense of security. Um, I would think so. Yes, I would say that. You know, so it's either so you get in a blockchain, the mult, the distributed consensus decides what gets on to a blockchain, but if the VM has already decided that you know, even has decided that, you know, there's so many votes for party A and so many votes for party B and so many votes for party C from this polling booth. Where is the question of not get letting that data go into the blockchain. Right. And if you put up the data on a blockchain, then why why your blockchain, where's the distributed consensus here so the whole process of the concept was a little easy. And it's not at all clear that what happened. So when people design electronic voting system on based on blockchain. They typically assume internet voting. There are lots of paper in computer science where they assume internet voting and on the basis of blockchain and I think that systems have been tried in Estonia and Switzerland and and and a few other countries, but as I said internet voting is very very theoretical concept you know, nobody knows how to do it correctly. They're just fraught with dangerous. Absolutely. I liked when you said that they believe that if you put it on blockchain problems will go away. It tends to be the narrative about all things blockchain these days. It's a data structure it's a really slow database and yet it's being treated like a magic wand. But you know, I think in computer sciences and many fields. If you have a hammer, then every problem starts looking like a nail. But they would be good if if people were more specific about what exactly they mean when they promise various guarantees. But there's a question from Preetam. So have there been any, have there been attempts to have the same electronic EVM based setup for allowing voters to vote at any booth in India, probably with prior registration. If yes, what changes are required in the current setup. I think that India has been using AVM since 2004. And so you have an offline authentication. So you go with what is called a voter's ID card and somebody verifies your photograph and your name on the electoral role. And then you go inside and cast a vote till about 2014 that's the system was purely electronic system. So you have a machine with a button and you press to vote and believe that the machine is recorded correctly. As I said, that there was this public interest litigation by Subramanian Swami, BJP MP, and they changed that they changed that then to have a VB pack system and. Yes, so you go into India every vote is electronic with the paper audit. No, so, so, so, no, no, no, he is asking the migrant question. Yeah, exactly. So the migrant question, no, the migrant question was a big, big problem. So migrant, you know, so. So I think that there was a question and on its top last, last Saturday in Haske, where a past election commissioner, my friend. He answered this question by saying that they should be allowed to vote where they are right now. Right. So if they have migrated, if it's a migrant laborer who has come from Bihar to Delhi, and is working in Delhi, then he should vote in Delhi, I think that's too simplistic. That's that does not sound right to me. I would say that, you know, for a person who goes back and forth between Bihar and Delhi works in Delhi but his families in Bihar and he goes to Bihar in some seasons should be able to choose where he wants to work from, either Bihar or right and the state should facilitate the election commission of India should facilitate and so this in my opinion will require multi race voting systems. For example, if you set up a migrant voting polling booth in say Delhi somewhere in the Indian International Center in Delhi, then that polling booth needs to support 542 constraints. Like starting from Jharkhand to Bihar and so on so forth so this will call multiple races and the number of races will be fairly large. So if you design such a system. Correct me. There are two stages to it one will be a designing a correct registration system to record that the voter will vote from this constituency and not from this constituency and the second is the polling process by itself right. In my opinion are not that difficult to solve. It just requires a will to solve and anyway I think not solving it is a blasphemy for democracy, I think 200 million people not getting to vote is it makes our election process wrong. So I think that this is something that needs to be addressed. immediate and it's possible to address maybe in my next talk I'll give some I'll try to capture some thoughts on how this may be addressed. Yeah, I have a lot of questions about this because what you're saying kind of comes down to issues of, you know, what is citizenship how is it. How is it defined. How do you prove it how do you like, you know, in Iceland, for instance, we have a very. We know the name and then identity of every citizen it's and you don't have to specially register to vote, and this is easy here but you know in countries with well land borders for one but also just more complex histories and layouts. These questions become a lot more difficult, but I would say, you know, in this day and age if you cannot just go wherever you want and vote, you know, regardless of your geography, and have the problem of which, which polities you are a member of be be solved in the process, then that seems a little bit underdeveloped. I have a quick comment there. Putting it mildly. I know it's not an easy thing to solve but it does seem like if you if you pinpoint the topic too much towards migrants rather than saying why don't we just solve this in general for everybody it's you know, you know, add to that that comment that that is a question that is facing our country tremendously you know we. We have all hard, but that is not citizenship you know that is the citizenship so that doesn't define citizenship at all. The government came up with a proposal to define an RC national register first registry for citizens and there were huge fears of exclusion and there were fears that it may it is a political move to exclude some people and there may be some truth in it right so the amount of fear and amount of suspicion about that process was. The legitimacy of the process was completely completely incredible. And we also have a voter database system, an electoral rule system, which is the voter ID system. Now why the voter ID system is not the citizen register system. It's not clear to know how are they different and and the parliament and the and the electorate don't seem to find this or that we have a voter register system which is different from the citizen register system and and these questions need to be answered so so I think that. What is this fear about, you know, government taking away data that is disenfranchisement that is so so. So what is the correct way to do this, this identity problem is an open, is an open question in India and I think that it's much more complex than an isolate obviously and hence, hence the chaos. Yeah, absolutely. But I must admit every time I follow an election in the US or in various countries in Europe or, or even in India I do wonder like why, why these two databases are not just one database, it would simplify everything and and probably be less prone to abuse in fact. But if I can just put in a question and we have a comment also from the ACA, but so, because you were talking about provability and verifiability provability and any voting system really comes down to verify provability on one hand, but unlink ability on the other, meaning you want to not, you want to sever any link between the voter and the vote that is cast, and but verifiability requires being able to prove not only that a vote was cast but that it was registered correctly as you talked about in your talk. And, and, but also that it was correctly tallied and that it had the correct effect on what kind of Pharaoh called the social welfare function of the election. So being able to do that implies being able to link a particular vote to a particular voter thereby proving that the vote was cast correctly, which breaches the unlink ability concept so you have this kind of conflict there. By voting we achieve unlink ability by ballots being dropped in collection ballots into into a collection of ballots by the voter, thereby severing the link but paper elections are not formally verifiable. We can actually achieve Byzantine verifiability by having enough people with enough different objectives and interests in the process, keeping checks on each other. So, and then audit procedures and such but with electronic voting. It's much harder to attain. And as you say in your talk verifiability is still difficult. So it's like we lost the good feature that we had, and we also didn't gain the other feature so, and this comes down to identity and you got an endless number of people like criticizing the ad hoc system and, and all of these other methods of trying to figure out who's even allowed to participate. But so my question is, how are you going to bridge this gap and in particular like wouldn't the blockchain type solution specifically eliminate any possibility of unlink ability. I know that's a complex question but no no no I think that that's a no that's a that's a that's an incredibly correct question and it's a it's a question at the heart of computer science research over the last 30 years for on electronic voting. You know this is the primary question that what what smarty is saying that to give any kind of a proof to your water that I have recorded and counted correctly. I have to ultimately know how you voted, and there is a possibility that the very fact that I'm able to grip this proof mathematically. It presupposes that I know you voted for and hence, you know your vote is no longer secret right so so so there is a and that violates every principle of voting. So, so computer science users not necessarily blockchain but tries to use cryptography to solve this work, and it uses a most of the systems user technical zero knowledge groups, which is a part of blockchains also that. You grieve a proof of knowledge without so so you say that, you know, you say that I know how to factor number n, but I won't tell you what the factors are. You know, so you give up give a proof that you know what are the factors of number and prime factors of the number and SP by into Q, but you don't show what is the NQ that's a zero knowledge. You use typically zero knowledge books to get this very probably, but. And you make certain assumptions, so there are certain secrets that has to be held by somebody, and that's where the problem lies and you distribute the secrets over multiple trustees. In computer science so you cannot solve this problem without a regulator say if election Commission of India holds all the secrets, then this problem is unsolvable, but there has to be a regulator. The election Commission of India has to be under a regulatory oversight, and the two parties the regulatory election Commission of India, they can jointly give you a proof, and you have to assume that they're adversarial they won't call you to do the election. So under certain assumptions. You can achieve both verifiability and. But whether you can achieve public trust. That's the question that whether a public can trust such a system. So that's what I'm saying that a public will have to trust such a computer scientist so a computer scientist will have to declare an election to be safe and whether that is valid in a democracy. Actually, most computer scientists think no, no that is elitism it is worse so. So, I think that that is a primary reason that computer scientists have not advocated use of cryptography in elections. I think my personal feeling is that if not today, maybe 10 years, maybe 20 years is inevitable. When the general understanding of computer science and cryptographic goes up a little. The techniques are too sophisticated to not to be used. So what I thought I'll do is that that as an illustration I'll discuss one system, you know, I'll address that how to have both unlinkability and verifiability using a computer. And that's something I'll address in the next. Very exciting. Also electoral rules. Yeah. Yeah. So we have these comments from the ACA. And also just just to say, again, to everybody who's watching on the on the call, or if you're watching on YouTube just leave comments and and or questions and or raise your hand if you're on the call and we still have a little bit of time for a few more questions. But the ACA says, I'm thinking about the 5000 ruby a penalty for frivolous complaint. Early in the talk, you referenced the presence of the danda. Yeah, the actual booth. Yeah, so, you know, yeah, of course it is there, you know, I think it is reduced in India. The question. But, you know, growing up in Bengal, through the tumultuous time of 1970s I went to school in 1970s and I saw elections from 1971. When thousands of people were killed. The elections in Bengal are fought with swords. And so, so they're incredible, you know, the entire entire communities are told to vote in certain way, otherwise, they'll be enough, you know, that was that was a threat. So if you travel to rural India, you know, northern India, Bengal, we are a QP. You know, coercion is the name of the game, you know, coercion, bribery, giving, giving money to the entire community to vote in a certain way. It happened and there were complaints in the Bihar elections and the last Bengal elections as well. So, so these are tremendous social problems in a democracy such as ours. India is always there. It may have reduced a little bit, but I, I seriously doubt that whether whether it has gone away so easily. Whether I have changed any voting protocol in my academic work. See, in an academic work. The objective is to get a paper published somehow. Right. So we are a bunch of phonies and so, you know, academic world at many times as a little divorce from the real world. So, so the academic work makes certain assumptions sometimes which is untenable. It is untenable, but you know, we just try to demonstrate a technique many times and hence you. So every time you publish a paper on electronic voting, you don't advocate that that that method be used in public. So, so it is a little bit divorced from real life. But the commission whether it got complained about the Danda. Yes, it did. So there are these fierce journalists called Punam Agarwal who deposed to the committee. And the role of money power, the role of violence and election, and so on so forth. There were several depositions not related to the EVM team but the other things. And some of those depositions are quite shocking. So the depositions are public so you can go and read them. And it's an education it was an education for me at least I think that some of my other fellow commissioners were more worldly wise you know they have been, they've been the chief election commissioner and chief information commissioners and Supreme Court judges. So they were probably more used to the worldly ways but for me it was a shocking. One thing you mentioned that I think you're going to talk about this morning next session but you mentioned one of the proposals is homomorphic encryption and it's one of those wonderful holy grails of cryptography that people have been looking for and hoping for for years and it never really comes to fruition in the way that people want and in particular because it's very hard to depend against known plain text attacks and non cipher text attacks and that kind of encryption mode. Do you know of any, any research that suggests that this might actually be viable right now. So there is a most elegant paper by Ron Rivest and Ben Adida, his PhD student. Adida and Rivest is called scratch and vote. It's an incredibly elegant paper, which suggests homomorphic booting homomorphic tally homomorphic encryption and tally. And the paper is almost flawless. But they end the paper saying don't use it. In the conclusion it says that maybe the time is not ripe for such a method to be used. So this is a paper from 2009 to 010. You cannot follow the paper in terms of its elegance like much of Ron Rivest wrote this is also, you know, maybe a little complex to implement, perhaps, but probably, if you're not familiar with cryptography, complex to implement and I think that's good enough reason not to use. So, we have another question here from Bianca. So more of a general question on the role of the experts in the public sphere. So deeply appreciates Professor Banaji's skeptical approach. What would be your suggestion to cultivate a wider presence of self facing disposition among self by grandising experts. It's a bit of a pointy question but I think it's actually really valid and it says especially expert technocrats and tech moguls. I know where this question is coming from and you know I think that it is coming from a deep skepticism in which technical deployments have happened in India over the last 10-15 years and I sympathise with the views expressed completely because of the following reason that, you know, India has seen the deployment of digital technology in public life like no other countries. You know from a somewhat technologically backward country we have procced a little bit recklessly into digital technology, you know, so we are the only country that has a digital identity authentication system for 1.2 billion people. We have been doing electronic voting from 2004, no other country has done it. No other country that I know, definitely not at this scale but no other country has even ventured into electronic voting like this. You know our Prime Minister announced a national health ID system and digital health record, you know electronic health record system for everybody in the last Independence Day. England have failed to do it six times, you know, they tried it six times and six times they haven't done their databases, they have deleted their health record databases because of privacy concerns. Sweden has tried it and they have leaked data, you know, the entire data has become public last time in 2019. I think they leaked once in 2018 and leaked once in 2019. The United States of America hasn't even tried it, you know, they're too scared of such a thing, they have local small scale electronic health record but a nationwide electronic health record is unheard of. And, you know, our systems have been deployed by the process of what may be called crony expertise. You know, so you hand pick some experts from my institutions, IITs, you know, you pick a professor like me, and somebody like me, we get on to a committee and we write sign off a system of a mind boggling complexity. This clearly I agree that is a wrong way to do it. I think that, you know, when it comes to this kind of system, there are no experts. The design should be open to public consultation and public scrutiny. But even before the design the objective should be articulated clearly that you know what is it that you want to achieve with the blockchain, what is the object. First agree on the objective and then, you know, I, I have always believed that we reckless in ideation, but be super conservative in deployment. You know, you publish a paper, you have to be completely reckless, you know, whatever idea comes to your mind publish a junk paper and forget about it right and maybe get a promotion. But if it comes to deployment, you know, bringing it to public life, public consultation is the only make the design public and let expert non expert everybody comment on it for one year before you before you move on it. And there are no experts when it comes to a system like designing an electronic voting system for the country of the size. It should not even be done by experts is the ultimately public trust that matters. There's the only thing that should come. Yeah, that's a person. Yeah, I think you're absolutely right that ultimately comes down to trusting the systems. Here in Iceland we have most of this, these health records and everything available through public API is on online systems, and we have a common identity system so it we don't need a separate health ID and a separate voter ID you just have an ID. It's a simple way but yeah, that's worth a discussion of its own because there are certainly other problems. But I think we're running out of time. So I think at some point. Well, yeah, I suppose if you have any final comments and thoughts. Yeah, so I would have to get my phone it isn't here so maybe. Oh, well, we apparently have a bit more time so actually you say some clever words and I'm going to see if I can find my phone. No, but you know I can if there are more questions so you know just unmute and ask. I'm happy to discuss listen to your comments or whatever from anybody. Thank you for the wonderful talk. So I have a question. From what I could gather from your talk I think verifiability is a very important aspect of ensuring credibility of election. Yes, another low hanging fruit is how effectively we can use VB pads. It's what I absolutely see tangible steps that you can think of which we can readily implement so that we can increase the credit we can increase the verifiability and thereby credibility. You know the first thing is that make the VVPAT system voter verified. So which is to say that the voter must have an agency to approve what has been printed on the VVPAT slip. You just don't show it flash it for seven seconds and it drops into a box that's that's not that's not water verified. So, give a mechanism to the voter to cancel if the water is not satisfied. Right. And actually, the best would be for the voter to get the slip in their own hand and drop it somewhere so that is. So make the thing voter verified that's the first thing that needs to be done the second thing that needs to be done is ensure that the voter verified paper trails. They are secure between the election and the time of counting. Right. So there are standard processes outlined by election experts like Stark and Wagner paper tells you how to do it. These are semi cryptographic processes like putting seals in a certain way, putting putting parity checks to ensure that no motor verified slips have been injected or deleted from the system. Right. So putting just a seal and pulling agents to sign it is too simplistic. So right now what they do is that they put a seal a paper seal on it and pulling agent of everybody puts a signature on it. Now those things are easily hackable, you know fairly easily tamperable and there must be a better compliance audit process to ensure that the custody chain is trustworthy. From the time the election happened to the time that we've been put slips are counted there. The integrity is maintained. The third thing is that counter representative samples statistically representative sample of we've been put slips manually. In India they declared the results first and then count the way we've been put slips of five assembly constituencies. Now once you have declared the results and one party has started celebrating. There is no way you're going to change anything. Right. There'll be rights if you change if you try to change that election. So I think that that's a completely wrong process. I think that there is no hurry to declare the election. So this is the VVPAT audit of a necessary number of VVMs. Take your time. And only once the audit is the risk limiting audit is far satisfactory. So I think that in the short the low hanging fruit just these three steps will make the election process much more credible. Also, another addition to it will auditing at the moment for every assembly consensus there are five VVPAT. So can linking the margins to number of VVPAT that have to be audited will that be a right? Absolutely. You know, absolutely. And the evidence based election papers of Stark and Wagner tells you exactly how to do it. And amazingly this paper by Ashok Varadhan Reddy on Hindu Frontline you can Google it and find it. And he outlines it and as an advice to election commission of India in 2018. He he advises election commission of India on how to do it margin based company margin based audit and that process is simple and so correct. And then means that once somebody has pointed it out. Once the election commission of India has been pointed out by an IS officer serving IS officer that this is the correct way to do it. Why would such a recommendation be ignored? And why would you try to fight it in the Supreme Court I think that this is this is not the right way to do things. Yeah, I think somebody has posted it out there. Yeah, that is a paper by Ashok Varadhan Shetty and that tells you how to do VVPAT audit completely correctly. Yeah, thanks for that. So Kiran asked me to demonstrate the Icelandic authentication system. I showed him this last time I saw him in Germany so I wonder if the glitter has worn off a little bit but so just to explain the here in Iceland we have on one hand an identity number switch. Which individual has an identity number. It is for identification it is not for authentication. Right. So my identity number I can tell you 0702842789 knowing that will not help you impersonate me in any way, but it will. It is a semi public number that is used to identify me uniquely in every any database. But on the other hand we have authentication mechanisms are a couple that are competing with the one that's kind of one out is privately owned at the moment but there's the government has been moving to purchase it because it feels like the most common authentication infrastructure should be government owned, but using it. So, the way it works there's a cryptographic essentially a public private key that's on my on my SIM card, and it can be used so. It's bound to my identity. And I can go to virtually any website my bank my pension fund various health services you name it, or a variety of other things that just happen to use the same authentication system and login. So I won't show you my bank account or my health records, but I'll pick something relatively innocuous there's a database of all Icelandic people. Public and so I don't know if you'll see this but I can just click the login button. And it offers me to put in my phone number. So, I just put that in. It's pretty short. And so that after doing that, I just click login, and I get a message through the SIM card system. So, or through the phone system, asking whether I want to open this website, and then I put in a pin, which unlocks the credentials. And having done that, I send a response and boom I'm logged in. So this I did without any, any password I didn't have to create an account or anything like that. I just give it my, I just saw the website, my phone number, and the authentication system is handled through the phone system. And I just essentially allow my SIM card to sign a cryptographic challenge. That's it. This is I think a fairly sensible and wonderful way of doing things and really I don't understand why not everybody does this. So, I just want to clarify on this. In case people didn't get it. This was not an SMS. It was not an OTP login. This was software running on the SIM card embedded in the phone. It's not software on the phone. It's not an app. You can't replace it. It's hard-coded into the SIM card. And this is technology that every SIM card already has, including the SIM cards that we use in India. So essentially what's happened is this was a cryptographic exchange that happened on the SIM card, which essentially recorded consent to proceed with the login. And it's cryptographically verified. It is not OTP based. It cannot be fake. Right, exactly. And so there's a question about how digital the system is if you have to update the exchange details like address. There's really only one entity that needs to know my address, and that's the National Registry. Well, the Post Office, I suppose. And so it is entirely unrelated to my personal identity, where I live. So this is something I lived in England for a while, and it was amazing how difficult it was to authenticate myself because there was always this assumption that I would, where I lived was part of my identity. But, you know, I move around a lot. Why would that be part of my identity? That's a perennial confusion in India, you know that you're, why can't you have my son's biggest complaint is that why can't he have to address lives in Hyderabad and lives in Delhi. And he has to, he's not allowed to have two things, two addresses. And this is one of the ways in which migrants and people who live maybe non sedentary lives are being disenfranchised. So eliminating that notion of your address being part of your authentic identity and authentication is, I think, pretty important. So just, on the other hand, the downside is that this system means that my phone number is part of my identity, but, you know, okay, we have to accept something, I guess. That's like that's 10 digit versus 10 digit. Yeah. Yeah, yeah, exactly. So, but does this thing that I quite dislike, which is it conflates identity and authentication and fingerprints. Yeah, so, so that would be the authentication stuff right. So, so, yeah. Okay, so are there any other questions. Anybody want to join in those calls question on the digital aspect. I don't know if that was addressed, like how digital is the system if you have to update or change details like address. You can do all of that. Pretty much digitally like online but I guess my point there was that address and such are not part of the identity but yeah, we have government online services for a growing number of things. So, in practice you should be able to buy a house and complete the transaction without going to any offices or any anything like that. So just just do it all online through this, this authentication system and then the web services that are provided. And there's a question from on. Can I ask. Yeah, so. I think that I want to kind of bring out as like we're technologists talking about what technology can actually bring to the process but so one thing that actually worries me is we're generally talking about improving trust in the election process right should we start talking about doing more transparent extinct process for example, I don't know we have transparency about how the EVM has been manufactured, and then have an audit trail of how this is actually coming. And they're managed. So wouldn't that be something that we should actually be more worried about to begin with. The reason to that is that, of course, they should be transparent into the, you know, any public system whose design is not transparent as a blasphemy. So, of course, it should be transparent. But, you know, suppose they make it transparent. It still does not get you public trust, you know what will they do, they'll put up a design on a GitHub. They'll put up a source code on a GitHub, and they'll say that this is the way we design the EVM system know from there to proving that that source code is what is running on every EVM is an impossibility. Right so unless you have bring in some extraordinary cryptography, like proof carrying codes and so on so forth other complex stuff out there. So, while that is an essential step for public trust. If necessary, I don't think that is sufficient for public trust. So, making things transparent is something that should happen routinely. But public trust in an election cannot come from an opaque hardware or a software. The dependence has to go away from that somewhere. So, you know, I like, so Ron Rivest, if you if you read his papers, he has been writing for the longest time on elections. He thinks that election equipment should be based on public, you know, publicly available commodity electronic devices like maybe this one. Right, is the hardest to hack you've got to get it from Apple it is manufactured for one purpose. Just put some software out here like like smarty and Kiran's talk about right now. You know, put some cryptographic protocol and your election hardware should not be anything more than that or you know the Mac laptop or a Dell laptop that you that's that you can buy you know just come on dear this laptop from school. No, you know, every laptop is different. They're not manufactured for the election. So they are most difficult to hack and use public commodity hardware for running elections. That's the way you will get trust the most. So ultimately, ultimately, the requirement from the hardware should be minimized to the extent possible. And I think that you don't require anything more than smartphones ultimately to run elections. Smartphones as evm so not small enough people don't have to carry smart phone so so and if one smartphone goes back just replace it with another smart phone right so so I think that the evm ultimately has to be very simple. And the proof of correctness should be independent of the of the hardware. You need bad slips or writing for programming. Yeah, thanks. Yeah, so there are. I don't know. There's I guess one one more question from the ACA. So it would be good to know from sorry on what kind of technology debates or compensation has with his parliamentary colleagues. That would be a long conversation. I mean, one of the things that is happening in pretty much every country in the world right now is technology is rushing in very, very quickly, causing all sorts of madness. And countries are variably quick to to pick up and and adapt to this new world right. And so we, you know, we have these problems like, you know, India has done electronic or so electronic based voting for a long time, Estonia has done done it via even online. Most countries are, I think, naturally and sensibly going very slowly on this. And of course, you know, the pace at which different countries go is is always going to be a matter of public debate in Iceland. We have moved fast in a lot of ways we have a small population. I mean it's, you know, we're less than four lack of people. It is, you know, it is very easy to change the nature of Icelandic society very very quickly but at the same time, we tend not to rush into things that fundamentally change the nature of society very quickly so. So it's a nice balancing act between, you know, liking the new tech and being very in favor of new tech but also being somewhat conservative and a lot of the debates I have in parliament on technology mostly comes down to this thing exists it is new it is maybe somewhat scary but it's coming whether we want it or not. And the question is how quickly do we regulate it how quickly do we do we adopt it into, you know, into some kind of legal practice and you know what what kind of limitations to be put in force. So for instance, there was a change to the money laundering law a couple of years ago, which basically expanded it to include cryptocurrency based money laundering. What I made at the time was the definition that was being used was overly broad and also in specific in ways that might limit the use of blockchain technology for non cryptocurrency or non monetary purposes, such as maybe voting. I criticize cryptocurrencies a lot. But, you know, if we're going to regulate them we need to do it very slowly and very deliberately and make sure we don't make mistakes. So, as a technologist who happens to be in parliament for a spell. I end up trying to explain technology to people and you know it, it sometimes it works and sometimes it doesn't, but it's always a big mess. And that answers the question. But there's a question here about whether I own my phone number in Iceland. And how, how does ownership around the phone number across different countries manifest as a correlate to identity. So, honestly, I don't know. I suppose the numbering scheme is just a public resource and, and it gets allocated to me for the duration of my tendency. But so the question is that, suppose you stop using a cell phone and they say you want to have nothing to do with it. Can that number go to somebody else after 10 years. Yeah, that is a problematic thing, you know, so what why you will leave your digital footprint somewhere. Yes, confused. Yeah, yeah, absolutely. Yeah. I just honestly haven't thought very specifically about this so I don't know how different phone companies do this here. But I guess, are there any other questions I don't, I also have lost track of time a little bit. Yeah, I think that we have five minutes to close. Okay, okay. Yeah. So last chance if anybody wants to come in with with a question. I thought I can ask you maybe one more. And just about because you mentioned statistical testing, whether you have any opinions about using real time or time lag statistical sampling as a way of ensuring verifiability. Essentially, as the vote is being conducted, people are taking statistical like essentially something like a jackknife or bootstrap type sampling of the paper data and comparing it to the stored data. So in India there are logistic issues with that, you know, so to do something like that you need the internet, perhaps and I think that in many, many polling booths there are no internet there's no possibility of an internet in India. We conduct elections in pretty remote areas. The other difficulty with that is that our elections are over a long period of time. So you cannot really announce results or do any auditing till the all the phases of the elections are over. So to hold the elections results. Such a large countries of Bengal election, for example, happened in seven phases over a month and you are not allowed to do any auditing in between. So I think there's a logistic reason. The statistical auditing online is a possibility otherwise theoretically there is there's nothing to stop it, but fundamentally there are logistic reasons to stop it. My only concern with statistical audit is that do it correctly, you know, so, so don't come up with ad hoc means to be clear and statistical audit is the cornerstone of an election. Whether it is paper based or whether it is electronic and that step cannot be sidestepped or cannot be, you know, you cannot use uncertain measures for that. No, exactly.