PQC: Post-Quantum Collaboration?

Transcript

[music] [music] >> Okay. Hey everyone and to start first I'll tell you I'm Vijay and I'm one of the co-founders of Fileverse. Andres was also supposed to come but he had some engagement and couldn't come. And before I start how many of you here know about Fileverse? If not I'll just give you one liner and that's it. Okay. And for those of you that don't know Fileverse is like end-to-end encrypted workspace and collaboration protocol.

Think of it as us like as signal but for collaboration and workspace environment. So let's start. I'll go ahead. Yes. If you guys have built your apps like you already know this that internet is changing a lot and very fast. So what your apps need to do and even you need to do is to adapt to this environment to this new environment which has a lot of issues like as we are exploring you already know about the PHP hack that happened. You also know about the recent CVE that was released which

was affecting all the you know deployed boxes and you just needed to send a small TCP payload to get access to the root privileges of the box and so on and so forth. So like there are two big threats that we all see. The first threat is the quantum computers and this is not a new threat. It's something that we all have been preparing for that we all know that there are advances that are happening in computing space and so on and so forth.

And for this the main thing that it threatens is the encryption. Okay? And encryption being like all the keys that rely on prime number generation and so on and so forth for their internal working. RSA and other important ones being one of one of them. And uh e-elliptic key cryptography as well, elliptic curve cryptography as well. And this was something that was also mentioned in the DevCon keynote. Then there is another thing that came out of the blue. Uh I won't say out of the blue, but still like it's something that we did not anticipate as uh like people that were working in the web space. Uh it's basically malicious actors using AI to chain together

vulnerabilities and then gaining root access to either your app infrastructure or gaining root access to your accounts or so on and so forth. And then there is another thing that came up, which is cryptanalysis this crypt- cryptanalysis advancement that happened and there are a lot of breakthroughs that are not yet even published in the public space because of responsible disclosure and so on and so forth. So, yes. Now, the case that I'm going to make in this presentation, it's that if you protect against one of them, you're also protecting yourself against the other.

And both of these threats are sort of uh handled uh if you uh keep some of the key primitives in mind. And before I go ahead and explain the solution or the architecture which you can utilize in your apps, I would uh go to the making the case for the threats. So, uh you already know and if you guys are in security or if you would have seen uh on the internet everywhere uh this like this CV graph that you see, this it has gotten like uh an exponential growth.

And why did that happen? It's because AI agents have gotten smart enough to understand all the different kind of like old vulnerabilities and can patch together and increase the threat level of that vulnerability. So, and uh one more important stat that is here, which is 3% of all the critical bugs were disclosed in the last month itself on the CVE. Uh and another thing and it's not just something one It's not just things that I'm saying and that's why you should believe it. You can also go on the internet and you can see uh Matthew Green said that I think we might lose lose public key cryptography. And I

think the reason behind that is the cryptoanalysis uh breakthroughs and also the vulnerability uh chaining for stuff. And then Justin Drake also said that I think we all need to start living in a paranoia mode and uh bunker mode uh where we uh like don't show any of the public keys to anyone, no matter how much we trust, because it can be reverse engineered. Now, right now, there is one big thing that is uh like causing the problem. And in the threat uh thing I showed you that AI causes two types of problems. The first problem is

the vulnerability chaining and for that the centralized servers are the biggest uh you know, honey pots that can be uh what do you call that can be compromised. And uh most of the servers either have the personal data in the plain text because they assume that the server is safest place and uh so on and so forth. Or uh they they have their keys as well uh stored on those things. Now, in this AI hack into everything world, I don't think that's a good assumption and that's where we need to change stuff.

And then there was this for the Q-day and why did I have that as an important threat? There is this new uh publish uh report which uh Google team did and uh they said that uh earlier they were able to simulate some qubits uh but that number was like, you know, higher than what they expected and that's why they updated their internal PKC deadline from uh 2030 to 2029 which is 1 years of advancement just be like and even at that time like even the AI related stuff did not was not as advanced as it is today.

And you can also take a look at the uh NIST uh that NIST timeline where they recommend when by when you should uh become PKC compatible. Now, how do we adapt? So, I told you what are the threats and it's not just to I did not just uh come here to speak about and making you scared. Uh it's also to uh show you an alternate mechanism and what are the different areas that you can look at for inspiration. Now, the privacy apps in a have an advantage here. Why? Because these apps were built with this assumption in mind that the service compromised or server is going to be compromised one day or the other.

And uh it treats that environment as a hostile environment and it does not put any kind of plain text data on it and it does not put any kind of single point of failures in the system. And uh there is one more thing which is like even if these servers that these privacy apps are using are breached, what the uh attacker will have is just a encrypted set of cyber ciphertext and nothing else, right? And there was this quote by Vitalik as well that he recently where he said that treat all the uh like we really need to treat data as a toxic waste and we need to make sure that we dispose of it properly. We need

to really make sure that whatever the life cycle of the data is, uh you handle that properly on the side of implementation. Now, there is That's why I was saying about the privacy apps and you can take a look at any of these apps and uh for example, Session, uh Signal, SimpleX, uh ProtonMail and FileVerse. All of these have end-to-end encryption by default. And this architecture, I think, it's going to play a key role uh while we take care of either the defense setup for AI cryptanalysis cryptoanalysis things or uh to all of these threats that we are

talking about. Now, what do you need to make sure in your apps? Uh you need to make sure that server only sees the cipher text. You also need to limit visibility of the data so that if the like something is compromised, it's not the user data that is at risk. It's just the availability or uh reliability guarantees that are at risk. Right? Then, you should also favor uh local first architecture. And why local first architecture? It's because in the hostile environment where nothing uh like everything is going to get hacked by AI, let's say we have this assumption then we need to make sure that the users are still able to continue with their work, continue to have the have the digital

life intact uh and so on. So, favor apps or create apps that can work with the assumption that the server is not available. Now [sighs] so recently, like because of all of this, we uh at FileVerse, like we have end-to-end encryption and because of this, we also said that let's go one step further and we should also have the post-quantum uh like cryptography implemented in FileVerse and not just have the RSA because we work with those assumptions.

And what we did is like we are using ML-KEM-768 and like this X25519 just for confirming who that person is. But for the it it works in hybrid way, meaning you need to have both of these things correct for you to be able to either access the system or do changes to the system. Now, what does that mean? That means your quantum computers or AI both will need to break both of these things or get nothing.

Now, if you put all of these things that I told you together, what do you get? And it's something that we have worked on and heavily iterated on. It's Fileverse workspace where you have docs, sheets, and drive. So, you can start by using it and you get all of these guarantees that I told you in the previous things for your digital life. And it's end-to-end encrypted by default and post-quantum by design because of the recent upgrade that we did. The client is a local first client and data in unencrypted format never leaves your system. So, we are also local first from this angle.

And we want to onboard the world and this is also something that you should look forward to, but anyways, I'll move on. And there is this like the there is one emphasis that I wanted to do, which is on the hybrid mode encryption, which is you should use basically different algorithms so that if one algorithm is breached, you are able to you know, you have some time buffer for you to rotate those things because at the end of the day, that's what matters.

Then, it's And what does this mean for the work space like we have things like you need to prove membership by generating a semaphore proof. You need to basically in this case the server never learns what the identity of the person is and why that's important and it's also important for your apps as well that your server does not know the identity of the user is because it reduces the risk on that one person. So imagine this that you have one person that is using your app which is let's say X Y and Z and this X Y Z is targeted by the hacker and if they get the data of that particular server they can actually

reverse engineer and just focus or target all of their efforts on decrypting the data of one particular account and so on and so forth and I think this is a vector that you should keep in mind and why it's important that even your server doesn't know what the identity of the person that is using your app is. And in all of this the most complex part that we had to solve was end-to-end encryption in the collaborative space and for that like we have done a lot of research and finally implemented it into Fileverse as well. If you have any questions I'm happy to answer and all of these follows all the principles of crops and that we have heard or seen everywhere.

Now I'll go to the I think this is the final slide but before I go I'll just say mention one more thing. In this case like we treat every device that you have that you connect to Fileverse as a client and they are just two collaborators and there is no special privileges that come with it and that's that's the reason why we are able to say that it is a local first app and it works local first. So anyways and this is the Twitter of Fileverse and I think you guys should follow for more updates and I'm happy to answer if have any questions >> [snorts]

>> as well. >> Awesome. Thank you, Vijay. We have one or two questions here already. Yeah, thank you. Thank you. Fantastic. Okay. Because D Docs runs local first and encrypts data directly on the user's device, do mobile or lower-end hardware devices experience latency during key generation or document loading? >> Uh that's a good question. Um To answer it very simply, yes, there is. But that is why we need to be very aware when it comes to what type of algorithms we're using and how we are using them and what are the

places where we are going to take more time versus others. So, to give you an idea of it, like let's say if you are doing an account generation on Fileverse, it takes a bit of time for you to create an account. Now, at those places we make sure that we go with the larger key sizes and that's where you see some latencies and it can come from the hardware itself, but we we have gone with the libraries which are like which have very less difference even if you change the devices.

And uh I think that's also part of like optimizing your app for a hybrid architecture is so that it works locally on a computer as well and also or a laptop or PC as well and also on phone. Yeah. >> Got it. Okay. Next one we have here is how does Fileverse bridge quantum-resistant document encryption with classical wallet authentication? >> Yes. So, for this what we do is like if you think about it, like let's say you you have your email, you have whatever kind of thing that you want.

It can be a normal Um, know wallet that you are using, it can be uh email, it can be a social. So, all of these things uh get connected and like in the end yield a wallet. Let's say this is the one-to-one uh thing. And what we do is like we generate a smart wallet, and under this smart wallet we add these uh wallets as signers. So, if you want, you can even rotate those keys or rotate those wallets and uh change access from one wallet to the other and so on and so forth. So, that's how we uh like uh bridge the gap. And uh once this wallet is there, then we generate the sign, and this sign is used

to generate the quantum uh resistant keys. And that in turn controls your identity, that in turn controls all of the accounts that are there. >> Okay, I like this question. Next one coming up here. How do you plan to convince everyday corporate teams to leave Google Workspace for a decentralized post-quantum uh secure hardened alternative? >> Yes, uh that's an interesting question. Uh uh So, there are two things here. One is to make sure that the UX that you have is at a level where you're not uh asking the user to, you know, cut their veins to start using a privacy

alternative for whatever service they're using, right? So, let's say if you're using Google right now, uh you can just go to Dedoose, put your email, uh create your account, start using it as if you never left. Or like you can you can add other people as collaborators just by their email, you can add other people as collaborators by their wallets as well. So, making sure that the UX is one-to-one, that is one important vector. Then there's another vector, which is like how do you make sure that you convince them that it's something that is secure and everything is well. So, there are two aspects to it. One is that we make sure that we are we do audits and the recent one happened

with Nethamind for the PQC upgrade that we did. And that's one. Then there is another vector which is to make sure that you also give them an option to self-host that infrastructure if they need, right? So, these are the two ways. One you say that everything is good and the other you say that if you don't like our approach of everything is good, you can run it on your own and start using it and secure it like that. Yeah. >> Okay.

Next, we got how will formal verification play a role in Fileverse? >> Okay. For this like right now, no. We mostly use the proofs that are generated, the ZK proofs, and then we like how do we do that is we have a semaphore group and on that you verify that you're part of the group and that's when you get access to the keys and so on and so forth. So, right now we don't think that formal verification is going to play a role in Fileverse.

>> Okay. I see we are running out of time. I'm sure you can approach Vijay right after this if you have any other questions. These next few questions might require a little bit more than than 20 seconds. I do see one here about who's the anime anime expert at Fileverse is. I don't know if you want to if you want to keep that anonymous and and secret. Um but yeah. >> I'll take the Vitalik one and I think just simple thing. You can go to walkaway.fileverse.io and export your keys after creating an account on Fileverse, import your keys on walkaway.fileverse.io.

You'll see why it passes file why it passes the walkaway test and the page is a static or a very simple build which is like let's say I don't think it's even in MBs. So, you just run it on your own, go to that GitHub for kit everywhere. It just works exactly the same way. So yeah. >> Great. Fantastic. Okay. >> That one I'll keep private. >> [laughter] >> Perfect. >> [music]