Verifiable Open Hardware Methodology: Building User-Centric Trust Anchors

Transcript

Next, we would like to invite our next speaker, Mr. Graham Yu. Mr. Yu, Justice, the floor is yours. [Music] Uh hello everyone. Today I'm here to talk about the open hardware verifications. As we all know the famous idea from the ether don't trust verify. This idea change how people think about trust.

Instead of trusting banks, companies or governments we cannot rely on system that anyone can check. But here is a questions I want to ask today. Have we really apply this principle everywhere? In software I think yes protocol yes but in hardware like the machine and uh the the hardware wallet I think the answer is no. So that what I'm going to talk about how we can bring ability of a verify to the hardware.

Ethereum is built on the idea of openness. Smart contract are open to everyone. Client code is published as open source. The blockchain uh the blockchain state can be checked by all through consensus. This openness and the verifiabilities at the software layer is the foundation of blockchain security. But hardware is different. The hardware still work like a blackbox. Users cannot confirm that hardware match the published design. This shows a clear gap. We can verify software. At least we put lots of lots of energy and work to build a trust on the software layers.

But hardware is hard to see. I think we do need a complete method to make hardwareable. People sometimes say that uh let's talking about some uh uh uh uh open hardares. People sometimes say we we have uh open hardares. We can we can see and check all the designs from the internet. Uh sure that give us quite a transparent of the inside of the hardware. But uh let's look closer. The files doesn't mean the devices.

Publishing design files doesn't prove the device you buy is the same. Right? There could be lots of things happened when the uh really uh the real hardware delivered to your hand. Manufacturing chain lack transparencies. Parts can be swapped, chipped, replaced and toy added. Most important user have no tools. You you don't have the tools to verify these hardwares. So most people have no way to check what they received.

So even with open hardware projects, the user experience is still uh we we we can only trust the factories. Yeah, because we don't have another things to do. That's a problem. Uh let me share an example from uh a very interesting uh guy uh called uh Bernie. I believe many of us maybe already know this brilliant guy. He create a method using the infrared camera to look inside the chips. I I already show his blog link uh on the uh powerpoints and he create a uh uh this method allow people to compare the chips

structures with a golden reference and they don't even need to remove the chip from the PCB board or cut the chip open. Normally when you want to compare or verify a chips you need to cut that opens or using some very expensive huge machine to look through like X-ray or something. So it shows in in this Bernice product it shows something very important hardware is not a black box forever actually it can be inspected but here is issues this method still need a professional and kind of expensive tools and it needs a environment like a laboratories so it's

too complex to their normal users so the challenge is not It is possible to verify the hardware. It's changed to another questions. Can we make it possible for everyone? Okay. Uh that we already know some method to check hardware but the main issues is not whether method exist. The issues is whether user can really apply them in daily life. For most people complex tools and the laboratory environment test are not possible. So we need to

design a framework that works in a simple way. I think an ideal verification system should have three key features. First one it should work in a normal place. It must run it can be run in the homes, office or workshops not only in the laboratories. The secondary rules I think it should be a layer checks that means it has a different level of checks schedules you from the uh you can you can start with a light checks or allow deeper and deeper if you need it.

And the third rules is a toothpaste and automated. Even we have some cool method to verify or cool algorithm to verify the hardwares. We still need a tool to to do this work. So to the final users, they don't care about the algorithm in the backwards. They still need a like a easy way. No, no matter it's a website or applications, they just want to use that easily. So that's the three uh principle I think the the verifi ver verification system should have. So we need to build a step-by-step system and this system should help user move from easy check to

stronger audit. Now the the final goal is is very clear. Start with a lightweight verifications and grow to full high level auditings. Okay. So let's go through this one. Open hardware should not only means file are published. It must also mean that user can check the real device by themselves. We call this idea is a user centric trust anchors. It means that the trust does not stay with the vendors. is not only proved by the factories or who

built it, who said it. The trust move into the hand of the users to everyone. With this model, user can do verification with very low cost tools. They do not need expensive laboratory or hard or hard professional skills. Also the community could share the tools data and results. When user repeat the same checks and confirm the same result, the trust becomes stronger and is reinforced by many peoples.

Here is my approach. The uh I call it a verification pyramid. It's like four layer architectures and from a bottom to the to to the top is like uh from easy way to the difficult way. Yeah. I I will go through this one by one. The first layer is called nanotech. It's a very little cings actually. Nanotech is a micro dot uh is made of nicos as octagon shapes. uh the size of the micro dots is about 03 0.3 millimeters to 0.5 millimeters and we can have the unique ID inched in

each of the dot. So this dot can be checked on the different light source. You know when when you when you print something on the surface of the items and normally it cannot be it it cannot be seen and the infrared cameras if you have a filters to filter all the the visible light source. But if if this uh unique ID or hed into the into the micro dot it can be see uh from the uh uh infr red light.

So uh how how we use this uh micro dots and I think we can uh just uh spray this uh nanotech I mean the uh the micro dots in each of the PCB board and uh since there is a unique ID on each of the uh dots and you can easily check uh by the uh very cheap uh microscopes. So here is the image I I I I took from the uh uh like $20 uh microscopes and it's uh under the five 500s times. You can see very clearly there is ID on that. So this can pro provide anti-conting anti-tempering and the batch of

traceabilities. Uh this is the most the basic and lightweight processes easy for user to complete verify by themselves. The second layer is the layout of these micro dots. You see when we produce the uh uh uh when the uh nanotech has been spray on the PCB board uh the pattern act like a natural vual coordination system. I call it a recognition matrix.

This can be done by the factories. So when the factories spray the dots on the PCB board and it's uh create a randomly and unique patterns. So people can take photos of these dots uh of these patterns and upload the photo to the uh system or applications and compare with original reference and that will output match score or reference orientations. This can be easily prevent clinging and crude temporary. Here is the two uh photos I take uh from I try to spray some uh micro dots on the PCB board and the

algorithm you can easily check the pattern of each dot. The third layer is uh more like natural ways and uh you can take a photograph from both side of a circuit board the PCB board and uh we use the power of machine learning machine vision things and the system automatically compare with the original reference images. I believe the factory could provide the uh original true images of each powerboard.

So when you take the photos and upload to the uh system to the uh applications the applications running the uh uh machine vision algorithm in the backwards and output the compare result. Yeah that result can detect the tracing alterations chip rel uh chip replacement and the solder anonymous. This little bit the complex of to the uh final user but it can be still be easily to performed at the at by themselves.

You see in the in the previous level level one to level two level three and they did they all like uh face on the PCB board uh that sometimes is not enough. So the final the final way the highest level of the audit is still be focused on the chips. So we still need to use the infra inspections. So this level is called infrared inspections. At this stage the checks are done by community nodes. I believe that uh for the final users uh they they don't have the enough budget to buy uh uh uh devices that uh computer devices

to check the uh chips with the infrared light source and the cameras special cameras but for the communities and they can do that. So this make a hidden structures or chip visibles. The image are then compared with a golden reference images and this way we can confirm that no parts has been replaced and no tries have been added. Yeah, this give us a highest level of security and provide final assurance for the whole systems.

So I call this uh it's kind of a methodologologist. I know that this not enough but let's go through this one. I call is uh variable open hardware methodologologist. Uh the core principle is uh we must move user trust away from the vendors move from the uh factories. Instead trust should come from variable. Yeah. The instead the trust should come from the mechanics to to anyone then that anyone can check. Yeah. And this secondly we need a full stack open design.

That means hardware design file must be fully opened not only published but also usable by the community to reproduce and confirm. and also for the uh the manufacturer chains and all the chains should be traceable. Yeah. And uh from the uh user side verifications and there's uh some kind of method we can use and we can choose to from the easy way to difficult way to verify the hardwares. uh at least the uh uh and uh the the the other part is uh verification to chain and uh even we have a good method I'm not uh I'm not

confident of the previous four level of the uh algorithm is the best I I think there will be a better and keep more and more other new technology coming up to help us to verify the hardwares but no matter how many uh algorithm can can can can can uh coming out. uh I think the tools is the very important uh is help people to to to improve their user experience to easy to use this way to verify the hardwares and also we can we we in the final we need need to use the uh power of the communities and communities can uh provide lots of lots of uh help to

others uh if we don't have enough uh budget we don't have enough powers and we can we can gather together. So this filling the gap in the client side open hardware ver verifications and uh and uh uh let's keep short uh in these days we uh we always talking about user experience and uh by we said a user a user experience I I think it's uh more like make user do something very easily uh not only we we don't care about the back algorithm and how difficult it and how easily is it we need to give them a

different level of the uh uh experience and a different level of the uh method they can choose by themselves. So in the previous uh uh pyramid I provide four levels and uh user can complete level one to level three verification as easily as they they can do them by themselves and a community nodes provide a level four reports similar to the smart contract audit. Yeah. So that will make don't trust verify ideas extend to the hardware layers and for more words and I do believe I I I do believe there will be more and more

new uh algorithm and new method coming out once we realize the verify uh hardware is very important things. So uh uh this is just a start point. Uh I will wait uh I I would hope that uh we we will have a more and more method coming out to help user to verify the uh hardwares. Uh uh let me introduce our uh research group. My my group is a like a mallet research group. We found in 2019 uh it's long-term focus on Ethereum core research. We have uh um several topic like consensus security P2P

optimizations and now we have put our energies on the open hardware ecosystem and uh we have many uh uh uh technicals on the machine visions and the data visualizations. uh the core uh developers of our group is based in China as Hong Kong and uh Australia and uh we establish uh collaborations with many uh universities and research institutes from the both countries.

Yeah. Finally, uh our vision is uh make verifications simple and practically for everyone's and we want to promote two-way verifiabilities between hardware and software and we want to encourage academic collaborations and we want you to join us. So if anyone uh interested in the hardware verifications and anyone you want to discuss and welcome to join to join join us and you can add my uh uh telegram and that's it. Thank you. Thank you.