Joe FitzPatrick has spent his career digging into the lowest levels of computer hardware, figuring out how things work and how they fail. He knows the kind of mistakes that manufacturers make thst lead to vulnerabilities, and he understands the ways in which an attacker or someone along the supply chain could insert a backdoor. Those are two different things, but can look quite similar to the untrained eye. FitzPatrick discussed the questions that observers should ask when presented with assertions about a hardware implant during a talk at last week's CanSecWest conference, using figures from Greek mythology as guides. This Q&A has been edited and condensed from a podcast with FitzPatrick.
Dennis Fisher: How did you come up with that concept for using Pandora, Cassandra, Aristotle, all these mythic Greek figures to explain hardware implants?
Joe FitzPatrick: I've always been a fan of Greek mythology and there's a lot of stories that explain things in a very similar way to the things we encounter. The two that I already picked out, Pandora, she's curious, she opens the box and how different it is what she did versus, you know, what anybody does when they inspect something and find vulnerabilities. The recent news kind of inspired me to think about the coverage and the reception of some of this information. Cassandra, she's gifted with the gift of prophecy, but also cursed with the fact that no one will believe anything she says. The two of those kind of summed up a lot of my feelings the past fall, when people were making bold claims about hardware implants. But what I didn't see was the rational approach to, you know, understanding what's real, what's not. My background is hardware. I can look at a report of some hardware risk and, and make an informed guess about it, but the vast majority of people don't have that skill set.
Dennis Fisher: I liked the way that you use Pandora specifically to talk about the curiosity thing and you mention curiosity is essentially the core ethic of hacking. You've got to have this base curiosity to even get started. You have to want to know how things work, how they fail, how they succeed. Is that kind of why you got into hardware hacking in the first place?
Joe FitzPatrick: I was fascinated by computers. We were lucky enough to, to have several old PC clones, which I played with. I didn't really get into programming until later than all the people who started programming when they were five. But, I was always fascinated by what made the next level work. And I kept going down through software, you know, I learned more about assembly and then I kind of learned, you know, conceptually like CPU architecture and that really fascinated me. And rather than pick any one of those levels to stick on, I just keep going deeper, which kind of got me where I ended up, which, you know, I started my career doing tools for silicon debug for our server and desktop CPU, which is kind of pretty low level when you get down to it.
Dennis Fisher: There are these debug features in chips and other pieces of hardware and they're essentially back doors in that that's what they're meant to be. They need to function when everything else isn't working right. And you can't really differentiate between a malicious backdoor and debug feature. Is that a concept that you have to explain to people fairly regularly when they're thinking, oh, this is a backdoor?
Joe FitzPatrick: I'm thinking back to a presentation and I don't recall the presenter or the exact detail, but he showed basically the proof that profit and theft are functionally identical. Right? You know, you input stuff to a system, you take stuff out of the system and you get money in the process and, well, that's profit. When someone is unfamiliar with the way of low level hardware and a low level debug, right, it makes perfect sense that they would see a back door and think it's a back door. But then when they have a little more understanding about like, why this functionality has to exist, why in order to bring up a piece of silicon, you need the ability to debug the silicon in order to debug the silicon. You need observability into the silicon. You can't have one without the other. You can't have a manufacturable, sellable product without having a way to debug it. So that's kind of the detachment. One of the things I pointed out, if you encounter something at hardware and you don't understand it, find a domain expert. Because the domain expert, and I don't mean like top of the field of management, it's just a hardware engineer or someone who has built network adapters before or something, you know, who is familiar with the components you're dealing with. There are lots of people who aren't security people, but know about this. What's interesting is, whether we're talking about implants inside of network adapters or we're talking about Rowhammer, right. People who make memory cells have understood Rowhammer for ages, but they didn't put it together as like a software accessible vulnerability. Until we got to put the things together, glue it all together, open the box and take a look at what's inside.
Dennis Fisher: More broadly, how big of a concern do you think this whole supply chain attack thing is? To me it seems like a very big concern, depending on who the customer is and what their threat model is. But, it's something that people in the security community definitely think about. But I don't know how much the broader government and enterprise security communities think about it
Joe FitzPatrick: For a very long time, when I would talk about hardware security, people would assume that I'm talking about supply chain security. But that supply chain security from the government side, the military side of like, oh, we need to make sure we get these tools from this source and that they're reliable and everything. I get a very different approach to supply chain security. My perception of that was like, we have a bunch of people who've been thinking about supply chain security for hundreds of years. And they're looking to apply their approach to think that that's going to solve hardware security and it won't.. You know, the example I use is we could have a trusted network adapter made by trusted people and our trusted fab and putting on a trusted board and shipped by trusted couriers are installed by a trusted person.
But if it's network adapter that takes firmware updates over the wire, we still don't have secure hardware. How big is the problem? It's huge. Like I think the great part about recent news is it's making people aware of this. I've been trying to point out the reaction is what we need to be careful about. You know, the knee-jerk reaction is not good. The, Hey, we need to talk to our supply chain. We need to find out where our servers come from, where our boards come from, where the components on the boards come from. Up until this point, no one's really asked those questions in volume. Consumers haven't asked those questions. If they start asking those questions, then the suppliers are going to have to come up with answers and processes for documenting this stuff. So we're in a spot where we don't really have visibility into supply chain, but we have that from an accountability in a counterfeit avoidance perspective. We don't necessarily have that from a tamper maliciousness perspective.
"A motherboard manufacturer's not going to backdoor all their clients because they're going to lose every one of them if it comes to light."
Dennis Fisher: How much can enterprises really do about that? I mean, they have to rely on the companies that they're buying this gear from to go and inspect these plants and hope that they're vetting their employees and nine levels down the way you were describing software down in the hardware.
Joe FitzPatrick: What has worked so far essentially is the consumer demand and the commercial reputation. You know, people always talk about China and China's doing this. China does not want us to stop buying things from them. Right? They need our money. They're not going to do this, you know, flat across the board. A motherboard manufacturer's not going to backdoor all their clients because they're going to lose every one of them if it comes to light. I don’t know what Supermicro’s position is right now, but they've been damaged by this news. We still don't even know the details of what actually happened. The fact is that people in the end are looking to deliver products and looking to deliver secure products, whether they explicitly state that or not, that’s the economic incentive to keep this supply chain clear.
Dennis Fisher: Manufacturers even in countries that are controlled by an authoritarian government aren't going to backdoor all their stuff because they're selling things overseas. Unless they're ordered to, which seems, self-destructive for the government. Joe FitzPatrick: Exactly. You know, if you're an international company and you're ordered to backdoor everything, then this is the same argument that came up with the whole Apple case, right? Who's allowed to get those back doors? It's a slippery slope. It's unfortunate because it seems like the world is moving in that direction and it's going to be a big change from the past 20 years of free trade and everyone trusting everyone else's hardware and software to do the things they're supposed to do. So it's gonna be interesting.
Dennis Fisher: That erosion of trust is going quickly. It's not gradual anymore. It seems to be going off a cliff.
Joe FitzPatrick: Yeah. You know, I don't know where that's going. And I'm reluctant to give any time to arguments of national influence. Like, oh, we can't buy X because it's made by Y. Because once you start to that conversation, there is no end to that. Like you basically will realize if you want to actually follow through, you have no choice but to only purchase domestic things and that's going to be a very different world than what we've got right now.
Dennis Fisher: One of the other things you mentioned in your talk is this idea that yes, hardware implants exist. We know of examples, there are groups that can do this stuff. They mostly reside inside intelligence agencies. Probably like the NSA. We've actually seen their catalog of toys, which you mentioned. I would assume there's other foreign intelligence services that can do this stuff. Do you think that there are private teams that have this level of capability as well?
Joe FitzPatrick: I am certain that there are teams that have the same capability. I don't know how it compares. So, a good example of this is gas station and ATM skimmers, right? These are people who are criminal organizations that whether they're designing it or outsourcing it or contracting to someone to do it, but designing hardware implants their design is about devices that an unskilled worker can go and cram inside an ATM or a gas station pump in 30 seconds. And then they're getting people who just show up and you know, download a list of all the cards that have been scanned by Bluetooth. And we're talking about five dollars worth of hardware. It's very different from the ideal that people will imagine where you have some, you know, elite tactical person who crafts something themselves and goes and opens up my machine and spends 20 minutes finding the right wires. It's plug and play.
Dennis Fisher: And that's a volume business. Like if it's five dollars worth of hardware and it's unskilled labor to install it, if you developed a process to build these things, you build a bunch of them, you send these folks out to install them on a hundred gas stations or a hundred convenience store ATMs. It's not just a one off hit.
Joe FitzPatrick: Yeah. And I mean that's just that realm of things. What's really cool to me from a hardware designing perspective is the capability of what you have in such a small space anymore, right? You know, you can have a couple square inches or you know, fraction of a square inch and you can fit a full stack Linux operating system. Chances are you're implanting it in someplace you can already grab power from. And a full GSM radio takes another, you know, half a square inch. You can fit a lot inside of everything.
Image: CC By 2.0 license image from Texture X.