Original URL: http://www.theregister.co.uk/2014/03/24/interview_ciscos_chris_young_on_internet_of_everything_security/

Interview: Cisco's security supremo on the Internet of Everything

El Reg asks Chris Young how we can stop the IoT becoming a $19 TREELLLION honeypot

By Richard Chirgwin

Posted in Security, 24th March 2014 00:58 GMT

Among his many responsibilities, Chris Young is the Cisco executive charged with leading its security challenge. Last week at Cisco Live! Australia, Vulture South talked to Young about securing the Internet of Everything.

El Reg:Cisco has put a $19 trillion value on the Internet of Things: how do we stop it becoming a $19 trillion honeypot?

Young: What's important to know, at the end of the day, most of the behaviour that we're seeing – most of the risks and challenges of the Internet of Everything mirrors the risks and challenges of society. When I think about cyber-attacks and cyber-security issues. At the end of the day, what are people doing?

They're stealing money, they're stealing information, or they're trying to disrupt someone's operations. Those are all problems that we see in the physical world. It's just magnified and scaled in a way that we can't contemplate in our own physical world. But all the motivations are human, at the end of the day.

It's certainly reasonable to expect, as monetary opportunities are created, [that] there will be criminals that will seek to exploit those opportunities for their own gain.

That's going to happen as we see the evolution of IoE.

El Reg:Canonical's Mark Shuttleworth [recently] wrote that proprietary firmware is an incurable problem ... no matter how invisible the firmware looks when somebody writes it, and blows it onto silicon, sticks it in a home router, sends the router out to people.

It's only the determination of the attacker, whether somebody decides “I'm going to see if there is an embedded factory password in that particular device”, they'll find it.

How feasible is it to say “lift all the value-add one step away from the firmware, make the firmware a visible and open set of API handles, and stop trying to believe that the firmware is the foundation of our value-add.”

Young: The reality is – in security, a lot of the security problems, the criminals are going to go after where they're going to get the most return for whatever they have in mind. I focus less on vulnerabilities of any given element of the stack, if you will – hardware, software, etcetera – what's important is that there's always going to be vulnerabilities in any product that exists.

A good example – some vulnerabilities in products that could be exploited for security purposes, are things that have been put there so the product could be tested.

The point is that the context becomes very important in thinking about the security model. If we follow the context, we have to follow the value if we want to understand what we need to protect. That's really important.

Most of the movement in the industry right now is that we're moving to more software-based models, more value in software. Hardware's still important because you have to get performance and scalability, but I do expect that we can get the right level of security, whether it be on consumer devices or others, by thinking about where attacks might happen, and being in a position to mitigate that.

Do we need updates?

El Reg: If we look at the kinds of devices that are expected to prevail in the Internet of Things model, a lot of them are going to be small and not very smart. A lot of the stuff hasn't been created yet. Now that we have an opportunity to do the architecture before we do the product: if we're looking at a sensor that we're expecting to last two years on battery, talking over RPL and doing a very limited set of functions – doesn't it make sense to say “we only need a tiny bit of execution on that device”, and everybody would like to have updateability, but that's not a big value point.

Can't we just say “the best way to stop someone hijacking the device” is to freeze it. If we then want to change it – it's got to be a compelling enough reason to go out and put a new one there. Doesn't that make more sense than the security vulnerability that says “if we can push software out to those things, someone else can, too”?

Young: This is the conversation I have with customers all the time. All context is not the same – you have to optimise your security model for the business context, and the environment in which you operate.

If you've got a device that's 100 miles offshore on an oil platform, then yes, you're going to need to be able to remotely manage and update that, because you can't physically go and swap something out. But if you've got virtual machines in the data centre, maybe you want to kill those every night, and reboot them in the morning with a “golden image”, so you've got really good certainty that nothing bad has happened to that machine.

The context in which those two different environments operate is very deterministic around what the security model will and won't allow.

It's going to depend on the context – on what kinds of use-cases you're talking about. You talked about very small devices. For small consumer, wearable tech, if you've got a compromised device, maybe the best thing to do is treat it like a compromised credit card number.

As soon as the bank sees you've got a compromised credit card, what do they do? They send you a new one.

The Internet of i-Things

El Reg: The smart watch will get bought because it is cool. It will get connected to the smart phone, because that's the model that Google wants, that Samsung wants, that Apple wants, and so on.

Is the industry setting itself for a really big backlash on the basis that in six months' time, what you have is a lot more vulnerability than is balanced by the utility of the device. For a toy, my health records have popped up in Bulgaria, and most of the use-case was “wow, cool toy”. That's too much vulnerability for too little utility, isn't it?

Young: That decision – this is where I do think the user has to take some responsibility for their own interests. Security is still everyone's responsibility, whether you're talking about your personal security, your home, your business – the individual still has a role.

You can't assume that there's some way to outsource all security care-abouts. A good example, in my own home I spend a lot of time worrying about the things you just talked about – what vulnerabilities exist, I worry a lot about when people come to my house and do work, who gets what alarm code. Those are important considerations.

At the end of the day, the individual is still responsibility for their awareness around security. And it goes even further than that. Children – you raise your kids, you teach them who they should be suspicious of. “Don't talk to strangers”, “Make sure when you go on a class field trip, you hold hands with somebody in your class, so you don't get lost.”

As human beings, we start to learn these principles very early in life, and there's no reason we shouldn't start thinking about how that extends to a world of connected devices and information about ourselves.

It's inevitable that we have to do that. We can't just assume that because there's this technical world that exists in the ether, that we don't see every day, that we're not responsible for our security in that context.

El Reg: And here, we get an easy illustration, or at least some scope of that boundary. One of the solar equipment vendors, in every respect held as the gold standard in the industry [yet] every piece of their equipment ships with an unchangeable factory default password – well, we can never connect that to the Internet, can we?

Young: In that world you put the consumer in the position where they have to choose not to use that product, or live with that problem.

This goes back to my point that every company is a technology company, and every company is a security company. Here's the thing: in Cisco, I'm responsible for the security business, security is my vertical. We sell security products that help a customer solve a security problem.

But I also am responsible for security as a horizontal, so I have teams of guys that work with other groups – the data centre, enterprise networking, service provider networks – to ensure that we have some basic security practices about how we develop our products.

[This is] the Cisco Secure Development Life Cycle. We train people, security ninjas, we certify people and we teach our developers to build software and our hardware products with basic security tenets in mind, like not hard-coding passwords.

That's the kind thing that every company is going to have to do, particularly as they start to think about connecting their users and devices to some broader ecosystem, connected to the Internet.

Everybody's going to have to follow secure development life cycle. Everybody's going to need basic, foundational security. Identity is going to become important in all of this – and that's not too much to ask of any vendor, of any product.

Beyond the MAC address

El Reg: Let's look at the identity part. Conceptually, how do you expect identity of devices to evolve? At the moment, the nearest thing we have is a MAC address, really.

Young: That's probably right.

El Reg: The MAC address probably isn't immutable … but it's close. How do we say “we want to associate with the device, with other things we know about the devices, with what we know about the device should be doing,” so we can say “that device shouldn't be calling Bulgaria on port 443” or whatever?

How do we extend what we now think of as a device identity to a security-useful identity?

Young: It's a great point. Think about it this way: my name is Chris Young, but my identity is so much more than my name. It's the company I work for, the people I work with, the places I go, the things I do – all of that constitutes my identity.

Why would we assume that in the device context, where identity becomes more relevant, that the only thing you need is a name?

Even if you had a unique identifier – think of a social security number as a parallel – that identifier is only a facet of your identity.

For human beings, we've gotten to the point where we enrich the name with all these other attributes to ascribe someone an identity. We haven't yet gotten to that place for a machine – we just think of it as a MAC address, or an IP address.

What we're going to have to do, as machines become more connected, and the context in which they operate becomes more important, then we will have to ascribe other elements of identity to machines, in order to be able to make better decisions about what those machines can and cannot do.

That could be: “This is a machine, it has this image on it, it belongs to this group. Its normal behaviour looks like this, so if one day it behaves like this, then we have a problem”.

El Reg: If we're talking about a sensor that's nailed to a telegraph pole, it probably should not be sending out 40,000 e-mails in an hour.

Young: Yes. Think about it, we apply those principles today. Your credit card company is building an identity profile on you, and if a transaction happens that's outside that profile, what do they do? They block the transaction, and say “call me”.

“Were you at this gas station?” There's no reason that the same methodology could not be applied to machines, particularly machines that could have behaviour that could result in a data breach or some other malicious act.

Refrigerator permissions

El Reg: How does the industry also get better at the granularity of its disclosures? If I put an app on my phone, I will get maybe three advisories saying “memory, network, and power” and not the option of turning those off.

Young: You can make those determinations for a lot of apps …

El Reg: But when we broaden it out to the refrigerator or the washing machine ...

Young: It becomes too complex. That's why you can't be deterministic, you can't rely on the individual to make these decisions. You've got to have profiles of behaviour, and you've got to have visibility to behaviour and context.

Then when something gets out of the tolerance, the right profile tolerance, that's when you take corrective action.

I think that's the only way to deal with it, because if you get to the world where you've got to make all these decisions about “what can your refrigerator do, what can your washing machine do, what can talk to what”, you'll never … it's just not possible. The permutations just in your home are too complicated already, and we haven't even started.

We've only connected one percent of those fifty billion devices.

El Reg: Already a lot of the upstream data gathering is simply the Facebook model – you didn't buy the product, therefore you are the product. In a lot of cases, what seems to be happening is “fill an enormous data centre with the diaries of light-bulbs”.

For the data collector, what on earth is the value model that they hope to get to?

Young: That one's harder. It's hard for me to speculate …

El Reg: Go ahead, speculate!

Young: ...energy usage patterns, using that as information on the transaction of energy rights, when to provision energy. Just knowing a little about the power grid, there's massive amounts of money to be made or saved depending on when energy gets delivered.

It's much cheaper to consume electricity in the middle of the night. There's use-cases like that where I can see why the data might be more valuable than it seems at first.

But there are security implications to that, and it's all the way up and down the chains. The security on the device, the security of the data transmission, the identity and context of that data, the comings and goings of the person – those are things that are all very relevant security considerations when you consider the kind of information that's going to be generated.

El Reg: This isn't really a Cisco-specific question: it's really easy to do the big data stuff, have a good picture of circumstances, and get the wrong answer anyway. Are we even remotely ready to get the right answers out of the data when it really matters, or are we still at Base Camp One on Everest?

Young: I have found in my career that the right answer is as often a function of the right questions, as it is of the process and the output. So – the challenge with Big Data today is that it's become this buzzword that people focus on.

But if you've done any statistical research, you'll know that if you aren't asking the right question you won't get the right answer. ®

Disclosure: The author travelled to Cisco Live! as a guest of Cisco.