Perhaps it’s because of the thematic thread that I’ve been following at AusCERT today – along with some of the people I’ve spoken to during the course of the conference – but its hard to escape the conclusion that the “Internet of Things” will create a host of new attack vectors that will probably only become clear after we have enthusiastically adopted a new technology: that’s the way it always goes.
Adoption always comes before mitigation. Google is registering hundreds of thousands of Android devices each day, as Murray Goldschmidt of Sense of Security noted in his talk.
And it’s easy to see how adding smarts to cars will be irresistible to the corporate sector. Can you, for example, imagine a CFO avoiding using the information available if the car helps reconcile expense sheets? All it needs is for the vehicle to have an operating system that plays nicely with business IT systems – like Android – a suitable app, and a WiFi connection.
That kind of capability might even be completely invisible and seamless.
However, we lack any good way to deal with such things from a security point of view – not just because of the problems that Goldschmidt and Tim Vidas talked about in the Android ecosystem, but also because security capabilities are lagging so far behind the threats.
When people discuss endpoint security, they’re not talking about a new discipline. In the first half of the 1990s, a Unix expert of my acquaintance working for then-major Australian technology company Softway, said there was no point in ever trying to secure “a network”.
You can never guarantee that you have blocked all malicious traffic on a network, he explained, but if every single host on the network – in modern parlance, endpoint – is secure, then the secured hosts will ignore the malicious traffic.
In discussion with various vendors here at AusCERT – Palo Alto Networks’ Brian Tokuyoshi and Tal Be’ery of Imperva among them – the idea that a network of perfectly secured hosts is a secure network still holds true.
It’s securing the hosts that’s the problem. For example, Tokuyoshi said it’s impossible to secure an endpoint you can’t see and don’t control. In the current enthusiasm for BYOD, “you can’t see the endpoint because it’s in the user’s hands”. And, as Be’ery said to CSO separately, there are just too many endpoints; so individually securing each device would become an impossible burden.
And that’s only talking about the endpoints that people believe need to be secured. Many devices have a very small footprint in peoples’ consciousness. They’re there, but we don’t think about them. When Paul Vixie discussed the DNS Changer attack in Thursday’s keynote, he noted that broadband routers had been turned into part of the attack. They’re similar in design, frequently use identical code for key functions like the Web management interface, and if compromised, a simple change to the configuration that points the user to the wrong DNS will be invisible to most users.
An infosec professional will remember that a simple broadband router is an endpoint that needs to be protected, but many or most people do not.
Which returns me to the starting point of this discussion. Companies are already struggling under the burdens of BYOD, and we’re only a handful of years away from seeing the beginning of the next explosive multiplication of devices. Is the security business ready for a world in which the attack vector is a Mercedes?