Security, IoT, AI and the Built Environment
April 12, 2017 | Blaine Wishart
Uber is an early example of an Internet of Things (IoT) company. Sensors collect information about location, traffic flow, passenger goals, etc. and use it to augment driver plans and decisions. The resulting increase in functionality has allowed Uber to put more cars on the streets of San Francisco than there were taxis a few years ago.
That type of early success has many wanting to put various devices—sensors, controllers, and embedded computers—everywhere. Over the past few years, estimates of the number of Internet-connected devices have varied widely. Currently some estimates are as high as 30 billion by 2020 compared to 7-16 billion today.
Visions of smart buildings, smart cars, smart cities get technologists—and companies that sell technology— excited. Virtually every major Internet-based company that has emerged over the last 25 years and many old-line companies like GE and IBM have big plans.
Perhaps more significant than the number of estimated devices is the assumption that having billions of Internet-connected devices spread throughout the built environment is a good thing. Indeed, wonderful things seem possible.
Buildings can respond to their occupants
When coupled with evolving batteries and solar power, energy management can be effective and allow for significantly more sustainable buildings.
The entire built environment can become smart. Homes, work schedules, offices, personal and social transportation can all be optimized for cost, time, sustainability, etc.
That is one dream, at least.
Sharp curves ahead
Last November hackers attempted to hold San Francisco’s MUNI for ransom. They failed this time, but the system was shut down for hours.
Before the winter was seriously cold in Finland, an apartment building was held for ransom—with all heat turned off.
Last October a distributed denial of service attack (DDoS) closed several key Internet sites (including the financial service company PayPal) and most leading newspapers.
Consider those types of events in a few years when (if) Internet-connected devices permeate the built environment.
- Instead of losing access to Twitter, apartment dwellers lose access to elevators. This would be tough in NYC, Tokyo, London, etc.
- Instead of losing access to PayPal, commuters lose access to their cars and their Visa cards.
- And so on …
Let’s think about the part of the built environment closest to A/E/C: buildings.
Should we work to make them self-protective before trying to make them smart? For example, is energy and communication self-sufficiency important? That may mean designing battery backup systems and maximizing solar and wind power in case of power outages before inserting sensors everywhere.
Should buildings provide minimal ISP type services to their occupants so that when commonly used domain name servers go down, the building stays up? That may be a precondition for any responsible IoT integration at the building level.
Do building designers and owners have responsibilities to protect the privacy of their occupants? Perhaps instead of buying whatever the big players offer, A/E/C should work to design energy management and other IoT systems around the needs of their occupants rather than the needs of Internet companies.
While DDoS attacks and server breaches get most of the news, there are more sinister possibilities in the built environment. One type makes use of the complex supply chain involved in most technology. The best-known example is the Stuxnet attack against Iran’s centrifuges. Consider thousands of buildings with complex Internet-based control systems. A successful attack need not even target a specific building or city, but simply wait for the first exposed installation in a 50+ story building.
Recent breakthroughs in hardware raise the possibility of CPUs which can pass rigorous security testing upon being manufactured and installed in hardware, but secretly reconfigure their hardware to allow remote attacks.
Right now, the Internet itself is insecure. Do we really want to design buildings on such a fragile foundation?
Rethinking the I in IoT
Internet-based technology is not the only approach for combining sensors, actuators, and embedded computers to make smart buildings.
Even Bluetooth can be used to build non-Internet networks, but other more resilient networks are also possible.
Perhaps most important is that the architecture of such networks does not need to be server-based. Peer-to-peer networks are pretty well understood. They can, as a unit, still connect to the broader Internet, but they are much more capable of being self-protective, of protecting privacy, and of being isolated from Internet hacking.
Since the Internet core architecture was developed about 50 years ago, security technology has moved forward. New approaches such as blockchain-based security are under active investigation.
Machine learning and Artificial Intelligence
Almost everyone who uses WiFi is familiar with the often-perplexing time and uncertainly involved in making a connection. Recently researchers in China used machine learning (ML) to decrease both by a factor of 10. That suggest the possibility of learning systems to improve not just building systems, but the systems that operate them.
Unfortunately, the same ML can be used to find vulnerabilities. This reality applies to every topic in this paper. For some, it may say: “Don’t worry. ML and Artificial Intelligence (AI) will make our buildings safe.” For others, it may speak to the value of slowing down IoT adoption.
Wrapping it up
Not only the future or the present, but also the past is a concern as billions of hackable devices are already deployed.
One class of problem is malicious attacks on buildings, vehicles, and infrastructure. A second class of problems is hacking the computing and communications power in the built environment to mobilize malicious bot nets.
Two types of protection are frequently proposed:
- centralized control and standards
- resilient, decentralized, self-protective systems
Dependence on manufacturers and Internet vendors may be misplaced. They don’t always have the competence or sense of responsibility to safeguard their own equipment and systems. AI has the potential to help secure networks—and not just TCP/IP networks—as well as the ability to hack them.
Blaine Wishart is a senior principal of the Strategic Technologies practice of DI Strategic Advisors.