Abusing the Internet of Things (2015)
Chapter 8. Securely Enabling our Future: A Conversation on Upcoming Attack Vectors
We’ve seen new attack vectors and threat agents come into the technology landscape since the dawn of the Internet. Many of the threats have evolved in a fairly predictable way. But because the world around us has come to rely upon inter connected devices at an ever increasing rate, it’s important for us to ponder the threats to our future.
So far, we have taken a look at vulnerabilities and security principles pertaining to specific devices. We’ve learnt a lot about the state of security in IoT devices in the consumer market. We know the things we are doing wrong today that we need to improve upon to securely enable the devices of the future.
Based on our knowledge of the evolving threat landscape as well as vulnerabilities that have and are plaguing our computing systems and IoT devices, we stand on good ground to be able to predict upcoming scenarios that may come to pass given the new sets of abuse cases that come along with the use cases served by IoT products. In this chapter, we will predict plausable scenarios of attacks based upon our understanding of how IoT devices will serve our needs in the future.
The Thingbots Have Arrived
Botnets comprise of a group of workstations and laptops that have been compromised and are controlled by the botnet owner. Most often devices are infected by malware sent to the victim via email, by using a phishing website, or by software worms that exploit a vulnerability. A single botnet can comprise of thousands of devices giving the botnet owner tremendous power to launch denial of service attacks on other networks by directing traffic from infected machines toward a specified target, thereby overloading it from being able to serve legitimate requests. Botnets are also used to steal private information such as credit card numbers and credentials for bank and email accounts.
As discussed in the previous chapter, in 2014, a research firm discovered over 750,000 phishing and spam messages that were sent from more than 100,000 household devices, including televisions, wi-fi routers, and fridges. The term “thingbots” is being used in the cybersecurity research industry to describe botnets that include infected IoT devices that can also be leveraged to launch attacks and steal private information.
Thingbot owners are likely to leverage the ability of IoT devices to steal information they may not have had access to previously, such as capturing private conversation from infected Smart TVs, or the ability to control locks and lighting in thousands of homes.
As more IoT devices start to come online, attacks spawned by thingbots are only going to increase and threat agents are going to have increased access to our private information as well as the ability to cause physical disruptions in our lives. This means that the categories of IoT related vulnerabilities discussed in this book will become even more of an emergency as we look into our future.
The Rise of the Drones
Unmanned Ariel Vehicles (UAV), known in the mainstream as a “drones”, are aircrafts without a human pilots on board. There are various types of drones that range from larger, military grade ones to drones that are used for recreational purposes such as photography.
At the BlackHat conference in Singapore in 2014, researcher Glenn Wilkinson unveiled a proof of concept tool called Snoopy that uses an onboard computer, a GPS unit, a GSM cellular unit for network connectivity, and attached video cameras with a quadcopter drone to capture wireless network traffic and to follow targets in a defined area. In addition to WiFi, the tool also leverages Bluetooth and RFID network traffic to track devices and their owners.
Snoopy works by tracking network probes from devices such as smart phones that are constantly searching for WiFi networks they have previously associated with. The Snoopy drone then offers a WiFi network with the same name as the one being probed. When a smart phone joins this network, Snoopy proxies the network traffic and therefore can be used to capture data being transmitted by the phone. In addition to phones, Snoopy can also capture data from devices such as pacemakers that use WiFi, as well as fitness devices and smart cards.
Snoopy is designed to capture network data and transmit it to a remote server so the owner of the drones can analyze data in one place since the tool can be spanned across multiple drones that may be spread across a city. For areas where there is spotty cellular coverage, a single drone can be deployed to hover and capture network traffic while another drone can be sent over periodically to collect the captured data and bring it back to the attacker. This tool comes with the ability to use an accelerometer to detect if the drone has been captured by a third party. In that case, Snoopy can be configured to self-destruct by erasing the contents of the hard drive on the computer attached to the drone.
Apart from researchers being able to demonstrate how UAVs can be leveraged to capture devices from smart phones and potentially life sustaining devices such as pacemakers, it’s easy to imagine how drones can be leveraged by heavily funded groups such as state governments and sophisticated criminal gangs. As UAVs continue to evolve in the military and the private space, it is quite probable that they will be used by a variety of agents to gain access to devices and networks. In this book, we have seen many different examples of IoT devices that require no authentication or authorization if the attacker has access to the local WiFi network. Since many popular IoT devices are already in existence, they are going to be a juicy target for individuals and well funded criminal agencies whose aim is to capture data and possibly disrupt physical safety.
Cross Device Attacks
Many people utilize a slew of computing devices on a daily basis: smart phones, personal workstations, employer issued laptops, and tablets to get their professional and private work done. Quite often, data is synced across multiple devices so the user has access to his or her information regardless of what device is being used. For example, the user may backup his or her smart phone on to their personal laptop. Another example is using a service such as iCloud to sync documents, application settings, and contacts across devices. This creates a situation where an attacker may be able to leverage one device that has been compromised to access information stored on another device or to access information that is synced using a cloud-based service.
Imagine a situation where a physician stores information about a patient in a document hosted on DropBox. If the physician’s desktop were to be compromised using a phishing attack, the attacker could modify contents of the document to alter the dosage of a medication. This document would have its updates synced across other devices, such as a tablet, that the physician may use while on duty. In this situation, the doctor’s tablet may be configured to have full disk encryption and additional security controls deployed by his or her employer. Yet these controls would have no merit since the document was compromised on the doctor’s desktop at home and automatically synced to the same DropBox account on the physician’s tablet. This illustrates how the compromise of a single device in a user’s ecosystem can be leveraged to negatively affect the integrity of data on other devices.
Local backup files of smart phones and tablets that may be stored on workstations and laptops are also a juicy target for attackers. In Chapter 4, we analyzed the token called access_token by the SmartThings iOS App which was issued by the server upon successful authentication and valid for 18,250 days. Should the user’s workstation or laptop be compromised, an attacker can potentially steal the backup file of the user’s smart phone or tablet and collect the access_token which is likely to work since it is valid for so long.
The amount of devices used by a single user will increase the attack surface. Attackers who have access to a single device will be able to steal private information and influence data synced across devices, as well as steal information that can be used to command IoT devices. Users, system administrators, and IoT device and application designers should think through the ecosystem of devices that users are likely to have, along with the threat agents, to architect possible solutions to mitigate these potential scenarios.
In 2007, Microsoft came under fire for a security hole in the speech-recognition component of their newly released Windows Vista operating system. A malicious website could simply play an audio file commanding the computer to delete files and empty the recycle bin, and the operating system would readily comply. An attacker could alternatively email the audio file to the victim and lure him or her to play it. Microsoft played down the issue, stating that the victim would have to have their microphone placed next to the speaker so the malicious set of audio commands are clearly interpreted by Vista. Furthermore, Microsoft stated that the user would likely recognize the attack because they would hear the audio instructions play. This is assuming the user is in the vicinity of their computer should someone exploit this issue.
Perhaps one of the reason this issue wasn’t taken very seriously by the users is that not many people leverage the speech function in desktop and laptop computers (except for individuals affected by impairments and related difficulties). When Vista was released in 2007, users primarily used the keyboard, mouse, and trackpad as their modes of input. With the growing popularity of intelligent. voice-operated personal assistant services like Siri and Cortana, this is changing. Users are starting to enjoy and find value in commanding their smart phones and computers with their voices.
Jumping in on the digital personal assistant bandwagon, Amazon released a product called Echo (Figure 8-1) which is primarily voice operated along with a companion smart phone App to configure it.
Figure 8-1. The Amazon Echo
The device is a 9 inch tall speaker and a microphone that has its default wake word set to “Alexa”. Just like Siri, you can command the Echo to tell you the weather by saying “Alexa, what is the weather like today?” or to ask trivia questions such as “Alexa, how tall was Michael Jackson?”.
The Echo can also be configured to turn on or off a selection of hue lightbulbs (discussed in Chapter 1). As with the hue iOS app, the Echo app can access the hue bridge once you press the button on the bridge to prove physical ownership. It is possible to specify a selection of hue lights into a specific group called “lights,” as shown in Figure 8-2.
Figure 8-2. 17 hue lightbulbs under the group called “lights”
var IDLE_TIMEOUT = 60; //in seconds
var _idleSecondsCounter = 0;
document.onclick = function()
_idleSecondsCounter = 0;
document.onmousemove = function()
_idleSecondsCounter = 0;
document.onkeypress = function()
_idleSecondsCounter = 0;
if (_idleSecondsCounter >= IDLE_TIMEOUT)
var audio = new Audio('alexa_lights_off.m4a');
_idleSecondsCounter = 0;
Back in 2007, the Windows Vista security issue was not of particular interest to the cybersecurity community since the impact and the probability of an attacker being able to pull it off was low (the voice-activation feature had to be turned on and the microphone needed to be next to the speaker). Today, however, more and more people are relying on audio-based personal assistants such as the Echo. What makes this attack vector of particular concern is that we are going to depend upon devices like the Echo to command IoT devices that could have a physical impact on our safety, such as lights going off.
The Amazon Echo also works with IFTTT recipes and can command the WeMo light switch. This makes the Echo a powerful device that is able to turn off appliances and lighting at homes. The Echo only allows the user to select “Alexa” or “Amazon” as the “wake word” which needs to be uttered as the first word in every command so it knows the user has intended it for the device. Our rudimentary proof of concept would have been thwarted if Amazon required users to select a unique wake word. Of course, theat agents such as neighborhood bullies would probably be able to find out from the local social network what the unique wake word is set to, but it would substantially limit threat agents who are completely remote and may not have access to that information.
Designers of products such as the Echo should correlate malicious activity that leverages audio as a channel of implementing attack vectors since the product is primarily designed to communicate using audio. The speech-recognition security hole may not have been deemed impactful in the past, but product designers and users need to be extremely cognizant of increased avenues of abuse using audio channels as we continue to increase our reliance on assistants such as the Echo.
IoT Cloud Infrastructure Attacks
Devices that offer Internet connectivity require supporting cloud infrastructure. We’ve seen how the Hue lighting system can be controlled from anywhere in the world using the iOS app. We’ve seen how the WeMo baby monitor can be accessed remotely through supporting infrastructure hosted by Amazon’s cloud service. We’ve seen how the Tesla Model S maintains a persistent cellular connection with Tesla’s infrastructure to obtain over-the-air updates, send diagnostics, and be controlled using the iOS app. Such reliance of IoT devices upon cloud infrastructure makes it a juicy target for abuse.
In late 2014, hackers compromised iCloud accounts of several celebrities and exposed their private nude photographs and videos to the public. They tried various combinations of passwords for the target iCloud accounts until they guessed the right ones. Since most iPhone users elect to sync their photographs and videos across devices using the iCloud service, the attackers were able to obtain the images upon logging in.
Although no actual vulnerability in the iCloud service was discovered to have been exploited, the reason attackers were easily able to obtain access was that certain functionality of the service did not implement controls to lock out accounts if too many unsuccessful attempts were made in a given period of time.
This celebrity breach demonstrates how the use of a static password makes it easy for potential attackers to gain access to private information. IoT devices, such as the Tesla Model S, also use static passwords that can be easily guessed to obtain access to cars, allowing attackers to track vehicles, unlock them, and even start them and drive away.
In addition to the cloud infrastructure implemented by the IoT device manufacturers themselves, we have platforms such as IFTTT and Apple’s HomeKit that will be included in the proposed attack surface. We’ve already seen how easy it is to connect our online spaces such as email and social networks with IoT devices such as lightbulbs and door locks. Compromising someone’s IFTTT account gives the attacker control over all of the virtual and physical services tied to the victim’s account.
Apple’s HomeKit service that is built into iOS is another example of a platform that will be of interest to attackers. The HomeKit service allows IoT device manufacturers to seamlessly work with Apple devices, even allowing the users to control their devices remotely. The goal of HomeKit is to allow users to easily set up new devices and then control them using Siri. Other big software companies like Google and Microsoft are also implementing frameworks like HomeKit to enable the emergence of consumer-based IoT devices. Apple has done a good job of setting clear guidelines that developers who use HomeKit must not leverage the data for advertising and data-mining. Frameworks and services such as this will become popular since they allow users to seamlessly interact and control their IoT devices. As such, cybersecurity researchers and malicious attackers (including disgruntled employees who have access to these systems) will also be drawn to potential vulnerabilities in such services that can be exploited to gain access to data available from various devices in the victim’s home.
In the recent past, breaches of cloud services have contributed to the loss of privacy for victims and financial gain for attackera. In the near future, attackers will look into exploiting cloud services to gain access to and abuse the functionality of IoT devices to further invade our privacy and potentially compromise our physical safety.
There have been various reports that the NSA may have intercepted devices such as network routers and planted backdoors in them. (A backdoor is a software or hardware modification of a device that allows the modifier to monitor and control the device remotely.) American government agencies have aggressively lobbied for popular hardware and software manufacturers such as Apple, Google, and Microsoft to build in mechanisms that would allow low enforcement agencies to monitor and obtain data from personal devices such as smart phones.
Besides United States based government agencies, the Chinese government is routinely accused of building backdoors into hardware and software produced in that country. Given that China is a major hub of hardware production, many electronic companies have major supply chain presence there. The Chinese government, in return, recently issued new regulations requiring foreign companies to reveal source code and build backdoors into software and hardware sold to Chinese banks.
The amount of power that can be exerted by a threat agent who is able to influence installing a backdoor into an IoT device is clear. The one thing to note is that competing attackers and threat agents can and will leverage backdoors once the knowledge of their existence is made public.
The Lurking Heartbleed
Heartbleed is a flaw in the OpenSSL library that can be exploited remotely to gain access to memory, including cryptographic keys and user credentials. OpenSSL is a popular library that is used by millions of devices to implement the Transport Layer Security (TLS) protocol to securely encrypt electronic communications.
Heartbleed was announced to developers on April 1, 2014, and at the time of disclosure, about 17% (i.e. around half a million) of Internet facing web servers were vulnerable to attack. Bruce Schneier, a well known security expert, described Heartbleed as a “catastrophic” issue given how easily it can be abused to steal information by a remote attacker.
In addition to workstations, IoT devices such as the Nest thermostat also use OpenSSL. Nest released an update for their thermostat product that was affected by the security issue and also advised their customers to change their Nest passwords just in case they may have been compromised due to the issue (Figure 8-3).
Figure 8-3. Nest support website detailing the Heartbleed security issue
OpenSSL is widely used in embedded devices beyond the Nest thermostat. Heartbleed demonstrates to us the potential catastrophic nature of a remotely exploitable vulnerability that can suddenly put millions of IoT devices at risk because they utilize common source code that has a bug in it. Another issue to keep in mind here is that IoT devices without the ability to update firmware and client software will remain vulnerable to critical issues such as this for their lifetime, thereby putting the privacy and safety of their consumers in danger.
Diluting the Medical Record
Vulnerabilities that exploit life-sustaining, hospital-grade devices have been proven. Researcher Jerome Radcliffe has detailed how he was able to use radio communication to remotely instruct a insulin pump to change the dosage being administered. This attack could be abused by a malicious entity within wireless range to kill a patient.
Consumer devices such as the Fitbit activity trackers are also gaining attention from the medical community. Doctors find it useful to obtain granular information about patients, such as their blood pressure, the amount of daily exercise they get, and other vitals that can influence prescribed dosages and treatments. There is consensus in the medical and technology communities that data from personal activity trackers will be incorporated into patients’ medical records, giving doctors greater visibility into the health of their patients in addition what they are able to measure in medical facilities. Figure 8-4 shows a screenshot of BPM (Beats Per Minute) data collected by the Apple Watch using the IOS HealthKit functionality. This information can be extremely useful to medical professionals to help diagnose a patient.
Figure 8-4. Heart-beat data collected by the Apple Watch
Government regulations and required health approvals will delay the convergence of data collected from personal devices into medical records, yet it is likely that it will eventually happen. Since devices such as the Fitbit and the Apple Watch are able to collect this information, it is valuable to begin to have a conversation on potential abuses of such information. One potential scenario for abuse in this case is the ability of a malicious entity to alter the stored information that is then relied upon by a medical professional. For example, tampering with an individual’s BPM statistics can result in a physician prescribing incorrect dosages for high or low blood pressure, which can potentially have significant negative impact to the patient’s health. In addition to monitoring activity, the iPhone lets users create an emergency Medical ID (Figure 8-5) that contains vital information such as known medical conditions, medications, blood type, and emergency contacts. This information is available even when the phone is locked so that medical professionals can access it in an emergency.
Figure 8-5. Emergency Medical ID Feature on the iPhone
Features such as the iPhone’s Medical ID will undoubtedly help save lives by giving doctors vital information in case of emergencies where the patient is unconscious or unable to communicate. This information can also put people’s lives at risk if the iPhone is compromised and the information is purposefully altered. The feature can also be abused by acquaintances who may have physical access to the iPhone and want to alter the information in unfortunate cases involving bad family dynamics and other psychological factors.
All information that is collected through the App will be sent to a secure data server run by Sage BioNetworks (“Sage”), a non-profit research organization. Sage will replace the direct identifiers listed above (your name, email address, and date of birth) with a code to help protect your identity—Sage will encrypt the direct identifiers and store them separately. Because the data are coded, researchers using the data will not be able readily to identify which information pertains to you. Stanford researchers will, however, maintain your consent and personal information and retain the ability to re-identify the information if doing so is needed for research integrity purposes or legal purposes, and they may share re-identified information with others at Stanford who need to see such information to ensure that the research meets legal, regulatory or institutional requirements.
In this case, information collected is sent to a remote data server. It is then replaced with a random identifier token so researchers using the data will not be able to identify the individual the data is collected from. However, another database is maintained that can be leveraged to re-identify the individual should Stanford researcher decide or need to do so. This is a solid example of how health data collected from sensors attached to our bodies is going to be leveraged and possibly distributed across cloud platforms around the world. The security of these platforms, as well as what access the researchers themselves have in terms of identification purposes, will define and possibly contribute to privacy violations that, until now, have been limited to traditional medical records stored by hospitals and dotor offices.
The Data Tsunami
Most people who use Facebook or Google have noticed targeted ads with a high creep factor, i.e. ads on these platforms are tailored to precicely suit people’s previous search queries, email contents, instant messages, and social network dynamics.
Services such as Google now go through your data to suggest events and even offer to check you in for your flights based on an email copy of your itinerary. The world of IoT will bring in additional sensor and behavior-based data that will be valuable to social network companies and extremely useful for marketing. We are likely to see advertisements targeted to us for blood-pressure medications based on our BPM readings from our smart watches or even anti-insomnia drugs based on data collected about what time we usually turn off our IoT-based lighting system.
As people start to consume IoT devices that they want to integrate and automate using platforms and frameworks provided by companies such as Google and Apple, information collected from various sensors will be available and used for marketing and stored across multiple cloud architectures. This data will be a gold-mine for malicious agents who have previously been limited to online platforms such as email and social networking. Besides privacy, the ability of a threat agent to tamper this information may have health or physical safety implications in cases where the altered information is consumed by IoT devices. It is likely that such violations of security and privacy will enrage customers, who will demand ability to granularly track what data is being collected, how it’s being used, and the ability to opt out.
Targeting Smart Cities
Researcher Billy Rios hacked into the building-control system of an Australian Google office. The building was found to use the Tridium Niagara AX platform that allows administrators to remotely control physical security alarms, physical access, water temperature, as well as heating and air conditioning. They were able to obtain access by using the default administrator password of “anyonesguess”. This password was stored in a configuration file that the researcher obtained due to a vulnerability in the system that exposed this information to unauthenticated users. Tridium systems are popular around the world, and according to Rios, he was able to use the Shodan tool to locate more than 25,000 such systems exposed to the Internet.
Besides industrial grade connected systems like those exploited by Rios, we are starting to see consumer-grade IoT devices such as the ones explored in this book thus far. The concept of the Smart City (also discussed in Chaoter 7) combines the use of industrial grade and consumer grade IoT devices to effectively manage energy, healthcare, transport, and waste across a geographical location. Smart cities will use a combination of industrial-grade connected devices (such as smart parking meters and traffic lights) along with consumer-grade IoT devices installed in homes and directly configurable by citizens (such as lighting, door locks, cars, that have been the focus of this book). Researcher Cesar Currodo’s paper An Emerging US (and World) Threat: Cities Wide Open to Cyber Attacks covers attack vectors against industrial-grade connected devices that will support the upcoming emergence of the smart city. Currodo’s research and the devices presented in this book will set the stage for attack vectors encompassing smart cities based on various categories of interconnected devices and services.
Efforts by society to construct smart cities are likely to include a curated selection of interconnected devices to provide for consistency and scalability. This brings the drawbacks of “monoculture” into the discussion. Much like in living species, the advantage of monoculture includes low variability in genetics, which in turn requires lower resources to find medical cures for diseases since experimentation wouldn’t require the analysis of a variety of specimens. The disadvantage of monoculture is that a new variant of a disease can wipe out the entire population because it will equally infect every individual.
This logic can also be applied to computer systems and IoT devices: increased monoculture will lead to lower costs and ease of interoperability, while increased variability will lead to lesser chance of a malicious attack being able to compromise an entire smarty city. There are areas of technology, such as the TCP/IP protocol, that are so fundamental to electronic communication that it may make offer no advantage to create additional protocols to decrease the monoculture aspect. The influence of monoculture in systems is an ongoing topic of discussion and debate in the cyber-security community. It has definite applicability to the concept and emergence of the smarty city, and it is ultimately dependent upon the total cost of implementation versus perceived risks that may be introduced.
Inter Space Communication Will be a Ripe Target
One of humankind’s greatest endeavors is our ongoing quest to colonize Mars. NASA plans to send humans to Mars by the 2030s. Their engineers and scientists are working hard to develop the technologies astronauts will use to one day live and work on Mars, and safely return home from the next giant leap for humanity.
Establishing communications between Earth and Mars will be critical in making sure that space agencies are able to successfully transmit crucial data related to the mission and that humans are able to communicate with each other. NASA is aware of the importance of securing communications during space missions:
“Reliable communication between ground and spacecraft is central to mission success, especially in the realms of digital communication (data and command links). Seen in the light of recent events, these communication links are vulnerable to malicious intrusion. If terrorists or hackers illegally listen to, or worse, modify communication content, disaster can occur. The consequences of a nuclear powered spacecraft under control of a hacker or terrorist could be devastating. Therefore, all communications to and between spacecraft must be extremely secure and reliable.”
There are various projects that are underway to facilitate space communication more efficiently. For example, NASA’s Optical Communications project is researching ways to use light to transmit data.
Space communication protocols need be robust enough to withstand delays, disruptions and disconnections in space. Glitches can happen when a spacecraft moves behind a planet, or when solar storms and long communication delays occur. It takes from 4 to 20 minutes to transmit data between Mars and Earth, so NASA’s systems have to tolerate such delays. Therefore, instead of using TCP/IP, NASA has developed a protocol called Disruption-Tolerant Networking (DTN) that is able to work seamlessly during delays and loss of connectivity. DTN is designed to incorporate cryptography and key management, signifying that space agencies are taking steps to make sure security is built in to the design of space communication protocols.
Elon Musk’s SpaceX is planning on launching 4,000 low-orbit satellites to provide global Internet access on planet earth. SpaceX plans to extend this network of satellites to include communications with satellites on Mars when its mission to send human to Mars comes to fruition.
It is easy to imagine how important communication is going to be to enable critical and risky space missions. Rockets and satellites (and other objects relevant to the mission) are “things” that are going to be available and accessible on space communication infrastructure. As NASA and SpaceX move forward with deploying a greater number of satellites to facilitate networks in space, their architecture will be a ripe target for a many threat agents. Terrorists and competing nation states are likely to attempt to exploit vulnerabilities that may be present in network protocols to steal intellectual property and to disrupt space missions. Such a security breach could result in the loss of human lives or even the failure of humankind to populate other planets. This is going to be an important area for security researchers to contribute to in order to make sure we are building our space communication infrastructure securely from the ground up.
The Dangers of Superintelligence
Irving John Good, a British mathematician who worked as a cryptologist at Bletchley Park with Alan Turing, is often quoted when discussing the perils of machines achieving greater intelligence than humans:
“Let an ultraintelligent machine be defined as a machine that can far surpass all the intellectual activities of any man however clever. Since the design of machines is one of these intellectual activities, an ultraintelligent machine could design even better machines; there would then unquestionably be an intelligence explosion, and the intelligence of man would be left far behind. Thus the first ultraintelligent machine is the last invention that man need ever make”.
Nick Bostrom, author and professor at Oxford, defines superintelligence as “an intellect that is much smarter than the best human brains in practically every field, including scientific creativity, general wisdom and social skills”. Bostrom and other leading scientists are worried that machines capable of superintelligence are going to be difficult to control and that they may have the ability to take over the world and eliminate humankind.
Well known intellectuals and leaders such as Bill Gates are worried about super intelligence:
“I am in the camp that is concerned about super intelligence. First the machines will do a lot of jobs for us and not be super intelligent. That should be positive if we manage it well. A few decades after that though the intelligence is strong enough to be a concern. I agree with Elon Musk and some others on this and don’t understand why some people are not concerned”.
The Future of Life institute is a volunteer-run research and outreach organization that has been set up to measure and mitigate existential risks facing humanity, including superintelligence. The institute is currently focusing on potential risks from the development of human-level artificial intelligence. The founders and advisory board of the Future of Life institute consists of indviduals such as Stephen Hawking, Alan Alda, and Elon Musk, to name a few. This demonstrates that some of the leading minds across various industries are genuinely worried of the perils of supertintelligence and that they want to contribute their time and effort to make sure we think through risks appropriately.
As we get closer to designing machines that are capable of superintelligence, it is likely that professionals with cybersecurity experience are going to be called upon to assist in designing algorithms that can help curtail potentially intelligent machine by running simulations of artificial intelligence in a controlled environment (i.e. a “sandbox”) that protects the safety of human beings.
Computing devices capable of greater levels of intelligence will have access to IoT devices that they can control. The intelligent code itself will be a “thing” on the network it is executed on and may have the intelligence to branch out onto other networks. The unique situation with the risk of superintelligence is that catastrophe is likely to be caused in the lab where scientists are not able to control a superintelligent machine in ways they originally thought they could. It is easy to see how the knowledge of how to create and unleash superintelligent computing can be attractive to terrorists who want to unleash destruction and havoc.
The threat of superintelligence is easily on top of mind of scientists and researchers in the area of computer science, and is quite likely the greatest human-made existential threat to humankind.
From thingbots to drones to device backdoors, the attack surface presented by interconnected devices is going to be attractive to threat agents and provide them with unparalleled access to our privacy and physical safety.
Vulnerabilities such as Heartbleed that suddenly affect millions of computing devices have already known to affect IoT devices such as the Nest thermostat. It is quite likely that many other IoT devices in the market are vulnerable to Heartbleed and will continue to be vulnerable due to lack of a security patch, either because the vendor has not released one or because the patch was not applied or failed. There are also devices that do not incorporate a mechanism for patches to be applied at all, and these devices are going to contribute to insecurity for years to come until they are decommissioned.
Information collected about citizens in terms of advertisement tracking as well as the compromise of health data has been an ongoing issue of concern. With IoT devices in the mix, we are bound to have cases where data collected from IoT sensors within our homes will be leveraged for marketing, contributing to intrusive privacy violations. Health information, traditionally trapped within instruments in hospitals, will be collected by consumer devices that may be prone to tampering. This health information will be utilized for research and stored in multiple locations in the cloud, thus increasing the probability of exposure.
Areas around the world are starting to leverage the concept of the smart city to serve their citizens efficiently. The combination of industrial- and consumer-grade IoT devices that empower these cities will introduce risk and open citizens up to privacy and security attacks that were previously limited to online spaces.
In terms of travelling to Mars and making our machines more intelligent, humankind faces great potential for triumph as well as great peril from threat agents—including superintelligent machines—who could destroy our population and curtail our ability to colonize other planets. As we make progress in the field of space travel and machine intelligence, we are going have to discuss and put a lot of thought into how to secure these platforms early on because vulnerabilities in these efforts can lead to the loss of human lives.
The ultimate goal of the cybersecurity profession is to enable connected technology as securely and swiftly as possible, and this enablement must begin with informed predictions of upcoming scenarios of attacks such as the ones discussed in this chapter.