Building a Zero Trust Security Model for Autonomous Systems

This sponsored article is delivered to you by Technology Innovation Institute.

Autonomous programs sit on the intersection of AI, IoT, cloud architectures, and agile software program improvement practices. Various streams of those programs have gotten outstanding, corresponding to unmanned drones, self-driving automobiles, automated warehouses, and managing capabilities in good cities. The drone business alone was estimated at US $100 billion in 2020, and autonomous programs are already driving considerably extra worth throughout different domains.
¹

But surprisingly little consideration has been paid to securing autonomous programs as programs composed of a number of automated parts. Various patchwork efforts have centered on particular person parts. In tandem, cloud providers are beginning to undertake a Zero Trust strategy for securing the chain of belief which may traverse a number of programs.

With that, it has turn into crucial to increase a Zero Trust structure to programs of autonomous programs to guard not solely drones but in addition industrial gear, provide chain automation, and good cities.

---

In the close to future, autonomous programs will convey a brand new degree of digital transformation to the business price trillions of {dollars}, together with automating transportation, visitors administration, municipal providers, legislation enforcement, delivery, port administration, building, agriculture, and extra.

Autonomous enterprise programs are additional enriching these extra bodily elements of autonomous programs. Gartner coined the time period
hyperautomation to explain instruments for scaling automation utilizing software program robots that have been valued at $534 billion in 2021.² Despite the significance of autonomous programs, surprisingly little analysis has centered on securing autonomous programs as a set of programs.

This is to not say that researchers are ignoring safety — in any case, safety infrastructure and instruments are a multi-billion greenback business. But in terms of securing bodily parts, a lot of the main target has been on securing particular person components corresponding to information, software program, {hardware}, and communications hyperlinks slightly than the conduct of an ensemble of autonomous programs.

Despite the significance of autonomous programs, surprisingly little analysis has centered on securing autonomous programs as a set of programs.

Similarly, researchers are simply beginning to scratch the floor of defending in opposition to swarms of autonomous issues guided with malicious intent. Just final yr, a half dozen exactly focused malicious drones managed to gradual oil manufacturing in Saudi Arabia for days, and extra just lately, a number of low-cost drones prompted important harm to grease tankers within the UAE. This illustrates the significance of detecting hostile drones getting into safe areas.

This type of safety is just the start of what is going to be required to maneuver in the direction of a bigger scale deployment of drones as envisioned by the U.S Federal Aviation Administration’s past visible line of sight (BVLOS) rules.
³ These rules promise to open immense industrial alternatives to enhance industrial inspection, delivery, and distant monitoring.

However, wider scale deployment would require a extra systemic strategy to guard in opposition to the impression of hundreds of low-cost autonomous drones working in live performance.

A extra complete strategy is required to guard the safety and resilience of autonomous programs and shield in opposition to cyber-physical assaults that leverage autonomous programs.TII

Autonomous Security for Smart Cities

Autonomous safety can also be a urgent difficulty for industrial programs and good metropolis use circumstances. Hackers have gotten higher at coordinating thousands and thousands of IoT gadgets to launch a devastating distributed denial of service assaults on laptop servers at the moment.

Similar ways that leveraged bodily and cell autonomous issues might lengthen the blast radius past IT infrastructure to destroy bodily infrastructure like factories, pipelines, electrical grids, or worse. A extra complete strategy is required to guard the safety and resilience of autonomous programs and shield in opposition to cyber-physical assaults that leverage autonomous programs.

The
Technology Innovation Institute (TII)’s Secure Systems Research Centre (SSRC) is main one promising strategy to constructing an autonomous safety testbed that explores the interaction between how {hardware}, software program, and communications programs might be exploited in order that they are often hardened.

The early phases of this work are centered on defending scalable swarms of unmanned aerial autos managed by the cloud. The long-term purpose is to create a framework for understanding and defending in opposition to autonomous safety dangers throughout all sorts of infrastructure, together with fleets of automobiles, automated warehouses, building websites, farms, and good cities.

Over the years, fundamental autonomous capabilities have grown into nearly each side of our bodily infrastructure, from automated braking in particular person automobiles to orchestrating energy stream throughout nationwide electrical grids with precision. Autonomous programs are already demonstrating great worth at the moment, and we’re simply scratching the floor. For instance, Goldman Sachs estimated that unmanned autonomous autos (UAV) had grown right into a $100 billion business in 2021.
⁴

Military functions accounted for about 70 % of this spending. However, industrial functions have been additionally substantial in building, agriculture, insurance coverage claims, offshore oil, gasoline and refining, pipelines, utilities, and mining. For instance, the development business makes use of drones to mechanically seize footage of building websites earlier than, throughout, and after the development course of.

Drones carrying lidar and high-resolution cameras can mechanically generate 3D fashions in minutes that will have beforehand taken people days or even weeks. This makes it sensible to maintain tabs on buildings as they’re being constructed, monitor progress, and determine errors when they’re cheaper to repair. After building, drones can even survey bodily infrastructure like bridges to determine cracks and different issues earlier than the entire construction suffers an even bigger downside.

Drones are additionally enhancing the planning and administration of huge farms. For instance, drones with spectral imaging cameras can rapidly determine nutrient deficiencies, pest outbreaks, and drought, permitting farmers to handle them extra exactly and cheaply. In the UAE, drones have additionally helped map your complete nation’s agricultural assets in a matter of days, which might not have been sensible utilizing bodily surveys alone.
⁵ In one other effort, UAE groups used drones to plant 6.25 million bushes in solely two days.
⁶

Scaling Secure Autonomous Systems

It is simple to get caught up in autonomous programs as a single self-driving automobile or particular person drone. However, the actual promise of autonomous programs comes when autonomous capabilities are concurrently scaled to enhance the management of particular person issues, the orchestration of a set of issues, and the understanding of issues at scale.

The individual-level might be thought of because the evolution from cruise management to automated braking and absolutely self-driving automobiles. The orchestration degree entails the evolution from synchronized visitors lights to dynamically adjusted visitors lights to superior mapping providers that route automobiles round visitors jams. Autonomous understanding programs embody visitors monitoring cameras to crowdsourcing dashcam video into dynamically up to date digital twins for enhancing total visitors.
⁷

These identical three components of management, orchestration, and understanding play out throughout varied use circumstances. A warehouse robotic would possibly scale back the necessity for workers. An autonomous warehouse administration system might optimize the scheduling and staging of things within the warehouse. In distinction, an autonomous understanding system might assist reengineer the warehouse design to additional enhance efficiency in the identical house.

This mixture of autonomous management, autonomous orchestration, and autonomous understanding is already displaying some promise within the UAE. For instance, one pilot venture has created an autonomous port truck system that automates the method of shifting delivery containers from boats to vehicles.
⁸

Gartner refers back to the simultaneous evolution of management, orchestration, and understanding in IT programs as
hyperautomation. In this case, enterprises use particular person robotic course of automation (RPA) software program robots (known as bots) to automate a set of human duties. Orchestration engines assist set up the stream of labor throughout a number of bots.

Then course of and process mining bots analyze enterprise functions and even watch over the shoulders of people to seek out additional alternatives for enchancment. Researchers are simply beginning to discover how comparable practices could also be prolonged to incorporate autonomous autos.

The Next Challenge in Systems Security

That is likely one of the causes ATRC’s ASPIRE selected to deal with autonomous swarm coordination as a part of its
subsequent grand problem venture.⁹ ASPIRE is tasked with internet hosting grand problem competitions loosely organized just like the US DARPA’s problem that spearheaded analysis on autonomous autos.

The upcoming problem duties researchers with discovering one of the simplest ways to orchestrate a swarm for drones to seek for and retrieve objects hidden on ships which are too heavy for any particular person drone. The want for end- to-end safety and resilience Enterprises and safety researchers are simply beginning to wrestle with defending particular person autonomous issues, a lot much less swarms. A brand new safety strategy is required for most of these swarms to scale for real- world functions.

The early technology of IoT gadgets have been rushed to market with solely fundamental issues on how they may be protected in opposition to hackers or securely up to date in opposition to new threats. Many of those early gadgets aren’t updateable after the actual fact. Consequently, they’re a well-liked goal for hackers desirous to create large-scale botnets for launching distributed denial of service assaults such because the Mirai botnet.
¹⁰

This has given rise to a secondary business of IoT safety gateways designed to detect and block malicious exercise exterior of poorly secured home equipment like lighting controllers, crockpots, TV set-top containers, and cameras. The safety posture of the primary linked automobiles is healthier, however there are nonetheless obvious vulnerabilities and gaps that should be addressed. Some of the vulnerabilities highlighted in Upstream’s 2021 Automotive Cyber Security Report
¹¹ embody:

• Hackers discovered 19 vulnerabilities in a Mercedes-Benz E-class automobile that allowed them to remotely management the car, open doorways, and begin the engine.
• Hackers took management of a automobile’s OEM company community by reverse engineering a automobile’s transmission management unit to infiltrate the community.
• Over 300 vulnerabilities have been found in 40 well-liked digital management models utilized in automobiles.
• Hackers managed to achieve management over Tesla’s total linked automobile fleet by exploiting a vulnerability within the communications protocol.

Modern automobiles enable the software program to be up to date after the actual fact however typically require shoppers to return to a store for an replace. Only just a few leaders, like Tesla, have mastered the power to securely replace software program at scale.

Building safe programs might want to handle {hardware}, software program, and protocols and their interaction. Hardware safety points want to guard in opposition to assaults by which a hacker can bodily replace a system to compromise safety or trigger harm.

For instance, the Stuxnet
¹² assault corrupted {hardware} in an Iranian uranium enrichment facility to ship miscalibrated timing information that confused the management programs. The end result was that the controller drove lots of of pricey centrifuge programs so quick that they exploded.

There are quite a lot of methods hackers might launch distant hardware-directed assaults on UAVs. For instance, centered beams of sound might confuse the inertial steering unit used to manage a drone. Directed EMF beams would possibly trigger a brief circuit on delicate electronics, and lasers or vibrant lights would possibly confuse or destroy digital camera sensors.

Vulnerabilities in software program programs enable hackers to spy on or take distant management of programs to launch additional assaults. Early examples in IT programs included malware just like the Zeus Trojan that allowed hackers to spy on banking interactions to seize credentials and steal $500 million.
¹³

In some circumstances, hackers are discovering methods to infiltrate software program provide chains to plant focused malware vulnerabilities. This was how hackers managed to burrow into hundreds of presidency, banking, and enterprise programs as a part of final yr’s Solar Winds breach.

The basic idea on the coronary heart of the Zero Trust safety mannequin is to by no means belief and all the time confirm the provenance of every request.TII

The Zero Trust Security Paradigm

The time period Zero Trust mannequin was coined by Forrester analysis in 2010 to indicate a brand new paradigm for securing distributed programs.
¹⁴

Security programs have historically been secured by hardening a bodily perimeter. But on the planet of cloud computing, the perimeter is extra nebulous. Zero belief safety connotes the thought of all the time authenticating and verifying each entry with a view to safe round a extra versatile perimeter. The Zero Trust paradigm permits safety groups to plan for the likelihood that vulnerabilities might exist all through a series of interactions amongst a number of programs, corresponding to throughout a number of cloud providers, information processes, storage providers, and networks.

The basic idea is to by no means belief and all the time confirm the provenance of every request. Another fundamental precept is to imagine {that a} breach has already occurred, making it important to restrict the blast radius of any breach.

Autonomous programs lengthen automated processes throughout a greater diversity and bodily vary of {hardware}, communications protocols, in addition to management and orchestration mechanisms. Each of those brings with them their very own assault floor. Thus, safety groups want to reduce the impression {that a} breach on one degree might have on different programs.

Zero belief safety connotes the thought of all the time authenticating and verifying each entry with a view to safe round a extra versatile perimeter.

Examples embody assaults on management servers, communication networks, embedded system functions, bodily gadgets, software program provide chains, and silicon provide chains.

Autonomous system safety must be constructed throughout a number of impartial safety partitions in order that if one key or system is breached, the integrity of the entire is protected. Each system needs to be designed to fail safely and securely in order to reduce the impression on adjoining parts.

This can even make it more durable for hackers to escalate an assault on a low-level system to extra important programs, as with the current Log4J assaults. For instance, autonomous programs like autonomous drones want attestation schemes to make sure that solely licensed software program runs on the drones. An attestation scheme makes use of cryptographically signed software program updates to make sure that solely legitimate code can run on distant programs.

This prevents hackers from reprogramming a drone by simulating a respectable program replace communication or changing respectable updates with a bogus software program improve staged on the command heart.

Vulnerabilities in communication protocols might enable hackers to spy on drone exercise or simulate management alerts to take management of a drone. Such assaults might occur at any degree of the communication stack, from hacking into communications throughout the cloud, the wi-fi alerts between the cloud and a drone, or between a number of drones.

In some circumstances, hackers could possibly assault programs by mimicking communications inside a drone or autonomous automobile. For instance, researchers have discovered methods to hearken to and simulate the unprotected wi-fi communications concerned in tire stress monitoring.
¹⁵ This allowed them to trick the automobile into indicating {that a} good tire had a flat, which could trigger a car to cease.

Autonomous Systems Infrastructure

Autonomous programs infrastructure wants to mix quite a lot of embedded computing platforms corresponding to drone navigation programs, CPU-based architectures, and different sorts of devoted {hardware}. Existing approaches are additionally mounted at design time, which results in utilizing untrusted software program to make use of peripherals in TEEs.

New approaches for composite enclaves can be required to increase TEEs to extra versatile designs. It’s additionally important to develop new instruments for detecting and responding to unknown and sudden adjustments attributable to novel assault strategies. A belief verification infrastructure might lengthen conventional API observability approaches to {hardware} via a mix of monitoring, logging, and tracing. These sorts of actions enable the development of steady verification mechanisms for anomaly and intrusion detection.

Offline profiling strategies might generate belief profiles that describe how the {hardware} is meant to function. During operation, ongoing logging might make sure that the conduct adheres to the belief profile.

However, this must be constructed to reduce the danger of logging instruments being leveraged as a part of a side-channel assault. We additionally have to discover new {hardware} capabilities and superior software program strategies to compartmentalize software program stacks throughout a number of ranges.

One of essentially the most promising approaches is CHERI (Capability Hardware Enhanced RISC Instructions), which is being explored by DARPA, Google, SRI International, and the University of Cambridge.
¹⁸ Further work is required to increase this strategy to enhance fine-grained compartmentalization on the working system degree. This might mix new middleware, OS libraries, unikernels, and varied mechanisms to grant and revoke authorization with a view to implement compartmental constraints.

This can be required placing the fitting steadiness between completely different levels of flexibility in each configuration and figuring out the suitable privileges. We are nonetheless within the early days of constructing large-scale autonomous programs, however as we scale them up, new issues like these can be required to increase zero-trust safety to embedded programs, autonomous programs, and programs of autonomous programs.

Enterprises and researchers are exploring methods to scale programs of particular person autonomous programs, with essentially the most promising analysis at the moment being centered on scaling programs of autonomous drones. In the long term, everybody needs to get to autonomous automobiles and factories and there’s a lot of experimentation occurring with unmanned autos that are likely to require a human driver or assistant within the case of supply autos.

But UAVs are already delivering worth at the moment, and regulators are beginning to open the skies for extra bold functions. In early 2021, the FAA granted American Robotics the primary license to fly drones past the visible line of sight (BVLOS). Around the world, enterprises are working with regulators to develop Unmanned Traffic Management (UTM) programs. Major aerospace corporations and revolutionary start-ups are working with regulators to point out how varied mixtures of AI, superior mapping, vehicle-to- car communications, and encrypted communication and management might facilitate protected drone administration at scale.

In the United States, Boeing has partnered with SparkCognition on SkyGrid. Airbus is main efforts to advertise SESAR for the EU. Guardian Angel just lately labored with U.Ok. regulators on Operation Zenith to show how a collection of on-airfield duties might be carried out with out endangering or disrupting airport operations. These are important efforts and are a mandatory first step in safely scaling fleets of trusted drones.

However, UTM programs typically begin with the belief that drones are all trusted. More work must be executed to know and analyze how these programs might be compromised and therefore trusted within the first place.

Researchers world wide are exploring how particular person parts of those programs might be compromised and hardened. For instance, researchers in Germany and Switzerland have experimented with implementing quantum-safe cryptographic algorithms to guard drone communications. They argue that long-running drones will even have to assist crypto-agility that enables dynamic updating of safety algorithms in response to the invention of latest vulnerabilities.
¹⁹

These researchers additionally explored easy methods to implement distant attestation schemes that shield drones from software program tampering. Other researchers have explored drone cloud management mechanisms. For instance, a group of researchers in Brazil has developed the Cloud-SPHERE platform as one strategy for integrating UAVs into IoT and Cloud Computing paradigms.

Architectures that try to manage every drone or autonomous system instantly will run into scalability challenges because the variety of people within the swarm grows.TII

Bringing Security to the Swarm

The subsequent part of autonomous drones would require creating structure to scale drone management and safety to assist autonomous swarms. For instance, a set of low-cost drones might be orchestrated into drone swarms managed by the cloud to discover new use circumstances like search and rescue, disinfecting public areas, and coordinating duties corresponding to lifting heavy gear past the capability of anybody drone.

One huge shift would be the want for extra distributed management mechanisms. Architectures that try to manage every drone or autonomous system instantly will run into scalability challenges because the variety of people within the swarm grows.

One strategy pursued by the TII’s Secure Systems Research Centre (SSRC) is the event of a dynamic hierarchy composed of drones with completely different capabilities for management and process execution. Similar group of drones has been described earlier than, and our focus goes to be on safety and resilience in such a hierarchy.

In this scheme, a tier of Fog Drones acts as intermediaries between much less refined Edge Drones and the cloud. The Fog Drone can even offload many duties corresponding to summarizing enter from many Edge Drones to cut back the quantity of communication required with the cloud and between drones.

This can even scale back the quantity of processing required on every Edge Drone. This work can also be exploring how mesh networks can additional optimize and safe communications between drones working in constrained conditions corresponding to a cave, fallen constructing, or hostile setting.

Cross-disciplinary analysis groups at SSRC are exploring methods to synthesize classes discovered from bodily testbeds into helpful and actionable safety fashions. Ultimately, these safety fashions might assist autonomous groups determine and enhance autonomous programs improvement that spans drone {hardware}, software program implementations, and communications decisions earlier within the launch cycle.

Several testbeds have additionally been developed at Masdar within the UAE and Purdue. One purpose is to develop machine studying strategies at each the drone and cloud ranges to detect safety points and allow resilience.

Another purpose is to develop instruments for testing these programs in augmented actuality environments for city settings. The groups are additionally exploring methods to enhance the power to seize security-related information into digital twins that mirror the safety implications of drones. This will assist automate the power to mirror new safety vulnerabilities found in the actual world within the fashions shared with researchers.

These researchers are additionally discovering methods to harden open-source {hardware}, software program, and communication protocols for creating and deploying drone programs. This strategy opens the structure to a variety of safety and drone researchers to seek out vulnerabilities sooner.

This open-source strategy might additionally profit from the speedy innovation that the open-source robotics group is already seeing. Some of the underpinnings of the present platform embody the PX4 superior autopilot, NuttX real-time working programs, and the Robot Operating System 2 (ROS2).

The group has additionally developed and applied an open-source RISC-V processor and system on chip with specialised security measures baked in.

The Future of Autonomous Systems Security

Today, nearly all drone functions contain the administration of particular person drones. The subsequent evolution of drone adoption would require discovering methods to scale each the command-and-control infrastructure , in addition to hardening the safety and resilience of those programs.

Ultimately, analysis round securing autonomous programs, and never simply particular person drones, will assist facilitate widespread industrial deployment. It is important for designers of autonomous programs to undertake parts which have been hardened and might be up to date recurrently as new issues are found.

Many enterprises are adopting DevSecOps practices by which safety issues are undertaken as a part of the software program improvement and deployment. In these circumstances, varied instruments are used to vet code updates in opposition to identified finest practices and reject updates that fail fundamental safety assessments.

Improvements in UAV architectures is also used to enhance the resilience of enterprise functions, autonomous warehouses, and good cities.

Afterwards, software program scanning instruments, corresponding to WhiteHat and Contrast OSS, construct a listing of libraries utilized by the apps, sending an alert when important vulnerabilities are detected inside lively programs. Similar approaches might want to lengthen to enhance the parts utilized in creating and deploying autonomous programs that scan not solely the software program, but in addition the {hardware} and communications protocols used.

The first outcomes of those sorts of collaborative drone safety applications are just the start. Eventually, enhancements in UAV architectures is also used to enhance the resilience of enterprise functions, autonomous warehouses, and good cities.

In addition, higher instruments for modelling drone safety points will even inform the event of methods to guard in opposition to largescale assaults by swarms of compromised drones. The FAA means that the evolution of UTM programs, which give safety for UAVs, different infrastructure, and other people, ought to comply with a spiral strategy, beginning with low complexity operations and steadily constructing modules to assist larger complexity operational ideas and necessities.

Similarly, the evolution of instruments for enhancing autonomous programs safety would require a spiral strategy as autonomous programs evolve.

References

¹ Goldman Sachs. “Drones: Reporting for Work.” Accessed March 16, 2022. https://www.goldmansachs.com/insights/technology-driving-innovation/drones/

² Gartner. “Gartner Forecasts Worldwide Hyperautomation-Enabling Software Market to Reach Nearly $600 Billion by 2022.” Accessed March 16, 2022. https://www.gartner.com/en/newsroom/press-releases/2021-04-28-gartner-forecasts-worldwide-hyperautomation-enabling-software-market-to-reach-nearly-600-billion-by-2022

³ “Advisory and Rulemaking Committees – Unmanned Aircraft Systems (UAS) Beyond Visual Line-of- Sight (BVLOS) Operations Aviation Rulemaking Committee (ARC).” Template. Accessed March 16, 2022. https://www.faa.gov/regulations_policies/rulemaking/committees/documents/index.cfm/committee/browse/committeeID/837

⁴ Goldman Sachs. “Drones: Reporting for Work.” Accessed March 16, 2022. https://www.goldmansachs.com/insights/technology-driving-innovation/drones/

⁵ Ford, Georgina. “Counting Camels in The Desert – A Drone-Powered Success Story.” Commercial Drone Professional (weblog), September 30, 2021.
https://www.commercialdroneprofessional.com/counting-camels-in-the-desert-a-drone-powered-success-story/

⁶ Douglas, Alex. “UAE to Emerge as World Leader in Using Drones, Predicts Falcon Eye.” Commercial Drone Professional (weblog), April 1, 2020. https://www.commercialdroneprofessional.com/uae-to-emerge-as-world-leader-in-using-drones-predicts-falcon-eye/

⁷ EnterpriseBeat. “Nexar and Las Vegas Tackle Traffic with Digital Twins,” September 27, 2021. https://venturebeat.com/business/nexar-and-las-vegas-tackle-traffic-with-digital-twins/

⁸ “Region’s First Autonomous Port Truck System to Be Implemented GulfToday.” Accessed March 16, 2022. https://www.gulftoday.ae/business/2021/07/06/regions-first-autonomous-port–truck-system-to-be-implemented

⁹ Defaiya, Al. “Al Defaiya | Abu Dhabi’s ASPIRE Launches Over US$3 Million MBZIRC Maritime Grand Challenge,” October 22, 2021. https://www.defaiya.com/news/Regional%20News/UAE/2021/10/22/abu-dhabi-s-aspire-launches-over-us-3-million-mbzirc-maritime-grand-challenge

¹⁰ “The Mirai Botnet Explained: How IoT Devices Almost Brought down the Internet | CSO Online.” Accessed March 16, 2022. https://www.csoonline.com/article/3258748/the-mirai-botnet-explained-how-teen-scammers-and-cctv-cameras-almost-brought-down-the-internet.html

¹¹ Upstream Security. “2021 Automotive Cybersecurity Report | Press Release | Upstream.” Accessed March 16, 2022. https://upstream.auto/press-releases/2021-report/

¹² Kushner, David (26 February 2013). “The Real Story of Stuxnet”. IEEE Spectrum. 50 (3): 48–53. doi:10.1109/MSPEC.2013.6471059. S2CID 29782870. https://spectrum.ieee.org/the-real-story-of-stuxnet

¹³ “$500 Million Botnet Citadel Attacked by Microsoft and the FBI | The Independent | The Independent.” Accessed March 16, 2022. https://www.independent.co.uk/tech/500-million-botnet-citadel-attacked-by-microsoft-and-the-fbi-8647594.html

¹⁴ September 17, Kelly Jackson Higgins Editor-in-Chief and 2010. “Forrester Pushes ‘Zero Trust’ Model For Security.” Dark Reading, September 17, 2010. https://www.darkreading.com/perimeter/forrester-pushes-zero-trust-model-for-security

¹⁵ BAE Systems | Cyber Security & Intelligence. “Security Challenges for Connected and Autonomous Vehicles.” Accessed March 16, 2022.
https://www.baesystems.com/en/cybersecurity/feature/security-challenges-for-connected-and-autonomous-vehicles

¹⁶ Yasin, Muhammad, and Ozgur Sinanoglu. “Evolution of logic locking.” In 2017 IFIP/IEEE International Conference on Very Large-Scale Integration (VLSI-SoC), pp. 1-6. IEEE, 2017.

¹⁷ Rashid, Fahmida Y. “The rise of confidential computing: Big tech companies are adopting a new security model to protect data while it’s in use-[news].” IEEE Spectrum 57, no. 6 (2020): 8-9. https://spectrum.ieee.org/what-is-confidential-computing