As Wireless Emergency Alerts (WEAs) and other new technologies have been deployed, research investments, in large part supported by the Department of Homeland Security (DHS), have provided some insight into questions around the use of new technologies for alerting (Appendix B includes summaries of this work). However, to reach the above-envisioned alert and warning system, additional research questions will need to be answered about the use and design of WEA as well as of evolving systems. Given that alerts and warning are inherently interdisciplinary, both a social science phenomenon (their goal is to change public behavior) and a technical phenomenon (technology is required for their assessment and dissemination) this research agenda includes a wide range of sociotechnical questions and highlights the need for social and behavioral scientists and technologists to interact frequently with each other. The agenda is divided into key sections: public response, feedback and monitoring, and technical-capabilities and their impact.
Ultimately, alert and warning systems need to be designed to elicit the most life- and safety-protecting response from the public. Research has evolved over the last several decades so that we have much more information about how individuals respond to alerts and warnings. Nevertheless, as technologies shift, so do public responses; therefore, continued research is invaluable. These responses rely on several things, including
characteristics of the messages themselves, demographics of the individuals who receive the messages, and, given our increased ability to geo-target messages, the understanding of an individual’s risk in relationship to their location and the hazard. Research is also needed to understand how best to educate the public about both alerting systems and impacts of and appropriate response to particular hazards.
Protective Guidance in Enhanced Media Links
Prior research called for including uniform resource locators (URLs) in WEAs to provide more complete information on the hazard and recommended protective guidance.1 Practitioners also spoke to the importance of enhanced warnings. For example, Christopher McIntosh, former Virginia statewide interoperable communications coordinator and current director for national government industries at the geographic information system firm Esri, observed that “without context, alerts are just noise.”2
At this point, it is unclear what information is best included in a WEA message and what information is best included in linked content. Prior DHS-funded research did not examine exactly 360-character messages because the research was conducted before the Federal Communications Commission (FCC) rulemaking that extended WEAs from 90 to 360 characters.3 Furthermore, concerns remain as network congestion could be caused by people accessing an included link within seconds of receiving a WEA that includes a URL.4 Alternatively, some research found that message recipients are unlikely to open linked content and that instead they read only a few words of WEA-like messages due to stress responses.5 Therefore, research is needed to understand what message content should be included in linked media. Also unknown is how to craft WEA mes-
1 M. Wood, H. Bean, B. Liu, and M. Boyd, 2015, Comprehensive Testing of Imminent Threat Public Messages for Mobile Devices: Final Report, College Park, MD: National Consortium for the Study of Terrorism and Responses to Terrorism.
2 C. McIntosh, Esri, presentation to the committee on January 26, 2017.
3 Federal Communications Commission, “FCC Strengthens Wireless Emergency Alerts as a Public Safety Tool,” release date September 29, 2016, https://apps.fcc.gov/edocs_public/attachmatch/DOC-341504A1.pdf.
4 Federal Communications Commission, “Improving Wireless Emergency Alerts and community-initiated alerting,” release date November 19, 2015, https://apps.fcc.gov/edocs_public/attachmatch/FCC-15-154A1.pdf
5 D. Glik, K. Harrison, M. Davoudi, and D. Riopelle, 2004, Public perceptions and risk communication for botulism, Biosecurity and Bioterrorism: Biodefense Strategy, Practice, and Science 2(3):216-223.
sages so that they galvanize people to read the entire message, including potentially life-saving linked content. Research is also needed on whether it is more effective to enhance protective action guidance by using longer, 360-character alerts or by adding links to additional information. A consistent finding across WEA research is that the American public needs education on what the WEA service is as well as what protective actions to take during a variety of hazards.
Expressing Time Until Hazard Impact
The Study of Terrorism and Response to Terrorism (START) research team found that the standard WEA message elements of guidance (what to do and how to do it) and time until impact (how much time people have to take the recommended action) play major roles relative to other message elements in the outcomes of public understanding and belief of the protective action recommendation and the ability to decide how to respond.6 The other WEA message content elements are hazard, location, and source. Importantly, the START research team found that WEAs should express time as how much time until impact rather than when the message expires, as is the current practice for WEAs.
WEA messages are designed to alert about imminent threats, which the START research team characterized as hazards occurring within one hour. Other research has extensively examined the optimal timing of warnings. For example, research on tornados finds that the optimal lead time for issuing a tornado warning is from 15 minutes to just over 30 minutes.7 If too much lead time is provided, people are less likely to follow the protective guidance in a timely manner. Therefore, it is important to understand how to best express lead time.
Current WEA guidelines allow for opting out of all categories of alerts except for those issued by the U.S. President (which have never been issued). However, individuals now receive messages from an increasing number of sources and delivery channels. Past research suggestions that
6 M. Wood, H. Bean, B. Liu, and M. Boyd, 2015, Comprehensive Testing of Imminent Threat Public Messages for Mobile Devices: Final Report, College Park, MD: National Consortium for the Study of Terrorism and Responses to Terrorism.
7 S. Hoekstra, R. Butterworth, K. Klockow, D.J. Drotzge, and S. Erickson, 2011, A social perspective of warn on forecast: Ideal tornado warning lead time and the general public’s perceptions of weather risks, Weather, Climate & Society 3(1):128-140; and K.M. Simmons, and D. Sutter, 2009, False alarms, tornado warnings, and tornado casualties, Weather, Climate & Society 1(1):38–53.
alerts and warnings should be sent through as many channels as possible, but new research is needed to explore what drives opt-in and opt-out behaviors on WEA and as well as on various platforms, such as third-party applications, or local text alerting systems. While past research8 supported the delivery of alert messages across as many channels as possible, it is unknown if the increasing number of alerting channels provides the same benefits or if it instead creates a situation of over-alerting, which may result in increased opt-out rates. Furthermore, once we understand optimal times and way to alert an individual, what are technical solutions to avoid over-alerting?
While we know a lot about what a message should contain, less is known about how to best present this information. Recent research has provided clear evidence that message length influences response; messages that can fit in the initial 90-character length of a WEA message and the 140 characters of Twitter foster milling9 behavior and delayed response.10 At this point, it is unclear what information is best included in a 360-character WEA message and what information is best included in linked content. Prior WEA research did not examine 360-character messages because the research was conducted before the pending FCC rulemaking that extended WEAs from 90 to 360 characters.11 However, research suggests that a mes-
8 Considering the sharing of emergency alerting messages, there are several benefits to an ecosystem that incorporates multiple different channels. One is redundancy—i.e., a message distributed via different channels (via different technological infrastructures) is more likely to reach its intended recipients in cases where some infrastructures are disrupted. The second benefit is diversity—i.e., people rely on different types of media platforms (e.g. due to cultural preference, education, or accessibility), and so messages spread across diverse channels will reach a greater number of people. A third benefit is that people are more likely to accept and respond to emergency messages when they receive them from multiple channels and in different formats. This latter benefit, which was identified in classic studies, requires more research to confirm and better understand in the context of this new media ecosystem.
9 “Milling” refers to the process in which people seek to confirm an alert or warning, a process that has been observed across all hazard types, warning delivery technology, or message sources.
10 H. Bean, M. Wood, D. Mileti, B.F. Liu, J. Sutton, and S. Madden, 2013, Phase II Interim Report on Results from Experiments, Think-out-Loud, and Focus Groups, Comprehensive Testing of Messages for Mobile Devices, Report to the Homeland Security Advanced Research Projects Agency, Science and Technology Directorate, U.S. Department of Homeland Security, College Park, MD: National Consortium for the Study of Terrorism and Responses to Terrorism, University of Maryland.
11 Federal Communications Commission, “FCC Strengthens Wireless Emergency Alerts as a Public Safety Tool,” release date September 29, 2016, https://apps.fcc.gov/edocs_public/attachmatch/DOC-341504A1.pdf.
sage length of 1,380, the maximum number of characters supported by the Common Alerting Protocol (CAP) standards additional information field, does reduce response time.12 Research is needed to understand public response to messages that fit into 360 characters and, given that optimal message length is relevant to any text-based alerting system, continued research should be done to understand the optimal minimum length that can elicit the appropriate protective action from an alerted population.
Language and Dialect
WEA currently supports messages in Spanish but this capability falls well short of the linguistic diversity of the U.S. population. What technical challenges exist in transmitting multiple languages, or relying on the receiving device to translate messages? Additionally, protective action language, such as “shelter in place,” might be challenging to translate to various languages and dialects. Research is needed to understand the limitations of language translations (in particular machine translations)—and determine what constitutes “good enough” language so that message templates can be automatically translated—for differing languages and dialects.
Adapting to Differing Abilities
Mobile devices exist that can be used by a wide set of differently abled individuals, these include a range of tools, such as use of vibration cadences to Braille phones and text-to-speech and speech-to-text tools based on needs of physically challenged individuals. Research is needed to understand best way for enabling specific customization (translating and delivering) alerts and warnings to physically and cognitively challenged individuals. What other technologies exist to support information dissemination to differently abled individuals? How can protective action instructions shift to support diverse populations—including those of differing ages and abilities—and their caregivers? Questions around literacy are also important, in terms of both age (given that children under 10 may receive an alert on their cell phone) and reading comprehension for older adults.
12 H. Bean, B.F. Liu, S. Madden, J. Sutton, M.M. Wood, and D.S. Mileti, 2016, Disaster warnings in your pocket: How audiences interpret mobile alerts for an unfamiliar hazard, Journal of Contingencies and Crisis Management 24(3):136-147.
While a large section of the population uses smartphones, there are still others who choose not to use smartphones or use them sporadically. Considering the diversity in communication habits and availability of technology, alert and warning systems will need to consider various technologies to reach at-risk populations.
Geotargeting Alerts and Warnings
Research suggests that people do not easily understand messages that contain a map reflecting the at-risk area.13 We know that a map that shows the risk area alone is not useful. In fact, it can be counterproductive. What is needed is research on how to best communicate, possibly through visualizations, about the location of the message receiver versus the area of impact. Research is needed to determine the best way to graphically display that an individual is in an at-risk location.
Determining Locations of Interest
Individuals want to be alerted not only when they are at risk but also, for example, when their children may be at risk at school or their home may be at risk. These locations of interest can be difficult to determine. Most systems rely on the receiving device being currently within the designated warning area; if a person works outside of a WEA alert area but their home is within the WEA alert area, they will not receive the message that there home is at risk. Most receive these alerts via subscription services, such as those provided by a school system or county. Are there technical solutions so that locations of interest can be dynamically updated (rather than manually updated by the end-user)?
Location-Based Protective Action
The best protective action for an individual may vary across the impacted area—shelter in place versus evacuation. Furthermore, individuals could be prescribed specific evacuation routes to spread traffic over different routes. What are the technical challenges to these technologies? Applications such as Waze could provide some of these capabilities. What
13 M. Wood, H. Bean, B. Liu, and M. Boyd, 2015, Comprehensive Testing of Imminent Threat Public Messages for Mobile Devices: Final Report, College Park, MD: National Consortium for the Study of Terrorism and Responses to Terrorism.
are limitations to implementing these for disaster responses? How might we encourage use of these tools?
Indoor location capabilities are already being deployed in some areas, chiefly for marketing reasons as well as for meeting wireless E911 location requirements. The FCC now requires that carriers provide location information to within 50 meters of a caller’s location (inside or outside of buildings) for 40 percent of the cases currently and in the near future for 60 percent of the cases. This requirement includes both horizontal and vertical location information. However, determining elevation (which floor) is challenging and is an important research topic. Knowledge of a person’s location within a building could be used to determine the best evacuation route or if the individual should instead shelter in place.
New tools and technologies support communications between members of a community; for example, NextDoor allows people to quickly identify neighbors and communicate with those people who reside either in their neighborhood or nearby. NextDoor is already being used by public safety organizations to educate the public;14 however, little is known about how these tools are used during hazards. As discussed in Chapter 1, collaborative tools have been used to provide assistance during a disaster but mostly fueled by volunteers outside of the area. Less is known about how local residents interact with each other to build resilience and educate others about particular hazards and how that information propagates throughout a social group. This issue of course intersects with other research areas around reposting information on social media and disaster education, but is itself an important area of research.
Disaster and Alerting Education
Across the DHS-funded studies, research participants were found to be unfamiliar with WEAs15 despite the Ad Council having partnered
14 M. Helft, A Facebook for crime fighters, Fortune.com, July 1, 2014, http://fortune.com/2014/07/01/nextdoor-local-neighborhood-social-network-police/.
15 M. Wood, H. Bean, B. Liu, and M. Boyd, 2015, Comprehensive Testing of Imminent Threat Public Messages for Mobile Devices: Final Report, College Park, MD: National Consortium for the Study of Terrorism and Responses to Terrorism; B. Kar, University of Southern Mississippi, “An Integrated Approach to Geo-Target At-Risk Communities and Deploy Effective Crisis Communication Approaches,” presentation to the committee on September 1, 2016;
Very limited research exists to determine what makes for effective disaster public education. Research finds that current public education campaigns typically are ineffective because they are not specific enough and do not contain content that motivates behavior change.18 More research is needed to determine how to motivate behavior change as well as what other factors contribute to successful public disaster education campaigns.
In terms of in-person training, research points to positive results from these public education initiatives. For example, in one study researchers found that, after receiving instruction on relevant meteorological principles, participants successfully applied their new knowledge to make risk inferences from hazard graphics.19 As another example, an evaluation of research on disaster education programs concluded that these programs are effective at increasing children’s disaster knowledge and preparedness as well as household preparedness.20
Finally, emergency managers have just begun to integrate gamification into public education, and it is too soon to tell how effective this approach is for increasing individual, family, and community disaster preparedness. For example, in 2013 the Centers for Disease Control and Prevention launched the “Solve the Outbreak” mobile app, which allows users to be “disease detectives” through obtaining clues, analyzing data, solving scenarios, and saving lives in the game. So far, the app has been downloaded more than 12,000 times. Research on whether this app or
D. Glik, University of California, Los Angeles, “WEA Messages: Impact on Physiological, Emotional, Cognitive and Behavioral Responses,” presentation to committee on September 1, 2016; and others.
16 Ad Council, “Emergency Preparedness – Wireless Alerts,” https://www.adcouncil.org/Our-Campaigns/Safety/Emergency-Preparedness-Wireless-Alerts, accessed August 22, 2017.
17 M. Wood, H. Bean, B. Liu, and M. Boyd, 2015, Comprehensive Testing of Imminent Threat Public Messages for Mobile Devices: Final Report, College Park, MD: National Consortium for the Study of Terrorism and Responses to Terrorism.
18 B.J. Adame and C.H. Miller, 2015, Vested interest, disaster preparedness, and strategic campaign message design, Health Communication 30(3):271-281; J.D. Fraustino and L. Ma, 2015, CDC’s use of social media and humor in a risk campaign – Preparedness 101: Zombie apocalypse, Journal of Applied Communication Research 43(2):222-241; and M.M. Turner and J.C. Underhill, 2012, Motivating emergency preparedness behaviors: The differential effects of guild appeals and actually anticipating guilty feelings, Communication Quarterly 60(4):545-559.
19 M. Canham and M. Hegarty, 2010, Effects of knowledge and display design on comprehension of complex graphics, Learning and Instruction 20(2):155-166.
20 V.A. Johnson, K.R. Ronan, D.M. Johnston, and R. Peace, 2014, Evaluations of disaster education programs for children: A methodological review, International Journal of Disaster Reduction 9(1):107-123.
others improve preparedness and capacity to effectively respond to warnings during disasters remains to be documented. A similar research topic is how best to use tools, such as video and animation, to model protective actions and the use of tools that provide education as a disaster unfolds.
Technology is needed that solicits feedback from message recipients to help understand how the public is responding to messages and what additional information might be needed. While some tools exist to help individuals in the emergency operations center to harvest information from social media and potential feedback mechanisms in alerting applications on mobile devices, these tools will need to be more readily available. Perhaps more importantly, research is needed to understand what information would be most helpful to emergency managers and social science researchers and how best to collect the information. Not only would these additional data help emergency managers during disasters, but they could also serve to validate laboratory experiments. Several researchers have conducted experiments to explore word choice, message content, and character length. While these experiments provide valuable data, real-world analysis could provide validation and further information on public response. Tools, including those that employ machine learning and other artificial intelligence techniques, are also needed that can quickly understand and process collected information.
Consistent, well-understood, and insightful measurements can inform (and improve) response to future hazards (Box 3.1 lists information that could be valuable). Such a data-driven experimental framework would be of great interest to multiple stakeholders, including emergency managers, researchers, technologists, and so on. By building measurement into the alerts and warning system itself, researchers can gain supporting evidence for findings made in lab studies (e.g., what is the optimal message length? should we include a map or not?). By sharing information across hazards, we can learn from past experiences to create new best practices. Feedback during the life cycle of a hazard can also be integrated into future responses within the same incident. For example, low response rates to an initial message could lead to more aggressive message content in a follow-on message.
In parallel with measurements of the messages themselves, we also encourage new measurements of ancillary supporting technologies. For example, what level of engagement is seen on social media and local news websites? What content was most engaged with? What fraction of users in a region used Waze? And so on. We anticipate new data-sharing initiatives for the multiple stakeholders in the alerts and warnings ecosystem.
For example, it could be beneficial for social media companies to provide aggregate (anonymized) measurements in the aftermath of an alert.
Enhanced Cell Broadcast and Network Capabilities
Cellular networks have become the most popular access mechanism in the current scenario due to the wide use of the device. However, there are
limitations to the current implementation of cell broadcast—limitations in coverage, capability of handling message size, and inability to facilitate two-way communication. Research is needed to understand capabilities and features that may take advantage of newer broadcast technologies and supplement older technologies. Research questions include the following: Is it feasible to combine multiple cell broadcast messages core into a single, longer alert message?21 What will the impact of including a URL be on network capacity? Recent WEA testing has indicated that messages are not always delivered as expected;22 what is creating these errors? As we move to newer wireless network standards, is it possible to flexibly select among 90-, 360-, and 720-character-long messages to best make use of existing 2.5G to 3G networks as well as new upcoming 4G-LTE (long-term evolution) networks? (See Box 3.2.)
Multimodal Transport of Emergency Alerts
Today WEA is designed only for cellular transport. However, cell phones can receive data through a variety of wireless communications standards. For example, phones are commonly connected through home or public WiFi hotspots. How can WEA be adapted so that it can use multiple channels to increase the likelihood of successful delivery to the end-user? How can a single message be delivered through an increasing number of delivery channels—including government and private channels?
Bypassing Network Failure
During hazards, some cellular networks may not function properly, owing either to overload or to infrastructure damage.23 One existing alternative is the NOAA Weather Radio All Hazards, which relies on a national network of dedicated transmitters and users who obtain suitably equipped receivers.24 Several other technologies exist that might support message receipt, including mesh networks, peer-to-peer communication,
21 The Alliance for Telecommunications Industry Solutions completed feasibility studies on message length in 2015 and recommended the expansion to the current 360-character limit for next-generation networks (https://access.atis.org/apps/group_public/download.php/25045/ATIS-0700023.pdf). Additional feasibility studies may be necessary to understand future limitations.
22 Federal Communications Commission, 2017, “Report: September 28, 2016 Nationwide EAS Test,” https://apps.fcc.gov/edocs_public/attachmatch/DOC-344518A1.pdf, and in briefing and discussions with the committee.
23 The National Research Council reviewed the impact of network usage during disasters in its report The Internet Under Crisis Conditions: Learning from September 11 (Washington, DC: The National Academies Press).
24 This text was modified after prepublication.
and FM radio transmission. For example, with peer-to-peer communication techniques, such as those used by FireChat, messages might be relayed to people lacking a direct network connection. Research is needed to validate the efficacy of these various technologies and understand the implementation of these tools.
Battery Life Management on End User Devices
Power resources can be quickly drained during disasters as users attempt to find information. If main power to a home is out, a user will also lose access to Internet service and forced to rely on cellular data service that uses more battery power than in-home wireless, and will be unable to recharge devices. Research is needed to determine possibilities for reserving power resources to support alert and warning delivery, implementing battery preservation technologies across available plat-
forms, and possibly providing information back to first responders about an inability to receive messages owing to power concerns.
Role of Connected Devices
As the Internet of Things (IoT) grows, more devices in homes and throughout the environment will be available not only as an alerting channel but also to detect emergencies and potential risks. To make most effective use of these opportunities, several questions will need to be explored.
If each home were to be outfitted with a mini-weather station or stream gauges were to be deployed pervasively, what data would be most helpful and how can data be trusted enough to automatically issue an alert and warning? Automation of some alerting, based on aggregate data, would resolve latency issues around fast-moving hazards. For example, in several areas of Japan, earthquakes are detected and elevators, trains, and gas valves are immediately cut off. Could similar automation be used for other events such as an active shooter event—gunfire on campus is detected and classroom doors are locked? Machine learning is needed to understand how systems can become sufficiently data intensive and have enough situational awareness to suggest the best protective action to take—i.e., such as when to shelter in place, when to evacuate, and where to evacuate.
Best Devices for Alerting
If most electronics can deliver some sort of information to users, which devices should be used to issue which hazards? Could the increasing number of smart devices better communicate appropriate protective action? For example, if a disaster results in a boil water order, could a smart refrigerator present an alert when the water dispenser is used? (See Box 3.3.)
Milling with Virtual Assistants
As more homes are equipped with virtual assistants, such as in an Alexa or Google Home, what role will they play in milling behavior? One can envision receiving an alert from Alexa that a tornado warning has been issued and a user has asked Alexa to explain the difference between a tornado watch and a tornado warning or is asking what the best protective action might be. This might on the one hand reduce milling times because confirming information can be obtained quickly and might on
the other hand extend milling because it introduces new avenues for extended information seeking.
Security, Trust, and Privacy
A system that instructs large populations to take a particular action may represent a significant target for attacks on service availability, compromises of the integrity of valid messages, and spoofed messages. Emergency alerting systems have been directly compromised already, including the use of false Emergency Alert System (EAS) tones, resulting in the issuance of false alerts on radio and TV stations in several states in 2014.25 (Box 3.4 lists some breaches that have already occurred.) Indeed, a 2014
25 R. Wimberly, “$1 Million Fine for Misusing the Emergency Alert System,” release date May 19, 2015, http://www.govtech.com/em/emergency-blogs/alerts/1000000-Fine-for-MisUsing-Emergency-Alert-System.html.
report from Carnegie Mellon University’s Software Engineering Institute26 identifies the following attack vectors that specifically target WEA:
- An outsider obtaining the proper credentials to send malicious CAP-compliant messages, for example, to direct people toward a dangerous location rather than away from it.
- Malicious code that has infected an alert origination service could prevent an operator from posting a new alert, and hence delaying notification.
- An insider may spoof a colleague’s identity to send an illegitimate CAP-compliant message, thereby spreading false information and undermining the colleague’s reputation.
26 Carnegie Mellon University, 2014, Wireless Emergency Alerts (WEA) Cybersecurity Risk Management Strategy for Alert Originators, CMU/SE I-2013-SR-018, Pittsburgh, PA.
- Malicious attacks could make communication channels unavailable, so that an operator could not distribute an alert message through the Integrated Public Alert and Warning System.
These and related threats highlight the critical cybersecurity challenges with maintaining the integrity of existing and future alert and warning systems.
Alerts that can be reliably and undeniably attributed to their actual author is essential to warning system use. This is a particular problem when warning credentials are issued to a jurisdiction or agency rather than to the individual that should be held accountable. If the only feasible sanction for a misuse of a warning system is to threaten to cut off warning system access to an entire jurisdiction, such a sanction is unlikely to be perceived as a genuine threat. Barriers to providing personal credentials need to be studied and remedies devised. Furthermore, research on password management and security could be incorporated into system training tools.
Spoofing, in particular, has been recognized as a threat to the validity of many communication channels, including email, the web, GPS data,27 and sensor data, among many others. Already, we have seen spoofing attacks on social media to post fake messages as in the April 2013 case of hackers taking control of the official Associated Press (AP) Twitter account to post false reports about an explosion at the White House. Though discovered and corrected within minutes, the spoofed post led to a 100-point drop in the Dow Jones Industrial Average. While the drop was only momentary—the unsophisticated tweet did not follow AP style and lacked other indicators of legitimacy—more sophisticated spoofs in the future could attack multiple channels at once (e.g., the AP Twitter account, Facebook presence, and their wire service) while adopting more credible language to create even more uncertainty.
Furthermore, as the system takes advantage of these large data sets and harnessing information from the public, misinformation can pose a challenge. A misunderstanding by the public and poor reporting can create misinformation, but it can also be inserted intentionally. Quickly detecting and correcting poor information will be a valuable system capability. (See Box 3.5.)
As emergency managers begin harnessing information from users and social media and provide geographically relevant information, concerns around user privacy arise. How can we take advantage of these tools while still protecting end-user privacy?
27 N.O. Tippenhauer, C. Pöpper, K.B. Rasmussen, and S. Čapkun, 2011, On the requirements for successful GPS spoofing attacks, pp. 75-86 in CCS ’11 Proceedings of the 18th ACM Conference on Computer and Communications Security; https://dl.acm.org.