Disrupt the Spectrum With AI

Technology is advancing at a dizzying pace. We are seeing breakthroughs in materials, medicine, energy, computing, and robotics, some of which have the potential to fundamentally change entire components of industry and defense. Identifying disruptive technologies that could have a major impact on the Navy requires a careful look at what the battleground of the future will look like and where the key fights will take place. While technological capabilities in the air, surface, subsurface, and space domains have advanced steadily and will continue to do so, the greatest disruption in the near future likely will occur in the electromagnetic spectrum (EMS). The reason for this is simple: technology is creating new ways to wage war in the EMS.

Despite the Department of Defense’s (DOD’s) historical view of the EMS as a resource and not a physical space to be contested and denied, the military is beginning to recognize the opportunities as well as the potential threats in the EMS. DOD’s draft directive on electromagnetic spectrum operations (EMSO) states that the armed services will now “recognize the EMS as an operational domain [emphasis added] comprised of all electromagnetic energy.”¹ The Navy must embrace this view and take advantage of new technologies to dominate this domain.

The Navy relies heavily on the spectrum to operate effectively and to wage war. We use the EMS to sense our operating environment, communicate (both voice and data), navigate, attack, and defend ourselves. We have tapped eagerly into the EMS to support every mission area, from strike warfare to logistics to cyber warfare. With electromagnetic benefits, however, also come vulnerabilities—both for us to exploit against our adversaries and vice versa. Command of the spectrum will enable decisive actions that can cripple and even destroy an adversary’s war-making capability without firing a single kinetic shot. Overlaying the tenets of modern warfare—maneuver, seizing the initiative, maintaining a superior decision cycle, and massing fires—on the EMS reveals game-changing advantages that favor the side that masters this domain.

The draft DOD EMSO directive acknowledges the importance of EMS superiority and defines it “as the degree of advantage in the EMS domain that permits the conduct of military operations at a given time and place without prohibitive interference, while affecting an adversary’s ability to do the same.” Accordingly, we must identify and develop technologies that will allow us to master the EMS faster than our adversaries. Artificial intelligence (AI), supported by quantum computing, is that technology.

The concept of AI is not new. For decades people have written about it, even devoting themselves to the pursuit of machines or computers that can think and learn for themselves. B. J. Copeland defines AI as “the ability of a digital computer or computer-controlled robot to perform tasks commonly associated with intelligent beings.”² To date, the limiting factor in the development of AI has been computing power. The sheer number of calculations necessary to execute AI code prevented significant developments, but processing speed and capacity now have progressed to the point that AI programs have become viable. Advances such as quantum computing will provide AI with the muscle required to apply this technology to spectrum operations. B. Wang states,

Some researchers are predicting that the market for “universal” quantum computers that do everything a supercomputer can do plus everything a supercomputer cannot do—in a chip that fits in the palm of your hand—are on the verge of emerging. . . . Quantum computer power scales far more than classical computer systems. . . . Quantum computers are already helping to improve machine learning. Universal Quantum computers that would drive high performance computing would drive artificial intelligence as well.³

AI already is being incorporated into the fields of visual and speech recognition, logistics, medicine, and games. The most widespread application of AI thus far is in speech recognition, including applications we use daily (Apple’s Siri, Hey Google, and Amazon’s Alexa). AI’s ability to learn speech patterns, understand dialects, and make highly educated guesses by analyzing digitized sound waves makes its application to the EMS particularly suitable. The distinguishing feature that gives AI the ability to learn is based on the concept of deep learning:

Deep Learning allows computational models composed of multiple processing layers to learn representations of data with multiple levels of abstraction. These methods have dramatically improved the state-of-the-art in speech recognition, visual object recognition, object detection, and many other domains. . . . Deep convolutional nets have brought about dramatic improvements in processing images, video, speech and audio. . . Representation learning is a set of methods that allows a machine to be fed with raw data and to automatically discover the representations needed for detection or classification.⁴

To understand how AI will be the critical game changer in tomorrow’s conflict, we need to examine the technological advancements of new EMS-dependent systems. The radars and communications gear of the 1940s through the 1990s were predominantly hardware-based and used electronics such as vacuum tubes and crystal sets to generate carrier waves sent through the EMS. These signals had a fixed or limited selection of parameters, such as frequency, pulse repetition interval, amplitude, modulation, and power. Each system’s transmission signal was unique and discrete, allowing databases to be developed that would help operators associate emitters to platforms. Advanced EMS-dependent systems are adopting software-based digital waveforms that can dynamically change, rendering older, database-driven identification systems less effective. Software-defined radios, solid-state power amplifiers, electronically scanned antennas, and spread spectrum technology are making activities in the EMS harder to detect and classify. The tempo of EMS agility developing in advanced spectrum-dependent systems is rapidly moving beyond the capabilities of legacy intelligence, surveillance, and reconnaissance systems.

Control of the EMS will require commanders to seize and maintain the initiative—to observe, orient, decide, and act (the classic OODA loop) in and through the EMS with superior information and at a faster rate than the adversary. EMS control therefore will require the technological capability to sense and digitize the EMS in real time (observe), identify patterns and understand EMS activity at machine levels (orient), take into account all factors in the electromagnetic operating environment (EMOE) to determine the best course of action (decide), and then implement changes to our emissions (act). AI will be the disruptive technology that allows us to accomplish this highly complex and difficult task, providing the Navy with a level of information superiority unmatched by any other fighting force.

The first step in gaining command of the EMS is sensing and digitizing the spectrum in real-time, a feat that until recently has been out of the reach of technology. Military systems operate from the lowest parts of the spectrum (ultra-low frequencies) all the way up to millimeter wave and above. The Defense Advanced Research Projects Agency (DARPA) recently announced it has developed analog-to-digital converters (ADCs) able to accomplish this task. DARPA’s Arrays at Commercial Timescales (ACT) program supports the development of an ADC with a processing speed nearly ten times that of commercially available, state-of-the-art alternatives. The new ADC samples and digitizes spectrum signals at a rate of over 60 billion times per second—fast enough to directly detect and analyze any signal at 30 GHz or below—a range that encompasses most operating frequencies of interest. Scanning through frequencies today requires costly hardware with long development cycles, and the new ADC can provide a “one-stop shop” for processing radar, communications and electronic warfare signals.⁵

Once the spectrum is digitized, the power of AI can be applied to this data: identifying patterns, matching frequency-hopping signals with discrete platforms, picking up spread spectrum waveforms that would otherwise remain undetected, and (through the iterative process of deep learning) developing an understanding of the EMOE. Deep learning algorithms currently employed to recognize speech and visual patterns can be advanced to detect and understand EMS activity, creating an unprecedented level of situational awareness in real time. As AI programs learn the unique characteristics of a given geographic EMOE, patterns of all periodicities will be detected, organized, and stored. Signals with characteristics that change at intervals ranging from microseconds all the way up to days, months, and even seasons will reveal patterns, leading to EMOE understanding. Repeated “patterns of life,” such as the emissions of command-and-control nodes, ships, and aircraft, will contribute to a better understanding of adversary operations, even to the level of how a platform performs its kill-chain process of find-fix-target-track-engage-assess (F2T2EA). This knowledge will provide commanders with superior information on which to make decisions, providing much more timely indications and warning of impending attack.

AI also will be applied to cybersecurity and cryptology in support of Navy information superiority. To this purpose, the latest commercial quantum computer already is being put to use:

D-Wave Systems Inc., the leader in quantum computing systems and software, today [24 January 2017] announced general commercial availability of the D-Wave 2000Q™ quantum computer. D-Wave also announced the first customer for the new system, Temporal Defense Systems Inc. (TDS), a cutting-edge cyber security firm. With 2000 qubits and new control features, the new system can solve larger problems than was previously possible, with faster performance, providing a big step toward production applications in optimization, cybersecurity, machine learning, and sampling.⁶

AI will play a central role in decrypting signals and protecting our networks, weapons, mechanical systems, and navigational systems from cyberattack. AI will pick apart detected signals, decoding them for exploitation and preventing malicious code from penetrating our networks and systems. To accomplish this, AI will stand as a gate guard at the EM entry points of our deployed platforms—their EMS apertures.

To detect, localize, characterize, and understand relevant activity in the EMS, the significant impact of atmospheric conditions on the transmission of EM waves must be considered. Space and planetary weather both have cause-and-effect relationships within the EMS, resulting in phenomena such as reflection, refraction, ducting, and trapping, which change the theoretical propagation of EM energy. The interrelationships between atmospheric weather and emissions are dynamic and complex across the three-dimensional EM environment. Applying AI to analyze weather patterns and conditions to understand their effects on EM propagation will drive the information wedge between the Navy and our adversaries deeper. AI will leverage weather sensors to iteratively model atmospherics in real time, providing highly accurate propagation estimates in support of operational planning and execution. This feedback loop will aggregate atmospheric sensor data, analyze atmospheric effects on existing transmissions of known locations and power levels (commercial radars, TV and radio stations, etc.), model current conditions to predict effects on friendly EM activities, and measure the actual impact of the atmosphere on friendly EM activities. With predicted-to-actual corrections applied, the loop will then be repeated to maintain a historical and real-time awareness of atmospheric impacts on the EMOE. Through deep learning, AI will develop a machine understanding of the relationship between atmospherics and EM propagation based on geography, climate, seasons, space weather, and planetary weather.

Armed with superior understanding of the EMOE, including friendly, neutral, adversary, and environmental activities, Navy commanders will be able to make rapid decisions across the entire range of operations, defending our forces from adversary attack (both in the EMS and in the traditional air, land, sea, space, and cyberspace domains) while bringing our full warfighting might to bear on the adversary’s weaknesses. These decisions will be made and implemented at machine speeds (through calculations performed by AI to optimize EM activities such as changing frequencies, power levels, and modulation types) and at the human level as recommended by AI (for example, best courses of advancement, target priorities, attack axes, engagement ranges). AI will be able to predict the performance of an adversary’s sensors and weapons given EMS activity and effects of weather, providing highly accurate counter-detection and self-defense estimates.

To leverage fully this decision-making process, AI systems must be networked across ship, air, and submarine platforms to coordinate and synchronize EM activities necessary to meet mission requirements (sensing, communicating, and transferring data) while preventing electromagnetic interference and avoiding detection. Finally, AI can help determine the most effective offensive operations in the EMOE, such as jamming, intrusion, lasing, meaconing, or EM cyberattack by analyzing the monumental amount of data and variables only AI can churn through. At every step, new information—digitized in real time—will be processed to provide commanders with an immediate understanding of effects achieved by these offensive activities. The adversary, continuously on his heels, will be unable to respond in kind.

To seize this EMOE advantage, development should be divided into two components that work in parallel: a quantum computing hardware/operational system team (HW/OS) and a software/graphic user interface (SW/GUI) team. Too often, acquisition programs fail to deliver because requirements are written too broadly and too much is asked of the final product—from the outset and during acquisition. Development of EMOE sensing should be viewed as a dual-evolutionary process that leverages cutting-edge, commercially available hardware (refreshed at a rate synchronized with ships’ maintenance availabilities—approximately every 18–24 months) to host AI software that is developed more rapidly and pushed more frequently—as quickly as every 6 months. This will be similar to how people buy a new smartphone every two to three years but get new software pushed to that phone every month or so.

To make possible such an aggressive HW/OS fielding schedule, strict nonproprietary specifications and standards must be established and enforced. Quantum processors, memory, and storage will be designed to be “plug and play,” so that upgrades take hours or days instead of months. HW/OS will follow a “waterfall” operational life, where high-value units always receive the latest technology and HW/OS gets moved to smaller and lower priority ships as it ages, to balance capability with capacity.

SW/GUI fielding time can be reduced significantly by shifting operational test and evaluation (OT&E) to ships that are in the work-up phase of deployment. This would introduce challenges in coordination, but the costs and time saved by leveraging unit and integrated phase training events could be significant. Crews would receive training on the new SW/GUI concurrently, and those same sailors could provide feedback on new system performance, tactics, techniques, and procedures (TTP) that is lacking in our existing OT&E construct.

Command and control of the EMS quickly is becoming acknowledged as critical high ground that must be taken to prevail in tomorrow’s conflicts. The draft DOD directive on EMSO states a key functionality of the EMS enterprise is to “develop and field electromagnetic spectrum command-and-control capabilities that provide commanders with reliable, relevant, and networked tools to conduct detailed and dynamic planning, execution, and assessment of EMSO to gain and maintain EMS superiority throughout the full range of military operations.” This tall order will be a key offset capability that will keep our Navy armed with the warfighting tools needed to dominate our adversaries for the next decade and beyond. Artificial intelligence, powered by supercomputing capacity, is the disruptive technology that will ensure the Navy gains and keeps the war-winning high ground of the electromagnetic spectrum.

1. “DOD Directive 3610.aa Electromagnetic Spectrum Operations (EMSO) Policy” (draft), Office of the Chief Information Officer, Department of Defense, www.dtic.mil/whs/directives.

2. B. J. Copeland, “Artificial Intelligence,” Encyclopedia Britannica, 12 January 2017, www.britannica.com/technology/artificial-intelligence.

3. Brian Wang, “After Quantum Dominance, A Surge in Quantum Computer Investment, Boosts to AI and Overall Technological Progress and a Quantum AI Singularity,” Next Big Future, 9 January 2017, www.nextbigfuture.com/2017/01/after-quantum-dominance-surge-in.html.

4. Geoffrey E. Hinton, Yann LeCun, and Yoshua Bengio, “Tutorial: Deep Learning (Overview),” Microsoft, 17 December 2015, www.microsoft.com/en-us/research/video/tutorial-deep-learning/.

5. “New Chips Ease Operations in Electromagnetic Environs,” DARPA, 11 January 2016, www.darpa.mil/news-events/2016-01-11.

6. “D-Wave Announces D-Wave 2000Q Quantum Computer and First System Order,” press release, D-Wave Systems, Inc., 24 January 2017, www.dwavesys.com/press-releases/d-wave%C2%A0announces%C2%A0d-wave-2000q-quantum-computer-and-first-system-order.