Estimated Reading Time: 8 minutes
By Jared King
Editor’s Note: In January we published an article describing a few of the technologies born from the Air Force Third Offset Initiative. The technologies described leverage the air domain to facilitate an advantage over peer adversaries. In response to the article, a reader asked about potential offset opportunities from the space and cyber domains—the two other primary domains the Air Force leverages to enable air superiority. Inspired by this thought, this article explores potential offset opportunities through the cyber (EMS) domain.
Commissioned in 2014, the Third Offset Strategy established initiatives in game-changing technologies such as autonomy, hypersonics, and directed energy to gain a technological advantage over peer adversaries China and Russia. Although the current administration may have placed a lower priority on these established initiatives, the acknowledgement in the current National Security Strategy that the US advantage over China and Russia is on the decline justifies the imperative to maintain and even expedite initiatives that may exploit offset opportunities.
Just as the Second Offset utilized gains in information technologies to facilitate the development of precision guided weapons, the Third Offset seeks to leverage the cyber domain (recognized as the EMS domain by OTH) to facilitate faster and smarter next-gen weapons. Much of the discussion surrounding the game-changing technologies that may allow for an “offset” capability revolve around leveraging the cyber domain to develop weapons with network-enabled or high-speed capabilities that provide an advantage in conventional warfare. Although the development of weapons capable of dominating the air domain will without a doubt play a significant role in ensuring future military superiority, the cyber domain will play an equally critical role in enabling these weapons.
A primary vector for the Third Offset Strategy is exploitation of advances in artificial intelligence and autonomy that include the use of deep learning systems, human-machine combat teaming, and network-enabled weapons. Then perhaps maintaining superiority over the cyber domain itself is where the real “offset” lies for future operations. This article makes the case for a “cyber offset” as a key imperative for the US to gain an asymmetrical advantage in the future operating environment.
“My CPU is a neural-net processor; a learning computer.”
– The Terminator
When thinking of artificial intelligence, one of the first associations that people make is with the 1984 “Terminator” movie. However, the Terminator demonstrates the ability to learn from its experience and interact with humans in a mostly natural manner, making the portrayal more akin to what’s commonly known as cognitive computing. Cognitive computing is artificial intelligence, but not all artificial intelligence is capable of being classified as cognitive. To understand why cognitive computing is an offset capability, it’s important to define the term.
IBM is the leader in investment in the artificial intelligence realm and also boasts the flagship computer for cognitive computing: Watson. Yes, I’m referring to the Jeopardy winning computer. Since IBM is the leader in the industry, I’m going to use their definition of cognitive computing as the foundation for the term. IBM defines cognitive computing as: “Interactive decision-making and reasoning over deep domain models and evidence-based explanations, using Artificial Intelligence/Machine Learning tools.”
Cognitive computing isn’t just a new capability; it’s widely published as the third era of computing. The first era of computing began with computers built to tabulate sums in the early 1900s, and then programmable systems of the 1950s marked the beginning of the second era of computing. Computing has made leaps and bounds in its capability and power, but it’s what computers can achieve that marks the beginning of new eras. As impressive as Quantum Computing is, it’s the evolution of processing power. Cognitive Computing is a system that can learn and reason with its experiences in order to provide arguments and hypotheses for complex and meaningful bodies of data.
With the evolution of computers, so will the evolution of malware follow. Cognitive malware has the potential to become the ultimate hacker. Hackers are currently limited by connectivity, sleep, bandwidth, and security. Cognitive malware has the potential to overcome all of these limitations. Our interpretation of cognitive malware is a program that can be given a task, gain access to its intended target(s), and then over time, by learning it’s environment, can map the network, establish a beachhead, spread to key targets, complete its task, and then clean up all evidence it existed on the network. A typical hacker would take months to years performing the same tasks, but humans have limitations. Cognitive malware would only sleep when its target slept, would always be connected (and therefore always learning), and would eventually seamlessly integrate with the target network. Without limitations like bandwidth, operating hours, connectivity, human error, and emotion (patience), cognitive malware becomes the ultimate hacker.
Is cognitive malware the silver bullet and/or atomic bomb of the cyber (EMS) domain? I don’t think so. With great computer capability must also come great computing power. This means that cognitive malware on a foreign network would need to be able to utilize local computing power to function. Currently, the resources necessary for cognitive computing are significant. IBM’s Watson, the flagship of cognitive computing, is ten racks of servers with 15 TBs of RAM, 2880 processor cores, and can operate at 80 teraflops. This means that the dream of an autonomous cognitive computing hacker is on hold until we figure how to shrink this computing power into a laptop, or until IBM figures out how to minimize the processing power required to foster AI learning. However, the requirement for large amounts of processing power turns cognitive computing into an early advantage for network defenders.
Defensive Cognitive Computing
The requirement for computers with large amounts of memory and processing power in order to train and operate a cognitive system lends itself to the defense of a network. Central defensive platforms like firewalls and intrusion detection/prevention systems follow a similar model. A network-based cognitive defense system could provide unprecedented oversight over a network. A single device with the computing power to support a cognitive defensive system will become a reality before cognitive malware is able to become efficient enough to achieve widespread application.
A cognitive defensive system could study the normal behavior of the network, and then key in on abnormalities to identify malicious activity. This type of analysis already exists, but cognitive computing will be able to study larger data sets and at a quicker rate then existing methodologies
The true power of cyber goes beyond digital applications, as cyber (EMS) serves as the backbone of the other domains by connecting Air, Space, Maritime, Land, and Human domains. There is of course the direct application of cognitive computing to traditional weapon systems, creating autonomous capability that removes the risk of blood on the battlefield. Beyond autonomous soldiers on the battlefield cognitive computing will provide benefits to many other enablers of the military:
- Planning – In 1997 Deep Blue, the predecessor to Watson, was able to best then-World Chess Champion Garry Kasparov by using algorithms to determine its next chess move based on probability of success. This inspired the creation of “freestyle” chess leagues where players could compete, assisted by computer programs. Some players would rely entirely on these programs, while others would compete unassisted. The players that combined their skills with the computers results dominated the leagues. The strategic guidance of the player mixed with the tactical acuity of the program generated an insurmountable outcome for the competitor. This same concept could be applied to planning staffs around the Department of Defense to create the most dynamic and well informed plans to achieve US objectives on the battlefield.
- ISR – Another area where cognitive computing could enhance the warfighter is by digesting immeasurable amounts of ISR collection data and consolidating that data for human interpretation. Cognitive computers can analyze various sources of raw data, identify trends, and develop evidence-based recommendations on the intelligence.
- Logistics – Cognitive computing’s ability to reason like a human over massive sums of data enhances military logistics by optimizing sourcing and predicting delays in delivery. Modeling supply chain data with cognitive computing can help logisticians increase delivery capability by comparing transportation modes, routes, and predicting delays based on experience with the system.
- Medical triage – A cognitive medical system could triage wounded combatants and provide evidence-based treatment solutions to combat medics, assessing urgency and level of care required. Additionally, based on experience, cognitive computers can rapidly improve results by sharing results as a collective system.
- Maintenance – Similar to medical triage, the maintenance system can become more efficient by employing cognitive computing to diagnose issues in mechanical weapon systems. By aggregating sensor data from the machine, results can be analyzed, reasoned, and suggested to human maintainers. Additionally, onboard cognitive computers could be used to instantly triage maintenance concerns while the weapons are operating on the battlefield. This type of live analysis could inform the operator of degraded capabilities and recommended courses of action to overcome or remediate these issues.
These are just a few very specific examples of how cognitive computing could be utilized to enhance operations across all domains. The possibilities grow as the understanding of cognitive computing increases, and the dependence of domains on the EMS continues to expand exponentially. Cognitive computing is truly an offset technology that will be created as a Cyber system in the EMS, but will inevitably spread across all warfighting domains as a critical capability.
As asymmetrical warfare is becoming the new normal, the next “offset” may not necessarily emanate from weapon systems traversing the traditional land, sea, and air domains as most of the Third Offset technologies suggest, but instead demand a complete superiority of the hub of these technologies—the cyber application of the EMS domain. Gains in cognitive computing in addition to reliable and resilient defensive systems against this technology may be where the EMS domain will have the most impact in future offensive and defensive operations. Utilizing cognitive computing for cyber applications in addition to acting as a powerful force multiplier for multi-domain operations may very well prove to be the next offset the US military is looking for.
Jared King is a cyber operator in the United States Air Force. He is currently a student at Air Command and Staff College at Maxwell AFB, Alabama.
The views expressed are those of the author and do not necessarily reflect the official policy or position of the Department of the Air Force or the US government.