Data really is the new ammo

Real-world data plays a critical role in the algorithms that support modern military operations, where the capacity to detect, identify, geolocate, and engage targets is the essence of conflict. Yet this data gathering and analysing capacity has not received the attention that it deserves. While academics and policymakers have recognised the critical role of data – epitomised in the oft-repeated statement,  ‘data is the new oil’ – such attention has not translated into meaningful consideration as to what type of data deterrence and defence require and of how such data can be acquired.      

Even events like Israel’s reliance on artificial intelligence for targeting Hamas operatives and the decision by the Trump administration to suspend the provision of critical intelligence to Ukraine have generated little discussion about how, in the digital age, the capacity to gather and analyse massive amounts of data is a key source of power. 

Signal and noise

The importance of operational data for modern warfare stems, among others, from three critical features of modern military systems. 

First, long-range high-precision munitions are completely useless without targeting data: if you don’t know what you need to strike, high-precision is no longer an asset. And, given the high price of precision munitions, using them in an indiscriminate manner is expensive and inefficient. 

Second, targeting data requires the capacity to detect, identify, and geolocate static targets; track mobile targets; and engage them. Such capacity improves as the amount of data increases. Much in the same way that the app Shazam requires an extensive library of songs to accurately identify the song being played, target identification and classification require data about enemy platforms and systems, as well as about the environment of operations. One notable example is the Zhousidun dataset, a Chinese-origin dataset that contains unique external features of American ships that could permit them to be identified and possibly targeted. 

Third, the capacity to detect, identify, geolocate, track, and engage targets, together with the capacity to transmit this information in real-time from sensors to shooters, relies on electromagnetic signals (eg, radio transmission). Because electromagnetic signals are so critical for long-range high-precision warfare, electronic warfare like blinding enemy radars has come to play a predominant role in modern military operations. For this reason, since the Second World War, countries have invested massive resources in electronic countermeasures like jamming and spoofing. 

Countering contemporary enemy electronic systems requires data about them, which, in an age of machine learning, is more accessible than ever before. This hunt for adversary data played a role in the decision of the US to expel Turkey from the F-35 fighter jet program in 2019, after Turkey bought Russian-made air defence systems that could capture and store data on the unique radar signature and electronic warfare countermeasures of the F-35. This would have made the data available for further analysis by Russian electronic warfare and radar engineers. 

Implications for European strategic independence

This data has important implications amidst the renewed European push for strategic independence. After the Trump administration interrupted military support to Ukraine, some European NATO countries expressed their concern about relying on American military technology and reiterated the need for European countries to become strategically independent. Some analysts, observers, and politicians went as far as to call for the cancellation of the acquisition of F-35. 

Such policy concerns are understandable, but they reveal some serious lacunae in the current debate: the operational data that the US has accumulated over the past decades, and over the past three years in Ukraine in particular, represents a key source of advantage that will be very difficult for European countries to replace. 

In addition to important capacity gaps – most prominently in airborne intelligence, surveillance, target acquisition, and reconnaissance, as well as in long-range surface-to-surface missiles and limited stocks of precision munitions – European countries would have a hard time building the library of signals that modern military operations require. In this area, they would have to remain dependent on the US. 

The same goes for the F-35 project.  European countries are presently concerned about the F-35’s potential ‘kill switch’ and the possibility that in the future the US might limit access to spare parts and software updates. Much less discussed is access to the critically important mission data the F-35 gathers. This incredible amount of data determines who can train and perfect machine algorithms and hence who can deal with enemy advances from new technologies to countermeasures. It is not clear whether European countries would have access to such mission data files. 

Data as a resource for algorithm development

Since the full-scale Russian invasion in 2022, the US has provided Ukraine with extensive military assistance, including military platforms, munitions, training, maintenance, spare parts, and, of course, intelligence. Such intelligence includes data from satellites, intelligence reconnaissance, surveillance drones, airborne early warning and control aircraft, ground sensor imagery, communications, and electronic signals. 

The public debate about US assistance to Ukraine has, however, paid little attention to the fact that the US has in turn gained a significant advantage in the form of a massive data gathering and analytics operation. Military conflicts generate vast amounts of unstructured data – images, electronic signals, radar returns, acoustic signatures, flight profiles, troop movements, and more. Ukraine has offered the US a live testing ground to collect this critical data and use it to refine algorithms. 

Combining intelligence from multiple sensors and using real combat data increases the probability of enemy asset detection, including of those that are camouflaged or hidden; the ability to distinguish military targets from civilian objects and from decoys; and identification of different equipment types. 

Moreover, the volume of data from a large-scale conflict is vastly greater than that from exercises and simulations, and the type of data is significantly different as well. Such real-world data comes from a dynamic and highly contested environment where algorithms face real adversarial countermeasures, tactics, and technologies. 

In Ukraine, the US has had the opportunity not only to learn about the patterns of employment by Russian forces, but also the functioning of Russian electronic warfare systems (by detecting jamming signals) and the approaches and choices of Russian electronic warfare engineers (by analysing Russian equipment captured by Ukraine). 

European countries have also operated platforms carrying intelligence sensors in Ukraine (eg, Swedish airborne early warning and control aircraft), but the scale of the US assistance is vastly larger than any other country, and has included the integration of multiple types of sensor data, from infrared cameras, electronic warfare systems, and airborne, ground, and satellite radars. This disparity means that European countries would not be able to fully compensate for the loss of American support, and the lack of access to prior data and analytics might reduce their military effectiveness. 

Looking beyond

Of course, access to past data is not a silver bullet. Algorithms trained in Ukraine’s unique environmental and operational context – a largely terrestrial conflict with specific fauna and flora, against mostly Soviet-era equipment – might not translate to a direct advantage to other theatres, like the Indo-Pacific, which is covered by large bodies of water and has very different fauna and flora, as well as technologies. Data is not necessarily fungible. 

While vital for modern warfare, these algorithms also cannot anticipate an adversary’s technological, tactical or operational countermeasures – jamming or disinformation could reduce the effectiveness of algorithms based on analysis of earlier data. This is an important reminder of the limits of technological solutions to human conflicts and of the inherent vulnerabilities of even highly advanced technologies.