Game-Changing Predictions for Cybersecurity in 2025

As the space domain continues to evolve, so do its threat actors. In the proverbial game of keeping data safe and secure, how is the cybersecurity world keeping up?

Via Satellite spoke with cybersecurity and space experts to predict what’s to come in 2025, including the impact of rapid advancements in Artificial Intelligence (AI) and quantum technologies.

AI and the Need for Human Supervision

AI continues to transform our world by automating everything from everyday tasks – such as restaurant recommendation or traffic routing – to more mission critical activities such as managing large satellite constellations, data processing on both the bird and on the ground, as well as enhancing edge-computing capabilities.

“There's a lot of excitement in the space industry to adopt things like AI,” SIXGEN Technical Director Jacob Oakley tells Via Satellite. AI “optimizes what happens at the edge in space. So instead of downloading a bunch of images, I can run a lightweight algorithm and only download one in 10 images – saving storage and bandwidth.”

With these benefits come an increased risk for failure, especially in a domain as challenging as space, and with more crowded orbits. According to a paper in the Journal of Space Safety Engineering, the resident orbital object (ROS) population in Low-Earth Orbit (LEO) grew from approximately 19,000 in Nov. 2018 to 28,000 in Nov. 2023, representing a nearly 50 percent increase.

When satellite fleets grow larger, AI becomes a natural choice for satellite management in order to reduce the human workload. Yet, Oakley warns: “if something goes wrong, the ability to recover [is compromised]. Because at some point physics becomes an issue – if one thousand satellites start spinning out of control, it will now take longer [for a human] to notice.”

With there still being a clear need for human supervision over AI, this concept is now known as human-on-the-loop (HOTL). This is opposed to the more recognized concept of human-in-the-loop (HITL) and is a fundamental shift in human involvement when AI is part of the operational equation.

Jennifer Krolikowski, former director of the Chief Information Office (CIO) for Space Systems Command, now CEO of Plan Z, shares that “instead of [humans] being in the loop with AI, a human-on-the-loop aspect should be considered to confirm and verify that what [the AI] is finding or doing is accurate and true.”

Supervising AI-enabled satellite management and data processing not only keeps the data moving but is essential to ensure accurate data outcomes and uncompromised training data sets.

Krolikowski further explains that “AI can help make the observe, orient, decide, and act (OODA) loop tighter, but it does not preclude us from making educated decisions on what is being produced from the AI. There’s a lot of [AI] hallucinations and there can training on the wrong data sets which could give you information slanted in a way it should not be.”

Cyberinflight Founder and CEO Florent Rizzo adds to the sentiment that humans are still a key element for decision making with AI, especially when it’s high stakes, such as integrating into a security operation center (SOC). “You may have a lot of false positives, especially at the beginning of using an AI tool,” he says.

“There is a learning curve. AI is like a kid learning a new language: It takes a little bit of time, but when the situation is known, the certainty grows higher and the level of entropy grows higher from the learning engine, and then you can actually trust it,” Rizzo says.

‘Security by Obscurity’ No Longer Applies

Software-defined satellites (SDS) were a breakthrough in the space industry, as Via Satellite reported on the development beginning around 2013. With satellites having the ability to be adjusted and reconfigured on-orbit, it increases lifespan and potential applications.

For many years, the space industry had assumed that because systems were too complex, nobody would be able to understand or hack them, according to Rizzo. He tells Via Satellite that the space industry is no longer protected by “security by obscurity.” Sectors like the banking sector have been considering such types of attacks for the past 20 years, but this was not the case in aviation or space.

“Now that the space industry is incorporating more commercial-off-the-shelf (COTS) libraries, software and hardware that are ready to use, and there may be vulnerabilities that could be exploited by a hacker,” Rizzo says. “The exploitation of these new capabilities could be a new type of sophisticated attack.”

“The democratization of space in general will make it a better target for attackers with the use of codes, and democratization of the usage of codes, for example,” he concludes.

Who’s in the Room with Us?

Dr. David Bray, chair of the Accelerator at the Loomis Council with the Stimson Center, highlights the importance of hardware verification, a departure in what is commonly considered cybersecurity vulnerabilities within the space industry. He cites a U.S. Department of Defense (DoD) “bake off” where some U.S. components failed deep hardware interrogation.

Bray elaborates: “About two years ago, the Department of Defense had a ‘bake off’ of about 37 companies to see what a deep hardware interrogation could do, meaning ‘Is the hardware really what you think it is and only what you think it is?”

In one section, 800 hundred pieces of U.S. components (built by an unnamed manufacturer) were scanned, and 65 percent of them failed the deep hardware interrogation. Although Bray clarifies, this “doesn’t necessarily mean that they’re all nation state exploits — it could just be memory that doesn’t match what you expect. But it shows that on the hardware level we are not verifying things before they are put in deployment, whether they’re in operational environments underwater or in the air.”

Failing deep hardware interrogation may not be important for less mission critical uses of AI, but Bray says that for things that are truly mission-critical like defense, banking or critical infrastructure, there is some evidence that things are compromised before they are received through insider threats.

“Check the hardware when you receive it then also periodically scan it, to make sure it is what you think it is,” Bray says. “We start with the hardware and build on up, because otherwise, you could do everything right in the software realm and still be exploited.”

Preparing for a Post-Quantum World

According to Google’s Cybersecurity Forecast 2025, it’s time to prepare for an age of post-quantum cryptography.

“Quantum computing is seeing impressive gains in the number and the time that you can keep qubits stable,” David Bray tells Via Satellite in response, as qubits are essential to quantum computing. “There are some obstacles, but we are making impressive progress.”

He continues: “The National Institute of Standards and Technology (NIST) has put out guidance saying that if you’re not thinking about quantum-resistant encryption now, you should be, because we know that there are nation states that are storing encrypted data with the belief that sometime in the next 10 years or so they will be able to decrypt it and it’ll be useful to them.”

The guidance he’s referring to is NIST's finalized post-quantum cryptography standards from 2024, so many organizations will naturally begin adopting these in 2025. These standards represent a culmination of eight years of work and include computer code from the encryption algorithms, and instructions on how to use and implement them.

“The latest guidance from NIST on quantum-safe encryption/key transport and cryptographic signing is designed to help mitigate attacks by adversaries and large-scale quantum computers. These attacks could potentially break encryption, and ultimately compromise sensitive data,” according to Google’s forecast.

Krolikowski agrees on the importance of keeping up with quantum: “Everyone is looking at quantum. We’re in a buffer zone right now. Everything is on the precipice, and we need to address and think about how we are going to solve or adopt to the problems quantum will have when it pops off. If we wait until quantum pops off then we’re too little too late, because we probably won’t have the techniques or the tools to help counter it.”

“We definitely have to be thinking about it now, and what ways we can make things ‘quantum-proof,” she emphasizes.

Phishing and Deepfakes

Aside from threat vectors that explicitly target the space industry, there are threats that target anyone with a phone or computer and an internet connection, such as phishing. A term coined in the 90s, the sophistication of phishing methods has only accelerated with the evolution of AI.

Bray is concerned that evolving generative AI capabilities “are good at taking training sets and producing things that look like the training sets. We already know that the quality of phishing emails has gone up dramatically. No longer can we use typos and misspellings to detect them. But now it's even more challenging, because these actors can scrape information found online and send you a targeted email including a reference to your name or even a family member.”

The evolution of AI has also enabled “deepfakes,” a newer term which according to a report from the U.S. Department of Homeland Security, is “an emergent type of threat falling under the greater and more pervasive umbrella of synthetic media, [that] utilize a form of artificial intelligence/machine learning (AI/ML) to create believable, realistic videos, pictures, audio, and text of events which never happened.”

Of note in the report is that the threat from Deepfakes is not necessarily the technology used to create it, but from the natural human inclination to believe what is seen. The resulting Deepfake, or “synthetic media,” does not have to be advanced or even believable “in order to be effective in spreading mis/disinformation.”

In addressing the threat from deepfakes, Bray comments: “There's evidence that with using generative AI, you only need at least 37 seconds to make a voice that sounds like that person … What we’re seeing is the consequence of the Turing Test in which the goal was for a machine to convince a human that it was human.”

“I worry that we've unleashed a whole set of technologies that are going to be really good for foreign state as well as non-state actors, to do some really nasty stuff,” he concludes. VS

The Next Game-changing Cybersecurity Development?

Via Satellite asks the experts: “If you could highlight one game-changing cybersecurity development that will redefine how we approach digital security in 2025, what would it be?”

Jacob Oakley: “It’s AI-related,” Oakley says. When companies raced to make models better and faster, “People were going to whichever [AI model] was the most effective, and didn’t care if it was expensive … So instead of who can build the sharper sword, it’s like, who can dump a box of butter knives on the floor?”

He continues: “With all these companies no longer focusing on achieving the bleeding edge, like ‘How do I push the boundary further?’ they are now asking ‘How do I get cheaper and lower to the ground and get this [AI] out on more things so that I permeate the market base more?”

Jennifer Krolikowski: “There’s a lot [of development] that’s going to be around blockchain,” Krolikowski says. She specifies that this isn’t about the cryptocurrency side of blockchain technology, but instead, it’s about data security and permissioning, and having data credentials archived to the blockchain. “The blockchain side even helps with the quantum problem to some extent, as well.”

Dr. David Bray: “Hardware! We have paid so much attention to software, so I’m going with hardware, because we may discover we were, in some cases, building castles on sand, and the sand was not solid underneath.”

Florent Rizzo: “Software-defined satellites will change the face of the space industry, and it will have pros and cons,” Rizzo says. “SDS has the capacity to optimize the performance and the life of the satellite, but on the other side, it could be exploited by attackers.”