The hide-and-seek game between stealth fighters and missile sensors just got a whole new dimension: artificial intelligence (AI), which could potentially rewrite the rules of engagement.
Stealth fighter jets are designed to conceal their radar signatures, making them extremely difficult to detect, track, and target.
However, despite stealth fighter jets’ platform-aligned faceted shaping, radar-absorbent materials (RAM), internal weapons carriage, and reduced thermal and radio frequency emissions, there is one thing they cannot hide: their heat signature.
When heat-seeking missiles lock onto fighter jets, the aircraft release flares to confuse the missiles’ infrared sensors and break the missiles’ locks.
It creates a recognition problem for the missile, as its seekers must determine whether the heat it detects is from the aircraft or from decoy flares released by the jet, often in milliseconds.
However, the infrared signature of an aircraft’s engine exhaust and the friction created due to aerodynamic heating are different from those generated by flares.
These unique heat signatures from aircraft, including stealth fighters, could provide valuable information to a missile’s IR seekers for target identification, tracking, and targeting, while rendering flares useless.
Exploiting this “weakness”, researchers in China have developed a lightweight AI system that could enable heat-seeking air-to-air missiles to recognize the infrared signatures of advanced fighter jets in milliseconds.
The system recorded high accuracy in laboratory tests against mock-up F-22 and F-35 targets.
The new system reportedly achieved up to 97.1 percent accuracy in identifying F-22 and F-35 targets during simulated laboratory tests.
“Lightweight recognition models could become widely used in future air-to-air missiles because they can provide high-speed recognition while maintaining strong identification capabilities,” An Jiangshan, first author of the study, said.
The study, published in the Chinese peer-reviewed Journal of Electronic Measurement and Instrumentation, involved researchers from the Beijing Institute of Technology. The institute is known for its strong military research background and is among the few Chinese universities subject to US sanctions.
The collaborators also included researchers from the China Airborne Missile Academy and a national key laboratory for near-surface detection, the Hong Kong-based South China Morning Post reported.
“The model enables efficient classification and recognition of targets in missile-borne scanning infrared imaging systems, achieving a recognition accuracy of 97.1 percent during testing,” the researchers wrote in the paper.
To train the system, researchers created a pool of 3,245 infrared images collected by a missile-borne scanning system. The data set included three categories of airborne targets: two types of aircraft mimicking the F-22 and F-35, and a loitering munition.
After statistical error processing, the hardware accelerator achieved 96.4 percent recognition accuracy, with an average inference time of about 1.5 milliseconds and overall power consumption of 2.2 watts.
The time frame is crucial, as a missile has only a few milliseconds in flight to detect the IR signature and identify the target. The power requirements are also crucial, as the AI accelerator must fit within a compact missile.
The researcher, therefore, also needed to solve a hardware problem. How to make the AI accelerator compact, lightweight, and not a power-hungry system.
Conventional AI models require excessive computational resources, making their integration with weapon systems and missiles impractical.
Though such conventional deep learning models can achieve high recognition accuracy, their large parameter counts and heavy computational requirements make them difficult to deploy on small embedded devices.
The team therefore developed a lightweight neural network.
The researchers reduced the model to 16.1 percent of the parameters used by previous methods. It also brought computational requirements down to 19.2 percent, according to the study.
The model uses structural optimization, batch normalization fusion, and 8-bit quantization to reduce computing costs.
Through these measures, the team reduced overall power consumption to just 2.2 watts.
To make the AI system suitable for missile-mounted platforms, the researchers also designed a dedicated AI accelerator.
The accelerator included optimized convolution modules, parallel computing structures, and data buffering methods to improve efficiency.
“The final system achieved a balance between recognition accuracy, inference speed, hardware power consumption, and hardware resources,” corresponding author Liu Ming and his team wrote.
The system is also relatively inexpensive.
“The Zynq7020 [AI] platform used in the paper costs only a few hundred yuan, and the hardware cost is relatively low,” An said.
The researchers said the proposed solution provided a lightweight structural design for missile-borne infrared scanning imaging systems.
However, the high accuracy rate of 97.1 percent achieved during testing against mock stealth targets, such as the F-22 and F-35, does not mean the system can detect operational F-22s and F-35s in real-world conditions.
When asked specifically about the method’s potential accuracy in identifying fifth-generation fighters such as the US F-22 and F-35, An said the relevant test data set involved “confidentiality issues and could not be disclosed”.
“For the simulated targets in the current [test] data set, the recognition rate can reach about 90 percent,” An added.
The experiments used simulated targets, and according to researchers, more tests and data will be needed to improve both recognition accuracy and processing speed.
The study also had other limitations.
For instance, the researchers limited their study to close-range air-to-air missile fuzes.
Furthermore, “the effect on surface-to-air missiles remains to be verified,” An added.
Nevertheless, the study points to an emerging challenge for stealth aviation. No country has solved the heat-exhaustion problem, and the evolution of AI-based compact infrared seekers for missiles could render stealth irrelevant in the future.
Notably, this is not the first time China has claimed to have developed radars or other technologies that could effectively blunt the stealth advantage of US fighter jets, such as the F-22 and the F-35.
For instance, in May last year, China unveiled its mobile meter-wave radar, which Chinese media claimed could detect American fifth-generation stealth fighters, such as the F-22 and F-35.
Showcasing its advancements in anti-stealth radar systems, China presented the JY-27V radar system, manufactured by the state-owned China Electronics Technology Group Corp (CETC), at the World Radio Detection and Ranging Expo in Hefei, East China’s Anhui Province.
Mounted on a military truck, the JY-27V is a high-mobility air surveillance radar. This next-generation meter-wave anti-stealth radar incorporates three advanced technologies: a low-frequency band, a high-power aperture, and sophisticated intelligent algorithms.
According to the state-owned Global Times, these advanced technologies enable the radar system to precisely detect stealth targets and guide an air defense system for a precision strike.
However, this same ‘stealth-killer’ radar failed miserably in Venezuela in January this year.
On January 3, 2026, US forces executed a precision military raid codenamed Operation Absolute Resolve, resulting in the capture of Venezuelan President Nicolás Maduro and his wife, Cilia Flores.
Venezuela possessed a variety of advanced radars, including the JY-27, whose capabilities observers have now called into question.
These radars were integrated into Venezuela’s air defense network alongside Russian systems like the S-300VM surface-to-air missiles, forming a layered defense around key sites, including Caracas.
However, this entire defense network collapsed as US jets entered Venezuelan airspace and operated freely.
Again, in October last year, Chinese researchers claimed to have developed a “revolutionary” dual-satellite radar system capable of detecting and tracking stealth aerial objects around the clock.
Earlier, in September 2024, China claimed that it could successfully detect a stealth aircraft using Elon Musk’s Starlink satellites.
As part of this experiment, the group launched a DJI Phantom 4 Pro drone off the coast of Guangdong. The drone was roughly the size of a bird and had a radar cross-section similar to a stealth fighter.
Despite its stealth characteristics, the target unexpectedly appeared on the screen, even though the ground-based radar did not emit any radio waves that would have produced an echo. The scientists explained that this was possible because the drone was illuminated by electromagnetic radiation from a Starlink satellite passing over the Philippines.
However, it remains to be seen how successful these claimed stealth-killer radars would be in a real-world war.
- Sumit Ahlawat has over a decade of experience in news media. He has worked with Press Trust of India, Times Now, Zee News, Economic Times, and Microsoft News. He holds a Master’s Degree in International Media and Modern History from the University of Sheffield, UK.
- He can be reached at ahlawat.sumit85 (at) gmail.com
