A modern processor is not just a die in a plastic package. On a common substrate, side by side, sit the processor, memory and power chips, often made by different companies in different countries. They are connected by copper traces thinner than a human hair. Whoever can "read" these traces will learn how the system is built and will be able to copy it, counterfeit it or alter it unnoticed.
This is exactly the task - seeing the internal structure without opening the package - that reliability engineers tackle. Using X-ray tomography (a method of layer-by-layer scanning) or an acoustic microscope, they find defects. But the same tools are available to attackers. Artificial intelligence accelerators, defense electronics and critical infrastructure systems are especially vulnerable: their components pass through many hands, and at every stage copying or tampering is possible.
Developers are already trying to confuse attackers: they add decoy traces and dead-end bridges. However, all these tricks are fixed at manufacture. If a cracker solves the scheme of one board, he can repeat it for all copies.
A group of researchers led by Himanandan Reddy Kottur of the University of Florida has proposed a different approach. They created a software environment, SUBSTRATE-CLOAK, that allows the configuration of connections to be changed using a key. The work was presented at the IEEE Electronic Components and Technology Conference (ECTC) 2026 and published in its proceedings.
Microelectromechanical switches (MEMS) are embedded in the substrate - tiny mechanical contacts that physically open and close a circuit. When the device is turned on, a microcontroller checks a digital certificate confirming authenticity, then extracts the key from protected storage. If the key is correct, the switches close in the required combination, connecting some components and leaving decoy paths open. Without the key, an X-ray image will not show which traces actually work.
The number of possible configurations grows very quickly. Even a small switch matrix (a grid where each intersection can connect, turn or remain open) gives 120 variants in the simplest case, and with flexible routing more than 3000. In one analyzed example with four components, the number of possible connections reached 2 to the 119th power - brute-forcing that is impossible.
The tool works with standard board design files. It reads the netlist, assesses which connections are critical for security, and selects locations for switches taking into account component importance, additional trace length and mounting density. The rest of the design remains unchanged. The authors demonstrated the work on boards designed in KiCad and confirmed that the modified designs pass standard checks.
It is important to understand: SUBSTRATE-CLOAK is so far only a software environment, verified at the schematic level. No substrate with MEMS switches has yet been manufactured. The next steps are physical design, manufacturing and laboratory testing. In addition, the switches add a small resistance and capacitance (parasitic parameters), so the approach is suitable for lines at speeds of roughly up to 50 Gbit/s, but not for the fastest channels. For secure computing, avionics, sensors and infrastructure control systems, where trust matters more than peak speed, such a trade-off may be justified.
According to the authors, the main thing is a paradigm shift: the substrate stops being a passive "wire" and becomes an active line of defense. The board itself turns into a lock that can be reprogrammed. And if a device is lost, stolen or resold, it will not give up its secrets just because someone looked inside.