Single Transistor Holds Two Memories, Turning Time Itself Into an Encryption Key

Some secrets are not meant to be broken, only delayed. Exam questions, sealed bid prices, and embargoed announcements all share a peculiar security requirement: the information may already be in your hands, but it should remain unreadable until an agreed moment arrives. Conventional cryptography struggles with this problem because it treats access as a binary condition — either you possess the key or you do not. Now, researchers at the Korea Advanced Institute of Science and Technology (KAIST) have demonstrated a fundamentally different approach, one in which the passage of time is enforced not by software or external clocks, but by the physical behavior of a single semiconductor device. The work, led by Professor Kyung Min Kim of the Department of Materials Science and Engineering, was published online in the journal Advanced Materials on September 10, 2026.

The heart of the breakthrough is a device the team calls a dual-state programmable oxide transistor, or DUPOT. A transistor, at its simplest, is a tiny semiconductor component that controls the flow of electric current, acting as the fundamental switch and amplifier of modern electronics. What makes the DUPOT remarkable is that it can be set — or programmed — into two distinct electrical states that operate independently of one another. One state behaves like long-term memory, holding its configuration for an extended period. The other behaves like short-term memory, changing gradually over time once it has been set. In this context, memory does not refer to storing files; it refers to the device’s ability to retain an electrical state that was established by an applied voltage.

The two states arise from two entirely different physical mechanisms inside the oxide material, and understanding them is key to appreciating why the device works. When a positive voltage is applied to the gate — the terminal that controls the current — electrons become trapped in hafnium oxide, or HfO₂, an insulating layer that does not easily conduct electricity. This charge trapping shifts the threshold voltage of the transistor, which is the voltage at which current begins to flow through the channel. Because the trapped charges remain in place for a long time, this state persists and functions as the long-term memory. It is stable, reliable, and essentially frozen until deliberately rewritten.

The short-term memory operates on a different principle altogether. When a negative voltage is applied to the gate instead, oxygen ions — electrically charged oxygen particles — migrate to the interface between the insulating layer and the channel through which current flows. This accumulation reduces the saturation current, the level of current the device maintains under fixed operating conditions. Crucially, once the voltage is removed, the oxygen ions do not stay put. They gradually redistribute themselves back through the material, and the saturation current slowly recovers toward its original value. This slow, predictable relaxation is what the researchers exploited as a time-varying signal: the device literally embodies a countdown in the drift of its own ions.

What sets the DUPOT apart from conventional memory transistors is the independence of these two control knobs. Traditional memory transistors store information mainly by adjusting the threshold voltage alone. The DUPOT, by contrast, can set the threshold voltage and the saturation current separately, meaning the voltage at which current begins to flow and the amount of current that ultimately flows can be programmed as two distinct parameters. The long-lasting state serves as the reference for the encoded information, while the time-varying state governs when the target information actually becomes visible. In other words, one knob defines what the message is, and the other defines when the message can be read.

When these operating characteristics are extended from a single device to an array of transistors, the basis for time-based cryptography emerges. The researchers configured how the current in each device changed over time and established a readout criterion for each element of the array. Depending on how the array was programmed, they demonstrated two distinct encryption schemes. The first is time-release encryption, in which information appears only after a certain amount of time has elapsed. In their experiment, the difference in current between the target pattern and the background grew steadily over time, and the letter “T” became readable after 8.4 seconds. Before that moment, the pattern was indistinguishable from noise; after it, the message resolved into view.

The second scheme is more striking: time-bound encryption, in which the correct information can be read only within a specified time window. In the demonstration, a decoy pattern — the letter “K” — appeared at 1.5 seconds, while the target letter “T” emerged only at the designated time of 37 seconds. The same physical array therefore displays different information depending on when it is read. An observer who looks too early sees a plausible but false message; an observer who looks at the right moment sees the true one. The researchers note that these specific time points were chosen for the purposes of the experiment, but the principle they illustrate is general: the readout time itself becomes a condition of decryption, layered on top of whatever key is required to restore the original information.

The team also proposed a way to apply this concept to cloud environments, where data are stored on remote servers and accessed over the internet. In this scheme, different time conditions are applied to the ciphertext stored in the cloud and to the decryption key held by the user. Ciphertext, in cryptographic terms, is data that has been converted into a form in which the original information cannot be recognized, and a decryption key is the information needed to convert it back. In a demonstration using the device array, the researchers restored the original information by combining the ciphertext and the key that matched each time condition. This architecture means that even if an attacker intercepts the ciphertext, the decryption process remains gated by a physical timing condition that the attacker cannot simply accelerate.

“The significance of this work is that long-lasting memory and time-varying memory can each be controlled independently within a single transistor and used for time-based encryption,” said Professor Kyung Min Kim. He added that the team plans to build on the time-dependent changes of the semiconductor itself to develop the technology further for cloud and smart-device security, as well as for a range of information-processing technologies. The statement underscores a conceptual shift that the study represents: a semiconductor is conventionally understood as a device that stores or processes information, but the DUPOT shows that it can also determine when that information is read. The technology could protect information that must not be disclosed before a certain time, or that must remain readable only under specific temporal conditions — scenarios that range from scheduled disclosures to time-limited access credentials.

The researchers are candid about the current limitations. The technology cannot revoke access to copies of information that have already been obtained; once a message has been read within its permitted window, the scheme cannot reach into a recipient’s records and erase it. Moreover, the present study has demonstrated feasibility at the array level rather than in a fully deployed security system. The team’s next steps are clearly defined: lower the operating voltage, improve device-to-device uniformity, and enhance consistency over repeated operating cycles so that the devices can work stably across a variety of environments. Beyond security, the researchers aim to extend the underlying physics to neuromorphic devices, which mimic how neurons in the brain work and connect — a fitting destination, since the same short-term memory dynamics that enable a cryptographic countdown also resemble the fading traces of biological synapses. The study, with Ph.D. candidate Huisu Noh as first author and Professor Kyung Min Kim as corresponding author, was supported by the Next-Generation Intelligent Semiconductor Technology Development Program of the National Research Foundation of Korea. If the remaining engineering hurdles can be cleared, the humble transistor may soon do something no software lock can: make time itself the gatekeeper of a secret.

Subject of Research: A dual-state programmable oxide transistor implementing time-based cryptography through independent long-term and short-term memory states

Article Title: KAIST develops transistor-based encryption that controls when information can be read

Article References: KAIST develops transistor-based encryption that controls when information can be read. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: KAIST, transistor, encryption, time-based cryptography, oxide semiconductor, hafnium oxide, threshold voltage, saturation current, oxygen ion migration, cloud security, neuromorphic computing, Advanced Materials