|
Advanced nano-optical security systems for secure IoT devices
|
|
In the modern world, where the number of Internet-of-things devices in society is increasing exponentially,
the development and dissemination of security technology that guarantees the safety and security of individual devices is a universal and urgent research task.
To address this, we are investigating “nano-optical metrics”, where we will construct a new security layer that functions based on nano-optical technology,
in addition to the existing physical security layer that includes watermarks and holograms.
|
|
|
|
We propose using a nonduplicable hyperfine structure processed for IC chips to identify information unique to each chip for their authentication.
Advanced measuring instruments unsuitable for general use, such as electron microscopes and atomic force microscopes, are typically used to read such hyperfine structures.
However, in this study, by applying the interference measurement technique using a common white light source, we have simplified the reading and achieved a high-accuracy authentication system using the reading results.
|
|
|
|
Next-generation optical architecture for physical AI
|
|
While parallelization is a simple and effective measure for achieving highly efficient information processing, it is challenging to implement parallelization at the architectural level in existing electronics.
Nevertheless, since light is an elemental medium that can reflect spatial parallelism and high dimensionality, parallel processing is often applied in optical technology.
We will develop a dedicated nano-optical architecture for machine learning based on nano-optical technology, in which the utilization of large-scale computing systems through parallelization is desirable.
|
|
|
|
We will create a neural network, a typical execution model in machine learning, using a dispersed structure of fluorescent nanoparticles and quantum dots.
The energy of the localized field excited inside a quantum dot network by an optical signal input from the outside exhibits various behaviors,
depending on the network structure, such as immediate emission as fluorescence or emission after autonomously propagating in the network.
Consequently, the fluorescence output from the entire quantum dot network does not show a simple linear relationship with the input but instead, shows a nonlinear input−output relationship essential for operation as a neural network.
In this study, we will evaluate the performance of a quantum dot network by measuring the ultrafast fluorescence phenomenon caused by the incidence of an ultrashort pulse laser on the prototyped quantum dot network at a single-photon level.
We are also working on demonstrating machine learning based on the obtained optical input−output relationship.
|
|
|
E-mail: tate[at]ed.kyushu-u.ac.jp
