New FeFET Device Advances Homomorphic Encryption For Secure Cloud Computing

As data privacy concerns grow in an increasingly interconnected world, homomorphic encryption offers a potential solution, as reported today by Spectrum.

This advanced encryption method allows data to be processed without ever revealing the original information to the server. However, the computational power needed for this encryption has long been a limitation for devices like smartphones and IoT devices, says Spectrum.

Now, a team of engineers from Peking University in Beijing may have found a breakthrough with a new device that makes homomorphic encryption more feasible for commercial electronics.

Homomorphic encryption works by encrypting data on a device before it is sent for processing. The data is then processed in its encrypted form, ensuring that no one—including the server processing it—can access the original information, reports Spectrum.

After the computations are complete, the data is decrypted, leaving the user’s privacy intact.The main challenge with homomorphic encryption has been the significant computational power required to perform the necessary calculations, notes Spectrum.

Current Internet of Things (IoT) devices lack the processing capabilities to carry out the complex mathematical operations needed. To overcome this, the Peking University team developed a device using arrays of ferroelectric field-effect transistors (FeFETs), notes Spectrum.

These transistors are optimized to handle the encryption and decryption process with high accuracy while minimizing the computational load, says Spectrum.

Kechao Tang, an assistant professor of integrated circuits at Peking University, explained to Spectrum that by implementing these novel semiconductor devices, they can enable commercial electronics like smartphones to use cloud computing power while still protecting data.

FeFETs differ from traditional transistors by incorporating a layer of ferroelectric material, which can store electric polarization without needing an electric field. This allows FeFETs to produce more unpredictable random numbers, making encryption harder to crack, as reported by Spectrum.

The device can encrypt and decrypt data more efficiently by simplifying the process into a single step, compared to the two-step process typically required, notes Spectrum.

The FeFET array can process the data more effectively by combining multiple inputs—such as the encryption matrix, data to be encrypted, and an additional vector—into a single computation. This results in faster encryption with lower power consumption, reports Spectrum.

Tang also noted to Spectrum that the new device offers a higher accuracy rate compared to other encryption methods using resistive random-access memory (RRAM), making it a more reliable option for secure data processing.

The FeFET array achieved a 99.6% accuracy rate, a significant improvement over previous RRAM solutions. Looking ahead, Tang hopes to see this technology integrated into smartphones, allowing users to securely encrypt data before uploading it to the cloud and decrypt it when needed, as reported by Spectrum.