SWIFT: Intelligent Spatio-Temporal Metamaterial Massive MIMO Aperture Arrays with Hybrid Learning-based Channel Classifiers for Spectrum-Efficient Secured Wireless Communication

Dates: 10/01/22 โ€“ 09/30/25
Award Amount:ย  $750,000.00
Award #: 2229384

PI: Chung-Tse Wu
Co-PIs: Narayan Mandayam, Waheed Bajwa


For the next-generation network systems, it is envisioned that billions of mobile and Internet-of-Things (IoT) devices will communicate with each other to provide unprecedented connectivity in a highly smart and secure manner. It is thus of paramount importance to ensure spectrum utilization efficiency as well as physical-layer (PHY) security of mobile and IoT networks against any attack. While spectrum efficiency can be improved by using many antennas in a massive multiple-input multiple-output (MIMO) fashion, the power consumption, hardware complexity and cost will increase drastically as each antenna element requires a dedicated RF transceiver. In parallel, although end-to-end encrypted sessions between the edge devices and the gateway can be used for secure communications, high computational and battery burden of such cryptographic strategies are usually a concern in large-scale systems. To overcome these challenges, this research will harness the spatial dispersion control capability and introduce time modulation for subwavelength metamaterial (MTM) unit cells to create a new intelligent space-time modulated MTM (IST-MTM) antenna aperture, which not only can provide dynamic control of radiation characteristics allowing the optimal spectrum utilization, but also PHY secure transmission for wireless links enabled by directional modulation (DM). Moreover, a new hybrid model-based and learning-based approach (HyPhyLearn) will be incorporated to conduct channel classification even when limited training samples are available. The IST-MTM-based secure communication scheme along with the HyPhyLearn channel classifier will have a profound impact in next-generation wireless networks by providing a highly secured and spectrum-efficient communication scheme, which can be deployed in future 6G networks for smart homes/cities and vehicle-to-vehicle communications. By leveraging outreach activities at Rutgers University, the educational plan of this project aims to broaden participation of graduate, undergraduate and high school students, including underrepresented minority groups in STEM, in relevant research on microwave/antenna technologies, signal processing, machine learning, and wireless communications.

Current state-of-the-art antenna systems have treated antennas as fixed radiator hardware, where extensive signal processing is required to achieve the desired specifications, e.g., cryptography-based security or digital beamforming, thereby increasing system cost and power consumption. Such fixed antenna hardware design also hinders the optimal utilization of the spectrum owing to the incapability of dynamically compensating channel effects, thereby reducing the spectral efficiency. Likewise, conventional data-driven classification methods used for authentication need large amount of sampling data to achieve a certain degree of accuracy. To address these issues, the research of IST-MTM MIMO antenna array with hybrid learning-based classifiers aims to achieve the following synergistic outcomes: (1) At the transmitter side, the IST-MTM antenna array will distort the signals towards unintended directions, resulting in a high bit-error-rate (BER) and thus preventing eavesdropping, while preserving the original signals along the directions for authenticated users, and thereby forming a secure communication link. Moreover, the IST-MTM antenna can reconfigure its dispersion characteristics through programming each sub-wavelength MTM unit cell for optimal spectral utilization. (2) At the receiver side, the HyPhyLearn method can conduct classification and authentication even with limited number of training samples to avoid spoofing attacks without the need of using additional spectrum, which is particularly useful for DM, since the transmission link of the IST-MTM array may result in limited training data. (3) In a multipath environment, DM may cause interference for legitimate users, resulting in a degradation in SNR. This multipath effect can be mitigated by using HyPhyLearn to perform accurate classification between desired signals and interference. As such, the unique integration of the IST-MTM array and HyPhyLearn will form a synergistic and complementary effect towards achieving spectrally efficient secured wireless communication networks in the future.

This award reflects NSF’s statutory mission and has been deemed worthy of support through evaluation using the Foundation’s intellectual merit and broader impacts review criteria.