Congress & Conference Contributions

Recent research has shown that the joint processing of simultaneously recorded EEG-MEG signals can be beneficial compared to the separate analysis of data if the corresponding tensors have one of their factor matrices in common. In this study, we perform a joint CP decomposition of simultaneously recorded EEG and MEG magnetometer as well as gradiometer measurements via a new extension of the coupled SEmi-Algebraic framework for the approximate CP decomposition via SImultaneous matrix diagonalization (SECSI) to jointly factorize several tensors with at least one factor matrix in common. In case of the biomedical data, the four measured tensors have the factor matrix in the frequency domain in common. The developed coupled SECSI algorithm allows extracting the signal sources even in ill-conditioned scenarios. In the final processing step, coupled SECSI chooses the best out of a variety of possible factor matrix estimates, i.e., a combination that leads to the smallest value of the reconstruction error. However, for data tensors that do not exhibit significant coupling, the uncoupled solution can also be selected. The results are evaluated on measured as well as on synthetic data and compared with other state-of-the-art approaches.

In this work, we consider both channel estimation and reflection coefficient design problems in point-to-point reconfigurable intelligent surface (RIS)-aided millimeter-wave (mmWave) MIMO communication systems. First, we show that by exploiting the low-rank nature of mmWave MIMO channels, the received training signals can be written as a low-rank multi-way tensor admitting a canonical polyadic (CP) decomposition. Utilizing such a structure, a tensor-based RIS channel estimation method (termed TenRICE) is proposed, wherein the tensor factor matrices are estimated using an alternating least squares method. Using TenRICE, the transmitter-to-RIS and the RIS-to-receiver channels are efficiently and separately estimated, up to a trivial scaling factor. After that, we formulate the beamforming and RIS reflection coefficient design as a spectral efficiency maximization task. Due to its non-convexity, we propose a heuristic non-iterative two-step method, where the RIS reflection vector is obtained in a closed form using a Frobenius-norm maximization (FroMax) strategy. Our numerical results show that TenRICE has a superior performance, compared to benchmark methods, approaching the Cramér-Rao lower bound with a low training overhead. Moreover, we show that FroMax achieves a comparable performance to benchmark methods with a lower complexity.

In Software-Defined Networking (SDN) the controller is the only entity that has the complete view on the network, and it acts as the brain, which is responsible for traffic management based on its global knowledge of the network. Therefore, an attacker attempts to direct malicious traffic towards the controller, which could lead to paralyze the entire network. In this work, Deep Learning algorithms are used to protect the controller by applying high-security measures, which is essential for the continuous availability and connectivity in the network. Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) are proposed to recognize and prevent the intrusion attacks. We evaluate our models using a recently released dataset (InSDN dataset). Finally, our experiments manifest that our models achieve very high accuracy for the detection of Denial of Service (DoS) attacks. Thus, a significant improvement in attack detection can be shown compared to one of the benchmarking state of the art approaches.