Uplink Channel Estimation and Signal Extraction Against Malicious Intelligent Reflecting Surface in Massive MIMO System
share
This paper investigates the effect of malicious intelligent reflecting surfaces (IRSs) in a massive MIMO uplink.The IRSs are used by malicious users (MUs) for randomly reflecting pilot and data sequences of legitimate users (LUs) to a base station (BS), we consider the problem of channel estimation and signal extraction in the presence of the malicious IRSs. Firstly, we find that by using the IRSs, the reflected sequences are correlative to the data sequences of LUs. This correlation challenges traditional channel estimation techniques. To be more precisely, we prove that the right singular matrix of a received signal at the BS is a function of the correlation between the legitimate and reflection data in a large-scale antenna regime. As a result, the correlation caused by malicious IRSs degrades the performance of traditional channel estimation methods based on eigenvalue decomposition (EVD) of the received signals. To address this challenge, we propose a signal extraction and channel estimation method to combat malicious IRSs. More precisely, geometric arguments, such as convex hulls of extracted signals exhibit insensitive to the correlation. The convex hulls are thus utilized to provide signal extraction and channel estimation criteria. To optimize these criteria, we develop an extractor that can capture the convex hulls of desired signals from noisy signals. Based on this, we formulate two optimization problems, whose global minima are solved to perform signal extraction and channel estimation. Experimental results show that when the IRSs randomly flop reflection phases with probability of 0.8, the proposed method outperforms the EVD-based method by more than 5 dB in the sense of normalized mean square error (NMSE).
READ FULL TEXT