Attack from malicious access points in cell-free massive MIMO systems: Performance analysis and countermeasure

Abstract

This paper investigates physical-layer security (PLS) in a cell-free massive MIMO network. It has been demonstrated that malicious access points (APs) can significantly compromise the downlink achievable rate by conducting channel estimation during the uplink training phase and transmitting precoded interference signals during the downlink data phase. A performance analysis has been conducted to derive a closed-form expression for the downlink achievable rate. To ensure secure communication, this paper proposes the injection of artificial noise (AN) into the uplink training phase to prevent malicious APs from acquiring channel state information. Due to limitations in degrees of freedom, third-party nodes are proposed to transmit AN. Given the possibility of no null space existing between third-party nodes and legitimate APs, a singular value decomposition-based precoding method is proposed, and an expression of the achievable secrecy rate is obtained. Both analytical and numerical results indicate that the proposed AN-aided scheme can effectively counteract the downlink interference attacks of malicious APs and outperform traditional power allocation methods in maximizing the achievable rate.