Generalised space shift keying (GSSK) technique proposed for massive multi-input–multi-output systems has a higher information transmission rate due to the activation of multiple antennas at the same time, and also has significant advantage in terms of hardware cost. However, the maximum likelihood detector has a very high complexity which makes it computationally intractable for large-scale GSSK systems. In this Letter, a sparse detector is proposed by exploiting the inherent property of sparsity in GSSK. Different from the existing compressed sensing (CS)-based detectors, the proposed detector utilises Euclidean distance instead of inner product operation for antenna matching. Moreover, it can remove erroneous antenna indices by backtracking to promote the detection performance. The simulation results show that the proposed scheme outperforms the existing CS-based detectors while maintaining low complexity.