In current years, with the widespread commercial application of contactless palmprint recognition based on deep learning, the unique backdoor defect of deep neural networks (DNN) is seriously threatening its security. The attacker tries to embed hidden backdoors in DNN, ensuring the model performs correctly on clean data but outputs manipulated results when triggered by predefined inputs. However, existing palmprint backdoor attack methods usually adopt the same texture/principal line-based triggers for all palmprint samples, which makes the attacks easily detectable and mitigated by existing backdoor defenses. To tackle these issues, in this paper, we proposed an enhanced contactless palmprint backdoor attack method with invisible sample-specific triggers (ECPBA_SST). Specifically, the proposed ECPBA_SST explored an encoder-decoder architecture to generate the additive noise trigger fused with palmprint features. To this end, an asymmetric label poisoning strategy with non-uniform poisoning tactics is designed to avoid the latent space separation. Experimental results on a number of contactless palmprrint datasests demonstrate that the proposed ECPBA_SST not only outperforms mainstream methods in attack effectiveness, but also exhibits exceptional stealthiness.