This paper introduces a novel hybrid physical layer authentication (PLA) method designed to enhance security in multi-node networks by leveraging inherent hardware impairments. The approach specifically exploits carrier frequency offset (CFO), direct current offset (DCO), and phase offset (PO) as multi-attribute features, improving the verification process for authorized users and enhancing the detection of unauthorized devices. Machine learning (ML) models are developed to authenticate devices without prior knowledge of malicious characteristics, resulting in robust and reliable device authentication capabilities. Experimental evaluations conducted on a commercial software-defined radio (SDR) platform demonstrate the effectiveness of the proposed approach under varying signal-to-noise ratio (SNR) conditions. The hybrid PLA scheme integrates advanced feature extraction methods with finely-tuned ML models, optimized through controlled experiments to ensure high performance across diverse network conditions and attack scenarios. Real experimental tests validate the efficacy of the proposed scheme, achieving high authentication rates exceeding 96% and reliable detection rates for malicious device attacks surpassing 95%. Additionally, the approach is highly efficient, with a mean inference time of less than 3.75 milliseconds (ms) and power consumption below 25.5 millijoules (mJ), confirming its suitability for real-time applications in energy-constrained environments.