In the present scenario of cryptographic protocols, Physical Unclonable Functions(PUF) is regarded as the primary primitive due to its properties like uniqueness, robustness, unclonability, and unpredictability which makes a system more secure random number generators for light-weight security applications. In this paper we are proposing a PUF based chaotic pseudo random number generator as Multiple PUF-CPRNG, which is more resilient to state-of-the-art modeling attacks like Logistic Regression, Becker Attack and Deep Neural Network. On comparing with other PUF based architecture like XOR-Arbiter PUF, Double Arbiter PUF (DAPUF) and Ring Oscillator PUF-CPRNG structure, the proposed architecture is more resilient to the above mentioned attacks. In this paper, accuracy of the predicting response bits of the proposed structure has been reduced with at least 3% in comparison to the state-of-the-art PUF architectures. While comparing with the hardware complexity we are reducing more than 5% in terms of FPGA resources. The proposed design also passed all 15 statistical randomness test described in NIST 800-22 Test Suite. © 2022 IEEE.