The chaotic sequences of polynomials have a promising but underutilized basis on the design of high-entropy stream ciphers, to mitigate constraints of predictability and low randomness in the design of conventional generators. The present paper suggests a hybrid chaotic-polynomial model that incorporates nonlinear iterative map with adaptive coefficient perturbation. The model is also tested with NIST statistical test suites as well as real-time encryption benchmark on IoT data. The outcomes show the enhancement of entropy up to 0.998, which is 21.4 % more than with LFSR-based ciphers, and a 34.7 % decrease in correlation. The proposed method has an 18.2% better key sensitivity, and increased resistance to differential attacks when compared to the existing chaotic systems. Furthermore, the performance of the encryption process in the embedded devices (Raspberry Pi and ESP32) showed that high speed encryption (up to 224.7 Mbps) is possible with the proposed framework with relatively small memory consumption, which confirms the applicability of the proposed framework to the low resource-based secure IoT communication environment. The solution guarantees a scalable and secure communication provisioning a strong paradigm in next-generation cryptographic stream ciphers.