Abstract
A true random number generator (TRNG) is crucial to support information security in IoT devices. Conventional RO-based TRNGs depend on intrinsic jitter as an entropy source. It remains vulnerable to environmental and malicious attacks, such as temperature manipulation and frequency injection, which degrade randomness and compromise security. To address these issues, the proposed design integrates a Galois-based conditioning mechanism that enhances entropy quality by reducing bias and correlation through structured XOR feedback. The TRNG is implemented with two cascaded inverters (TSI) as a ring oscillator and a Galois inverter as an extrinsic jitter, named the Galois jitter ring oscillator (GJRO). Additionally, the time-variant XOR is used for tapping bits and acts as a random switch between oscillatory and Galois states for implementing a true source of randomness in the entropy. Experimental analyses are demonstrated in SilTerra 0.13 µm with low power of 46.4 pW, area of 129.61 〖µm〗^2, and high throughput of 2.9 Gbps. The hybrid reconfigurable strategy allows the source to generate faster random bits per clock cycle in the range of 12.8 Gbps for a 2^7 parallel GJRO implementation. The random sequences are evaluated using statistical test suites including NIST SP 800-22 along with entropy estimation and autocorrelation analysis. Consequently, the entropy source is tested for its robustness against frequency injection attacks in a voltage range of 0.6V to 1.2V and a temperature range of 0 °C to 100 °C. Furthermore, the generated bitstreams are introduced into a regression-based machine language model to monitor statistical features and identify deviations from normal behaviour, thereby enabling real-time attack detection.