Physical attacks are becoming more common due to the exponential expansion of connected devices, which is expected to reach 50 billion in 2025. This is because attackers may get rapid entry to the system being attacked. This risk is amplified by the increasing reliance on devices linked to the Internet of Battlefield Things (IoBT), industrial control systems, and vital infrastructure. Present anti-tamper approaches only protect against certain kinds of attacks and exhibit predictable responses when tampered with, which might limit system accessibility. Many advancements in physical assessment procedures, thinner attacks are now feasible. This work aims to enhance existing physical protection approaches by developing an intelligent anti-tamper system using machine learning algorithms. Therefore, there is an immediate need for more intelligent defences that can match the anticipated increase in adversary capabilities and ensure a longer operational lifespan. The implementation of a tiered reaction system and recovery plan led to a decrease in false alarms and an increase in operational time. Due to the increasing reliance on gadgets embedded in critical infrastructure and industrial control systems, as well as the proliferation of the World Wide Web of Battlefield Things (IoBT), the need for solutions that prevent tampering during design has surged. This research has also found a way to make the proposed system less vulnerable to adversarial learning attacks.