<p>Industrial Internet of Things (IIoT) deployments that rely on IEEE 802.11 are particularly exposed to selfish back-off manipulation attack. In such scenarios, a subset of nodes deliberately shortens its contention window to seize more transmission opportunities than the protocol intends. This behavior distorts medium-access fairness, increases collisions, and eventually degrades overall throughput, especially in dense industrial networks. In this work, we propose SAFE-MAC, a fairness enforcement framework that regulates medium access under such adversarial conditions. The design combines three elements that operate in sequence. First, an entropy-based detector highlights abnormal access patterns, and a Kalai–Smorodinsky bargaining step applies immediate contention window corrections. Second, a Generalized Bargaining Equilibrium component keeps track of past violations so that proportional fairness can be preserved over time, rather than only in isolated slots. Third, a Regret Matching⁺ learning module updates contention windows adaptively using cumulative behavioral feedback. Simulation studies indicate that SAFE-MAC improves Jain’s Fairness Index by about 25%, lowers collision probability by roughly 55%, and recovers nearly 98% of the throughput observed in non-adversarial baselines. Experiments with 100–500 nodes further suggest that the framework scales well and remains robust against diverse selfish MAC behaviors in IIoT settings.</p>