Machine-to-Machine (M2M) communications are expected to be one of the major drivers of future cellular networks, due to a plethora of services provided to operators and consumers. This leads to an explosively growth of simultaneous M2M arrivals, and then a bursty random access attempts that causes a severe random access congestion in addition to terminals' synchronization issues. In this paper, we proposed a novel implementation of the Access Class Barring (ACB) scheme, that mainly consists of dynamically adapting the ACB factor according to the network's overload conditions. Simulation results show that the proposed algorithm outperforms the existing solutions by improving significantly the access's success probability while minimizing radio resources' underutilization.