The invention relates to a
wireless communication anti-eavesdrop interference
power control algorithm based on
Q learning. The
wireless communication anti-eavesdrop interference
power control algorithm comprises steps that S1. transmission power Ps, safety evaluation coefficient Rho, interference power xi, and levels of working time ts are initialized, and working time ts of an
edge server is equally divided into k parts, and each part is used as a time slot t<(k)><s>; S2. the related quantity of the working state of the kth time slot t<(k)><s> is calculated; S3. a friendly jammer adopts a
Q learning algorithm for learning, and a decision is made in every time slot t<(k)><s> according to a
system state S(k), and then a corresponding motion xi(k) is selected; S4. after the
Q learning training of the
edge server is completed, the friendly jammer is used to schedule the motion X<k> to transmit a friendly interference
signal according to an equation V(S<k+1>,X<k>)corresponding to thecurrent state S(k). The jammer is trained by the
Q learning algorithm, and then the jammer continues trying to transmit interference power according to
information transmission power of legitimate transmitters until the optimal interference power is finally acquired, and therefore maximization of
information security and minimization of network
energy loss are realized, and ideal effect of network energy
utilization rate is improved.