This paper presents a solution to the problem of synchronization in Wireless Sensor Networks (WSN) by adopting the two-way message exchange mechanism. Using the affine model of the unsynchronized local clock, a Maximum Likelihood (ML) estimator is established to obtain the best consistent solution. However, directly implementing an ML estimator is often difficult because the ML cost function is highly nonlinear and non-convex. In order to effectively obtain the ML estimator, a new double iterative algorithm is proposed, which obtains the ML estimator by searching the sensor source node information and clock parameters (clock skew and clock offset). The convergence of the proposed algorithm is proved, and the estimation accuracy of the sensor and clock parameters of the node is adjusted close to Cramer-Rao Lower Bound (CRLB) under low noise conditions. Compared with the existing methods, the algorithm has the advantages of higher computational efficiency, fewer anchor nodes and lower communication cost.