Based on the broadcast nature of wireless signal, this paper proposes an approach that exploits the overhearing data omitted in conventional MAC protocols to eliminate the spatial correlation. Particularly, sensor nodes can prevent the transmission of redundant data in link layer by using the data they overhear to compress their own sensory data collaboratively before the transmission. Firstly, this paper formulates the problem that sensor nodes collaborate their compression to optimize their lifetime, and establish linear programming model for it. This paper also proposes a lower complexity (O(N²)) heuristic node selection algorithm while achieving a nearly optimal performance. Based on that, an efficient Collaborative ComPression-based MAC (CCP-MAC) protocol is designed to implement the above node selection algorithm in a distributed way. As a result, the corresponding node can receive the data it overhears from the selected sensor node subset to compress redundant data before the transmission. Experimental results show that by exploiting the data nodes overhear CCP-MAC can collaborate the nodes to compress sensory data, thereby conserve energy significantly to prolong the network lifetime.