This paper proposes a novel approach to separating background (BG) and foreground (FG) traffic based on periodicity analysis. As BG traffic is commonly periodically generated by applications, this trait is leveraged to effectively detect BG traffic. Concretely, the Period Candidate Array (PCA) approach is proposed to extract only necessary information from long and sparse traffic flows, hence quickly detects the flows' periodicity with low computational cost. The PCA works directly with "on-site" traffic without depending on historical data as in machine learning methods. As a result, the proposed approach can be immediately applied to the real world traffic management systems. In addition, the PCA properly works with latencyincluded traffic affected by network delays. Experimental results reveal the effectiveness and efficiency of the PCA compared to the conventional methods in terms of computational cost, memory usage, and independence to historical data. 978-1-4799-5952-5/15/$31.00 ©2015 IEEE II. RELATED WORK BG traffic detection is an emerging topic in network traffic analysis and management [6], [7], [8] aiming at improving energy efficiency [1], [3] and communication quality [4], [6]. BG traffic is traffic generated by devices to maintain their network connectivity such as network management packets (ARP, DHCP, IMCP,. . . ), network service handshake (Net-BIOS, DNS,. . . ), or applications heartbeats (Windows update, Yahoo News,. . . ) [1], [3], [4]. In contrast, FG traffic is generated by users on real communications such as web-surfing. As BG traffic is not immediately useful for users, it can be delayed to some extend to save the device (battery, memory, computational capability,. . . ) or network resources (bandwidth, radio channel,. . . ), specifically at critical times.User activity analysis is a suitable approach to improve QoS and QoE of mobile phone based network services. Net-Sense [9] and LiveLabs [10] are two interesting live projects that study user activities on smartphones including social networking, location-based services, screen state, etc. User activity involvement inferred from the screen state (on/off) is associated with traffic generated/received by the device's network interface to detect BG traffic [3], [4]. The essential issue of this method is that a specific application must be installed in the UE leading to user stressfulness, requiring more resources and computational cost. Moreover, BG traffic detected on UEs would be only useful for optimizations on individual devices (e.g., battery saving). It would be far to utilize this detection to manage the network traffic since it needs further information about the network load and the amount of BG traffic on the whole network. This work resolves this issue by a novel solution at the network edges to quickly and effectively capture, analyze and optimize a large amount of network traffic, without imposing any policy on the UE.