Malicious software attacks are gradually becoming more prevalent, posing a serious threat to the information security of nations and large enterprises. Currently, widely applied methods for malicious software detection include signature-based detection, behavior-based detection, and heuristic-based detection. However, malicious software has been optimized to evade these detection methods by encrypting traffic and attempting to conceal communication traces with noise. While signature-based and heuristic-based detection methods offer fast and efficient detection of known malicious software, they exhibit poorer performance in detecting unknown malicious software. This paper proposes the use of periodic communication behavior to detect the presence of malicious software. The prevalent methods for detecting periodic traffic are mostly statistically based, facing challenges in accuracy due to fluctuations in real-world traffic. To address the shortcomings of existing methods for detecting malicious software using periodic traffic, this paper designs and implements a malicious software periodic traffic detection system based on Fourier Transform. The system introduces the Zeek tool to convert initial traffic pcap data tables into log files. It then employs a time series extraction algorithm to extract communication time series for each internal and external IP address pair. Finally, it utilizes Fourier Transform to calculate the periodicity of malicious software traffic and employs malicious software detection algorithms to determine whether the IP address generating this periodic traffic is a remote control endpoint for malicious software. The paper evaluates the system's detection performance using three publicly available datasets, with experimental results demonstrating an 77.3% detection accuracy on the IoT dataset. Furthermore, to address the limitations of existing publicly available datasets, such as a limited variety of attack scenarios and insufficiently diverse data, we collect a new dataset from a university campus network. The system is further evaluated on this campus network dataset, showing good detection efficiency and discovering several malicious software remote IP addresses, thus proving the value of this malicious software detection system.