The internet we-media has large data characteristics, and negative energy words and deeds often break out. So public opinion detection becomes a major and difficult problem in network regulation. Based on software definition network, honeypot technology and distributed architecture, integrates "traffic level" detection and "process level" detection two levels, then negative energy public opinion test algorithm was designed and a virtual honeypot active public opinion detection system through constructing the exception behavior data set and the sensitive text type data set. Practice proved that virtual honeypot active public opinion detection system can better complete the thematic tendency monitoring tasks from we-media circle, and it can provide a new perspective for public opinion detection technology from we-media circle.