It　is　necessary　to　evaluate　the　credibility　of　software　user　behavior　for　the　timely　identification　and　control　of　abnormal　user　behavior　risks　in　order　to　ensure　the　security　of　the　software　system　and　the　credibility　of　user　behavior.　The　user　behavior　log　dataset　is　generated　by　using　the　simulated　county　and　city　government　office　system,　and　the　credibility　analysis　algorithm　model　of　numerous　individual　learning　and　ensemble　learning　are　constructed　to　evaluate　the　effect　of　the　credibility　analysis　model　in　the　study,　and　by　comparing　and　analyzing　the　heterogeneous　ensemble　learning　algorithm　models　of　different　combinations,　the　accuracy　of　software　user　behavior　credibility　analysis　has　been　improved　to　more　than　97%.　This　study　shows　that　the　model　is　able　to　identify　untrusted　users　in　a　flexible　yet　accurate　manner　and　guarantees　the　security　of　the　fixed　domain　work　system.