On security requirements analysis for multi-agent systems