On the Resilience of Multi-Agent Systems with Malicious Agents