Word-based exploits