How would you go about deciding reliably whether a given Javascript program would perform a malicious computation on certain inputs? As the concept of malice se

How would you go about deciding reliably whether a given Javascript program would perform a malicious computation on certain inputs? As the concept of malice seems somewhat hard to pin down, let's simplify the problem a bit: it's enough to decide whether that JS code stops at all or if it would just run forever. How to get that ad smasher written?