In Effective Java (Chapter 7), it says
Note also that we did not use Date’s clone method to make the defensive copies. Because Date is nonfinal, the c
Consider this code:
public class MaliciousDate extends Date { /** malicious code here **/ }
public class SomeClass {
public static void main(String[] args) {
MaliciousDate someDate = new MaliciousDate();
Date copyOfMaliciousDate = someDate;
Date anotherDate = copyOfMaliciousDate.clone();
}
}
Since copyOfMaliciousDate
is of type Date
, you can call clone()
and it will return a Date
object, but calling clone
on copyOfMaliciousDate
executes the code written in the MaliciousDate
class because the instance stored in copyOfMaliciousDate
is a MaliciousDate
.
I haven't read the book you quoted from, but that paragraph gives a poor justification and offers no protection against any sort of attack.
The quote mentions that an attacker with the ability to load code into your program could potentially submit a Date
subclass with malicious methods, for example returning a subclass of Date
from clone
.
But that's only a minor way an attacker with the ability to load code can cause harm. They could also:
System.exit()
to stop your program, andIf the attacker is running code in your process, the game's over and your process is compromised, and this silly little guard is not going to help.
Maybe you think that clone is bad from a design standpoint, and that's fine, but please don't pretend that not using it will protect you from some security threat, because it won't.
clone()
is widely regarded to have been a failed experiment for a number of reasons. In this case, someone passing in a Date
could have passed in an EvilDate extends Date
whose clone()
method sneakily returned a copy that was still mutable by someone else.