you gotta threat model this better or else it looks like fud. is it better/more secure than the previous implementation?