The vulnerable part will be the communications. If an enemy can decipher the communication channels then they can send false instructions. Having said this how is it different to the situation today when the soldiers are for the most part human. Would it really be that hard to convince a unit to retreat or to fire artillery on their own frontline troops? As we already occasionally see 'blue-on-blue' incidents I would argue it isn't.
The danger with robots would certainly be greater but it is not a new threat.
True, but it is rather difficult to convince an e.g. blue soldier that he is, in fact, red. Not so with robots, assuming you can reverse engineer their software.
I agree that the potential consequences of a breach would be significantly greater with robots than with human soldiers.
Most of the wars I'm aware about during the past few decades have been highly asymmetric technologically or have been between technologically limited forces. In the first case the likelihood is that only the more technologically advanced force would utilise robots and would be able to secure them beyond the capabilities of the technologically inferior opposing force to compromise. In the second case neither side is likely to have robots fighting.
If this trend holds true for the future then 'warbots' are going to be increasingly used.
The danger with robots would certainly be greater but it is not a new threat.