You cannot use the KDD dataset anymore. It is old (outdated) and has significant bias. Any results obtained with this dataset are irrelevant for modern intrusion detection systems. Furthermore, if you attempt to publish a paper including as validation only the KDD dataset, it will very likely be rejected.
Better datasets also depend on your goals. If you want to study false positives, real traffic archives (CAIDA, MAWI) may help you. In any case, I would recommend that any publication validates results on several datasets.