Only if the numbers are fairly comparable. If the number of "good" scientists looking to defend is much larger than the number of bad scientists, you're still probably better off publishing.
You have to guarantee 100% effective vaccines with 100% distribution.
This is false. A vaccine can effectively stop a virus with far fewer than 100% actually receiving the vaccination.
First, depending on the type of vaccine, the immunity may be communicated to other people through contact. Other people in the household, work, school, etc. are contaminated with the attenuated virus (in such a vaccination, as with the Salk polio virus).
Second, with a sufficient portion of the population immune, they act as firebreaks interfering with the transmission of the virus through the population. This is called "Herd Immunity" [1]. According to the linked article, for example, smallpox is effectively stopped with 83%-85% of the community vaccinated.
The concern is that it is much much easier for those who would cause harm to use this information than for those who would use it to create defenses. With such an asymmetry it makes sense to restrict access.
>I've got to believe that anyone skilled enough to be able to use the information
From what I've gathered from various articles is that the process described doesn't require any high level skills or facilities. Apparently it's frighteningly easy to nudge bird flu to be highly contagious in humans.