When choosing the right transistor for this job, first I'll eliminate the PNP transistors. They're a bit more complicated to use in your case. As you said, for a PNP transistor, active high becomes active low, meaning the transistor will switch on when you apply 0V from your Arduino, but it won't switch off when you apply 5V from the Arduino. You'll need to apply 12V to the base of the PNP transistor to switch off (VEB = 0).
Leaving PNP's behind, looking at the NPN's that you have availabe, only the BC547B (Ic = 100mA) couldn't handle the 480mA current that your siren needs. From the remaining 3 transistors, I'd choose the one that can handle the most current, just to be on the safe side. That would be the BC517 darlington, which can handle a maximum of 1.2A, more than enough for your siren.
Only now you'll have to worry about the gain of the BC517. But, because BC517 is a darlington transistor, it has a huge gain (hFE = 30,000), so you can easily switch on the transistor with a very small base current. If you chose to drive the base of the transistor with a 1KOhm resistor, you'll have a 3.6mA base current, which is sufficient for your purposes.
So the winner would be the BC517.
In this case the decision is fairly easy. Try a major supplier's part selector and select the parameters of interest.
As an example I went to Farnell to set the selector for Vceo=1.5kv.
It's not an option.
So you're not looking for a "widespread" part at all. By all means keep looking and if you find a suitable part, buy it regardless of source. (If you have any doubts about the source, consider doing a lifetime buy.)
Or learn from this little experiment and change the design approach.
There are ways of operating devices in series to achieve higher voltage : they are complex and relatively difficult to get right (for obvious reasons...)
There are devices with higher breakdown voltage (even from Farnell); but they are not BJTs.
Or you may find another approach such as restricting the working voltage to 1kv.
Best Answer
If the motor is very small toy motor (say 500 mA max), try a 2N2222a (maximum continuous collector current (Ic) = 800 mA). The transistor will probably need a heat sink. If the motor draws 500 mA, the headroom will be 300 mA.
Thanks @next-hack. I pulled out a Mims book to read more about this topic. On second thought, a MOSFET such as the IRF510PBF is cheaper and probably won't require a heatsink because it has a TO-220AB package.
Thanks @peufeu. After your comment, I looked closer at the IRF510. So, I now have to also say the IRF510 won't work because of the RdsON in this circuit.
If @user161778 were using a '130 size' DC hobby motor, the currents could be 70, 250 and 500 mA (no load, loaded and stalled). But, the current stated is 750 mA. So, it will be interesting to see if the circuit is built with an NPN (such as the 2N2222) or a MOSFET (such as the AO4480).
Great link for additional reading.