http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1882343
This framework applies to any multi-currency european style payoff, so worst-ofs and dual digitals qualify (they are the most common trade types in that category). This is the best method of pricing as it reprices the cross smile consistent with that pair's vanilla surface, so hedging is straightforward. Some people just run it through a copula MC, but that won't hit all the cross vanillas.
Example of a trade might be something like 3 month EURUSD less than 1.10, USDJPY less than 125 dual digital. You'd buy this if you think USD-based correlation is priced too high, you like both EURUSD lower and USDJPY lower, and therefore the EURJPY cross much lower. Because you sell correlation, you would be able to buy the dual digital well below the product of the individual digital prices. You would have short USD correlation exposure: long EURJPY vega, short EURUSD and USDJPY vega (the long EURJPY for sure, the 2 short legs are each 'most likely' - the sum of the 2 is definitely short, though). Net you would be long vega if you summed the three up. If the trade works out brilliantly and both EURUSD and USDJPY go lower, your correlation exposure would break down and could even flip signs. If this were a worst-of instead of a dual digital, and those were vanilla type strikes instead of digitals, the correlation exposure would have instead grown larger as the trade went into the money. That is the main difference in risk between dual digitals and worst-ofs of sufficiently long (1 month+) maturity, which means if you think the spot moves and break in correlation will be large, you should trade a worst-of; if you think it will be a more local move and break down in correlation, then a dual digital is more appropriate. For this reason, worst-ofs tend to look more expensive on a relative basis to dual digitals, and dual digitals in turn are slightly more common among the hedge fund community. Also for this reason, people tend to trade worst-ofs on 3 currency pairs instead of 2, and tend to buy correlation rather than sell. The idea being that correlation going to 100% happens more often than a correlation break down, but buying correlation doesn't "look cheap" in a dual digital, and if correlations do go to 100% it can mean a rather large move. It doesn't look cheap in a worst-of, either, but if you add on a 3rd currency pair, then it does start looking cheaper.
Hope this helps!
