Although I would agree that the NSA can hack anything, there is a big difference between situations where the NSA can hack with their basic kit of software tools and situations that require them to spend some quality time. If you are a high-profile target and they dedicate a team of cryptographers to your individual case, then yes, NSA can decrypt anything. But you don't have to make it easy. Layers of encryption are tedious to decrypt, especially when the inner layers are of an unknown protocol.
As for password security, it really DOES make a difference with WPA. Nothing can help WEP. I know because I did a security study in a laboratory environment, comparing the success or failure of readily available hacking tools against current wireless encryption protocols.
There are three common types of decryption attacks:
1. Mathematical shortcuts -- flaws in the encryption algorithm can shorten the decryption time dramatically
2. Dictionary attacks -- exactly as it sounds, a vast number of words are read from a file and tried in various combinations
3. Brute-force attacks -- every conceivable combination of characters (even unprintable ones) will be tried until something works
Method 1 is many orders of magnitude faster than method 2, which is many orders of magnitude faster than method 3. Depending on how long the password is, you might not live long enough for a brute-force attack to finish.
Attacking a wireless network starts with monitoring encrypted packets. There are some tools that deliberately inject malformed packets to force the wireless network to generate error packets in response and produce raw encrypted data to speed up the attack. But if you monitor any network long enough, you will eventually get enough data to attack it.
WEP is vulnerable to mathematical shortcuts. No matter what password you choose, it won't help. If you capture enough encrypted data, you can get a WEP password cracked in no time. My personal best in the lab is 2 seconds.
I am not aware of any WPA vulnerabilities to mathematical shortcuts. My research was done a few years ago, and I did not find any published exploitable attack vectors at that time. For this reason, conventional hacker tools for WPA networks rely on dictionary attacks. If you are dumb enough to use names or words that match the dictionary file, the attack will finish in a reasonable amount of time. If you use random characters, the only attack that will work is brute-force. This will hardly ever finish in a reasonable amount of time.
As I said in my original post: "At best, WPA security can resist ordinary eavesdroppers -- not government agencies." I would never bet against the NSA, as they would probably obstruct the deployment of any encryption algorithm that they can't decrypt at will. But it is definitely possible to "raise the bar" so that even the NSA would have to dedicate some resources to attacking your network.
IMHO, the gold standard of Wifi encryption is 802.1x. Several users are typically logged on to the network at any given time, and their username/password is part of the encryption. This means a typical network has several devices connected, each with different encryption keys for different streams of data. There is no single password that unlocks everything, so it's not easy to figure out which packets belong to which stream. For a number of technical reasons, I don't see 802.1x as a viable option for home preppers. But in a commercial environment, that's the way to go.
In a WPA environment, strong passwords mean the difference a brute-force attack and a dictionary attack. Not many people will wait for a brute-force attack to finish. But a teenager can download free programs from the Internet and run a dictionary attack in much less time. It is surprisingly easy to get free, pre-packaged, ready-to-use hacker tools on a bootable DVD. My network cannot be easily attacked with such tools. How about yours?