I am with teckel on the time scale argument. Just protect your project so an average person will take a long time to steal it. They copy you, they get the binary. They don't get the source C++ code. You may update the code to stay ahead of copy cats, offer a serial code on each device and request the number when users need tech support, etc. If you have a truly brilliant idea you should prototype with arduino, lawyer up, show it to a major player in the field, and get them to purchase your idea or give you percentage. Don't try to protect it yourself on an arduino platform. You may not be able to afford it (time, money, effort etc.).
A good example of this is ATmega processors used in ESCs (typically the ATmega8). They're almost always locked so you can't access the program (via conventional means) to keep competitors from copying the code and making their own copycat ESC. If you try to download the code, all you get is all zeros (I've tried). You can write over the programming with your own code even though it's locked. But, you can't "easily" get the code.
ESCs are simple devices with simple programming. So, this level of security is probably good enough to make competitors develop their own code rather than spend the time and money trying to crack into the microcontroller. A scope can tell you what the ESC is doing, so it's quite easy to copy the hardware and reverse engineer the software.
I'd consider the level of security in the ATmega (if done correctly) is above "kid sister" level. Not to a 256 bit encryption level, but I'd say plenty good for almost all purposes. For protecting your idea, it's fine. For a military project, not at all.
Lets put it this way. For even a HIPAA security system, I typically use 256 bit SHA-2 including a 320+ bit private key/salt done over 1,000 iterations to prevent rainbow table attacks as well as slowing down brute-force attempts. To date, there have been no known collisions found (even via theoretical attacks) with 256 bit SHA-2. While brute force is still possible, the time required (as in billions of years) makes it unrealistic to attempt. And this is not even to a level of military-level security. What I like to say is that it will be so secure that other means will be used to get what they want. Like, instead of hacking into the system, it would be easier to break into your office and look for passwords that people have written to a Post-It note in their desk. The same goes for an ATmega project. If your idea is really that good, it's probably easier and more rewarding for someone to break into your house and take your computer and get the source code rather than to only get the compiled code off the chip. How secure is your home's WiFi? I'd say something as simple as that is your weakest link, not the ATmega.