If you can't see how things are running (coding) wouldn't that make it harder to attack such code or network?
It might make things harder on scammers, but if they're determined enough that's just a small obstacle. Many malicious wallets aren't even that sophisticated, and in large part they don't need to be. In some cases they only need to
look like the original, and obviously the code is going to differ somewhat for the scam to occur.
On the other hand, of course, is the trust issue: How do we know that a rogue employee doesn't imbed some malicious code into Binance's wallet? Without being open-source, verifiable, and reproduceable by the general public something like that may months before it's caught.
Open source is especially critical in the crypto world, where we are expected to operate without the need to trust anyone.