[bradleyland]

No offense to Damian, but be very careful about accepting the assertion that something is secure. I'm not a security expert, but I am part of the way through the Matasano cyrpto challenges, and my #1 take-away has been that what you think is a secure design rarely is when held to a high level of scrutiny.

The approach the author has taken here would appear to suffer from similar weaknesses present in block ciphers (e.g., AES ECB). The context here is not specifically encryption, but recovering data from "wiped" drives shares some aspects with encryption. Specifically, statistical analysis of the drive state.

Repetition of a block of data will allow an attacker to exploit that pattern in developing statistics about the data that was there before.

In other words, writing a repeating string of random bits is no more secure than writing all zeros.