Comment 11 for bug 1157454

Revision history for this message
Chuck Thier (cthier) wrote : Re: Swift exposed to hash collision attacks

I just want to clarify, that scenario 2 is not feasible here. In order for a an attacker to fill up a drive, they would have to generate object paths that map to the same partition. The partition is derived from the first n bytes of the hash, where n is the partition power of the cluster. An attacker may try to generate a number of paths that hash to the same first n bytes, but adding the hash suffix will cause those hash to change, and thus the objects will not fall on the same partition.