Rating: 5.0

The authors actually noticed doing only rotation would be easy to attack, this is why they added matrices A1 and A2. However, as it turns out, A1/|A1| and A2/|A2| are rotation matrices. Then we can describe A1 and A2 as rotation matrices multiplied by some scalar value: A1 = k1 rot(λ1) and A2 = k2 rot(λ2).