Encryption is widely used to enforce usage rules for digital content. In many scenarios content is encrypted using a group key which is known to a group of users that are allowed to use the content. When users leave or join the group the group key must be changed. The LKH (Logical Key Hierarchy) algorithm is a very common method of managing these key changes. In this algorithm every user keeps a personal key composed of log n keys (for a group of n users). A key update message consists of O(log n) keys. A major drawback of the LKH algorithm is that users must update their state whenever users join or leave the group. When such an event happens a key update message is sent to all users. A user who is offline during t key updates, and which needs to learn the keys sent in these updates as well as update its personal key, should receive and process the t key update messages, of total length O(t log n) keys. In this paper we show how to reduce this overhead to a message of O(log t) keys. ...