Data storage technology has witnessed impressive, and still ongoing, advances since the old days of punched cards. Nevertheless, storage devices, such as hard disks or flash drives, are still bound to fail after long periods of usage risking the loss of valuable information.  To overcome this problem and increase the data reliability, multiple storage nodes can be joined together in a network to redundantly store the data, thus forming a distributed data storage system. Applications of such systems are innumerable and include large data centers, online peer-to-peer storage systems and wireless ad-hoc sensor networks.

To guarantee a targeted level of reliability, when the system experiences a node failure, it is repaired by adding a new replacement node that downloads data by connecting to the nodes that are still active in the system. As bandwidth is an expensive resource in many practical scenarios, such as in wireless networks, it is important that the repair process is accomplished with minimum bandwidth without flooding the systemÕs underlying network. In a recent work by Dimakis et al., it was shown, using network coding techniques, that the repair bandwidth can be reduced by slightly increasing the amount of data stored on the nodes and that, in general, there exists  a fundamental tradeoff between the node storage capacity and the bandwidth needed for repairing the system.

  In this talk, we focus on the case when the dynamic distributed storage system is ``hacked'' by a passive intruder that can observe a number of storage nodes. When accessing a node while it is being added to the system to repair it from a failure, the eavesdropper will have access to all its downloaded messages and not only to its stored content, which may compromise the entire stored data. What is then the maximum amount of data that can be stored without leaking any information to the eavesdropper?  We attempt to answer this question by deriving some upper bounds and focusing again on practical systems where bandwidth is a limited resource. Time permitting, we will also address the case where the original source data starts in a decentralized fashion in the system and analyze the value of interaction among the nodes.