12:16 Introduction | |
For the first time anonymous communication systems have been presented in the seminal work Chauma (Chaum 1981). The point is that anonymity is achieved by sending messages through a series of transmission units, called mix-nodes (mixing nodes). Each mix-node has two main tasks. The first - is to provide bit-indistinguishability (bitwise unlinkability) message, the second - to mix the message flow. To the output of the node striker was unable to identify the monitor reports on its contents, all incoming messages are the same size (short message are complemented with random data) and encrypted. This is "to ensure bit-indistinguishability. Stirring the message flow is used to ensure that winger could not match the login message to the node with the time of its release, and thus to understand what kind of an outgoing message matches the search. Node for some time accumulates incoming messages, mixes them and sends the next in random order. System anonymous communication over the Internet can be divided into two categories:
For example, electronic mail (e-mail) - a system with long delays, because for message delivery can be time consuming. If you want to interact in real time (or close to it) to use the system with low latency. SSH and instant messenger - examples of systems with small delays. Neotslezhivaemost communications systems of both categories is achieved by implementing the ideas Chauma: peredachtonyh chain of nodes between the sender and recipient, and encryption to hide data messages. Each node in the chain knows only its predecessor, from whom he received the message, and his successor, whom he would send a message. Systems with large delays specialize in sending single messages, while systems with small delays in the maintenance compound. This means that in systems with large delays for each message creates a new path, a new message - a new way. In systems with small delays one path is used for some time to send a packet stream. There is one important difference between systems with small delays. To meet the stringent requirements for time of delivery, have to abandon the accumulation phase and stirring messages. Consequently, such systems are more vulnerable to traffic analysis attacks and in particular timing-attacks. Simple timing-attacks can be reduced to the computation of the time it takes the package to pass the network. More complex timing-attacks can include an analysis of the distinctive features of the connection used victim - revealing a pattern trafika1. Timing-attacks use the fact that all nodes in the network introduce various delays. Knowing the time delay can speculate about the connection within the node and the outgoing flows from it. In other words, guess what the output stream corresponds to the tracks incoming stream. Striker may some time watching the connections between nodes, and then comparing the patterns of traffic of all sites, identify sites with similar patterns - perhaps these sites form a chain. Using statistical methods an attacker can obtain information about the sender and recipient of the flow, and even identify the entire flow path. To protect against this attack needs to be done so that the temporal characteristics of all streams were indistinguishable. However, this would require a significant amount of smeshivaniya2 operations and high volume of traffic covering (cover traffic) 3, hence increase the time delay. Determine the right balance between anonymity and delays - not an easy task. The success of the aforementioned attacks need a global observer, able to observe all flows passing through the network. It is believed that the anonymizing network with low latency, such as Tor, can successfully resist the weaker threat model which does not include a global observer. In this weaker model, the attacker can only see part of the bonds. If we talk about large public networks, such as the Internet, then on that assumption, suggesting the absence of a global observer, it can rely on. Recently, Merdoch (Murdoch) and Danezis (Danezis) showed an example of a successful attack on a system with low latency without the use of a global observer. Attack is based on an analysis of the proposed traffic Danezisom in 2004. They attack the anonymity afforded by Tor, an attacker can break that sees only part of the network or possesses a single node Tor. Attack of the work due to the fact that the developers have removed the Tor blending operation, which was in an early version, and now a part of all streams is processed in round robin fashion4. Controlled by an attacker Tor-node creates a connection to other network nodes and thus can indirectly estimate the volume of traffic passing through at any one time. Ratings are based on the difference in delays of flows sent and received on these connections. Based on the fact that
The authors note that their attack applies to all anonymous networks with low latency. This is a loud statement hints at the viability of the threat model is not used in the creation of many anonymous systems with small delays. Our contributionWe checked the attack Merdoch and Danezisa (Murdoch & Danezis 2005) on the other anonymous networks with low latency. Tarzan (Freedman & Morris 2002) and MorphMix (Rennhard & Plattner 2002) do not work like Tor. In particular, they follow the architecture of peer-to-peer, but Tor uses the dedicated server. Still, Tarzan uses some of the operations of blending and covering traffic, which is not in Tor. And MorphMix, intermediate nodes may choose to part ways to stream, unlike Tor, where the client initializes the stream itself specifies the entire way. Our analysis was moving in two directions. Initially, we focused on the delays encountered in the system, to make sure that it really can be used to estimate the volume of traffic passing through the nodes. We then examined the efficiency of attack for different architectures. Results of the study allowed us to identify principles of anonymous networks with low latency are able to withstand such attacks. OutlineIn the second section, we discuss three anonymizing network - Tor, Tarzan and Morphmix 5 - and show how they differ from each other. In the 3 rd section, we consider an attack on the Tor, and described Merdoch Danezisom (Murdoch & Danezis 2005). Note that the authors claim that the attacks are subject to all the anonymous network with low latency. In the fourth section, we refute this claim by Morphmix. In the fifth section, we derive some rules for constructing systems with small delays. And finally, in the second section will conclude. Note translators 1, the traffic pattern In the context of timing-attacks, it is the characteristic form of "burst" of activity on the schedule, "the amount of traffic from time to time, which may enable to distinguish between users or types of activity in the network . 2 Mixing flow There are many options cmeshivaniya flows (Mixing, mixing). For example in the current Tor Mixing in the absence of such a pattern: the X server included n chains and the same happened. Which belongs to which user, as it were unclear, but the volume of traffic spikes, and other circumstantial evidence, we can speculate. Mixing (only as a simple example) can be all of the chain from server X to server Y again to encrypt and pack it into a common channel is encrypted and will be clear: entered n, but rather on a server came out - not visible. From users to servers go solo chain, and between servers - a continuous stream. As the output of the Tor network is still manifest some of the chain, it is not very effective. 3 covers traffic (cover traffic) The generated gear units phony traffic with a special mark by which other nodes can distinguish it from real-world traffic 4 Round robin fashion The easiest for RR range of processes in one package danyh of each stream (assuming that the priority of all flows are equal). Thus, we achieve "equality" of all data streams. 5 Tarzan and Morphmix existed as a concept even at the beginning of the development of Tor. Currently, these networks are not really used. | |
|
| |
| Total comments: 0 | |