A Systematic Review of Anonymous Communication Systems
Ramzi A. Haraty
1
, Maram Assi
1
and Imad Rahal
2
1
Department of Computer Science and Mathematics, Lebanese American University, Beirut, Lebanon
2
Department of Computer Science, College of Saint Benedict & Saint John’s University, Collegeville, MN, U.S.A.
Keywords: Anonymous Systems, Crowds, FreeNet, NetCamo, Mixmaster, Tarzan, TOR.
Abstract: Privacy and anonymity are important concepts in the field of communication. Internet users seek to adopt
protective measures to ensure the privacy and security of the data transmitted over the network. Encryption
is one technique to secure critical information and protect its confidentiality. Although there exist many
encryption algorithms, hiding the identity of the sender can only be achieved through an anonymous
network. Different classifications of anonymous networks exist. Latency level and system model
architecture are two essential criteria. In this paper, we present a description of a set of anonymous systems
including NetCamo, TOR, I2P and many others. We will show how these systems work and contrast the
advantages and disadvantages of each one of them.
1 INTRODUCTION
Due to the increasing use of the Internet and the
emergence of wireless technologies, the value of
security and privacy is becoming more significant.
New online activities have arisen during the last
decade with the advancement of the electronic
communication. People can now shop online, send
and receive emails, pay their mobile bills, and make
diverse banking operations. These types of
electronic activities produced new challenges. Two
main goals the sender of information over the
network seeks to ensure: the privacy and the security
of the communicated information. Confidentiality
and protection of the data can be achieved through
encryption mechanisms. Encryption in general is
capable of hiding the content of the information in
the network. Moreover, in some cases, the sender
might wish to hide his/her identity. This objective
can be achieved through the use of anonymous
systems.
Traffic analysis is the art of examining and
intercepting messages transmitted over the network
to infer information, thus it violates user privacy.
Several technologies exist to ensure data integrity
and the security of the transmitted information that
might be very critical in certain cases. Anonymous
communication protects the identities of the sender
and the receiver from third parties and keeps the
identity of the user hidden from remote parties
(Mittal 2012). Hiding the user-server relationship is
another crucial goal behind any communication. For
example, let us consider a client that wishes to
communicate with a web server. This client might
prefer to stay anonymous. One of the protective
measures that help hiding the identity of users
communicating through the internet is anonymous
network. These networks allow users to surf the
Web without leaving any tracking information.
In (Chaum 1981), Chaum presented almost the
first architecture allowing the transmission of
untraceable email. The main idea behind the
proposed architecture is to allow communicating
peers to transmit data through cascade proxies
known as Onion Routers. Anonymity is achieved by
the use of public key cryptography. Most other
proposed anonymous systems nowadays are based
on Chaum’s scheme. While the main goal behind
anonymous system is to protect the identity of the
sender or the receiver, several other motivations
exist. Some common ones include freedom of
speech, censorships and personal privacy in order to
prevent data mining and tracking. Anonymous
systems can be classified into two main types: high
latency and low latency. In the former category of
networks, the transmitted message takes several
hours or even several days to reach the desired
destination. Quick response is not required for such
application including email systems for example
(Wiangsripanawan, 2007). For interactive and real-
time applications like instant messaging, a low
Haraty, R., Assi, M. and Rahal, I.
A Systematic Review of Anonymous Communication Systems.
DOI: 10.5220/0006216802110220
In Proceedings of the 19th International Conference on Enterprise Information Systems (ICEIS 2017) - Volume 2, pages 211-220
ISBN: 978-989-758-248-6
Copyright © 2017 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
211
latency communication network is required because
of the timing constraint. TOR and I2P are two
examples of low latency anonymous systems that
will be discussed in the next section (Zantout,
2011)(Haraty, 2014). From an architecture point of
view, anonymous systems can be divided into two
categories client-server communication system and
peer-to-peer based anonymous network. In fact, in
the client-server model, only few nodes are selected
to provide anonymity to the rest of the users. One
disadvantage of this architecture is that the number
of server nodes is small, and an attacker can easily
track the traffic. The P2P architecture overcomes
this challenge. The main idea behind this model is
that there is no distinction between a server and a
user (Zhang, 2011). In these systems, it is hard to
distinguish the sender and the receiver nodes. As a
matter of fact, all nodes in the network are
considered universal receivers and universal senders
making it difficult to detect whether a specific node
is transmitting or receiving data.
This paper investigates the network anonymous
systems that seek to protect the identity of the sender
of information transmitted over the Internet and that
provide secrecy. Each upcoming section describes
how a specific system works to achieve anonymity.
A contrast of the advantages and disadvantages of
each technology is illustrated later. Finally, the last
section summarizes the major ideas discussed in this
paper.
2 BACKROUND
Throughout the research that was conducted during
the preparation of this paper, a number of
observations where noted for the design and
implementation of the new methodology. They are
as follows:
1. No Real End-to-End Traffic Analysis
Prevention Assurance:
Although many of the previously mentioned
implementations claimed avoiding traffic analysis,
the possibility for this to occur is extremely high and
unavoidable in unmanaged Local Area Networks
(LAN).
Securing LAN environments could be a costly,
and sometimes an overkill (cost wise), for
organizations of different sizes. Using any of the
implementations in unsecured LAN environments
such as computer labs, work environments, or
wireless networks is somehow a hassle and rarely
found. Therefore man-in-the-middle attacks can
occur at any of these locations or even public
networks whereby a malicious attacker can sniff
packets being transmitted and received by a
particular user or a number of users, and then apply
traffic analysis techniques. One has to note here that
preventing traffic analysis at the end-to-end level is
realistically impossible if infrastructure network
security measurements are not implemented on the
infrastructure level.
2. Trust is in “Cathy”
In any security model example or illustration,
authors tend to use Bob and Alice as two entities
wishing to receive and send information from/to
each other with a trusted entity called Cathy, and a
malicious attacker called Eve. The aim of any traffic
analysis avoidance algorithm considers Eve as an
eavesdropper that will only sniff information.
Hence, the algorithm designed by security personnel
tries as much as possible to circumvent traffic being
passed to Alice and Bob through many and different
routes while camouflaging and encrypting data in
order not to allow Eve to sniff this information.
What is somewhat confusing is that sometimes one
only considers Eve to be on one of the routes that
information is being sent to and from Alice and Bob,
and that Eve is only capable of sniffing abilities and
not injecting information or even tampering with the
data being sent through a route or different routes.
Moreover, in any security model, the adoption of
a trusted entity, Cathy, is a must to verify the
identity of senders and receivers and later to validate
the data being transmitted and received from parties
involved. Cathy happens to be a fixed host that is
susceptible to attacks by Eve also, and any
compromise done to Cathy renders the whole
security model useless sometimes. As a simple
example, if Eve is capable of injecting information
onto a stream whereby Cathy has been compromised
by Eve, the receiving entity will try to validate this
information against Eve and not the trusted entity
Cathy. Data integrity is a vital part of any security
system and having a single point of failure is
ultimately a drawback in any security model. In an
ever growing world of communication and
networks, one has to consider alternatives to basic
security models and concepts. Decentralization of
trusted entities needs to be seriously considered in
anonymous systems hence the reason why I2P was
invented.
3. Questionable Host Reliability and Security
Almost every traffic analysis avoidance design and
implementation relies on hosts that belong to users
for creating different routes and therefore passing
ICEIS 2017 - 19th International Conference on Enterprise Information Systems
212
data through different hops on the network or the
Internet.
What some of the implementations lack, is
catering for the reliability and security of such hosts
mainly because of many factors such as:
a. How trustful are these hosts really? If a host
decides to join a network for anonymous
communication then should that host be trusted
immediately and therefore have data sent to it to
route to other hosts. Consider Eve to be a
distributed form of traffic analysis whereby a
number of malicious hosts join an anonymous
communication model at strategically selected
locations or routes. Data being transmitted on
this network would no longer be anonymous
because Eve can now collect information from
all hosts on the network and perform traffic
analysis on a compilation of streams instead of
one.
Some authors (like Tor developers) argue that
the more the number of hosts joining an
anonymous system and participating in traffic
then more anonymity of traffic being sent and
received becomes possible.
b. Hosts are usually personal computers and
workstations that could be located at users’
homes, labs, and work. These hosts usually have
limited bandwidth allocated to them due to
network lab restrictions or because of
asynchronous bandwidth limitations enforced by
Internet service providers (DSL).
Communication with these hosts could suffer
from factors like intermittent connections, lack
of reliability because of host reboots, signoffs,
power shutdowns, security
vulnerabilities/checks, or even policies enforced
by organizations’ firewall implementations, not
to mention downtime for hosts because of
day/night time making the number of available
nodes much less during non-congestion hours.
Accordingly, and although many
implementations have managed in deploying
their design successfully on the Internet (such as
Tor), they have introduced dedicated reliable
servers worldwide and continue to encourage
users to donate and deploy dedicated servers in
order to make their network reliable to
compensate users’ computer usage behaviors.
However how wise is this?
4. No Dynamic Hops
Some anonymous systems require a number of host
hops or anonymous-router hops for traffic to pass
through, before sending the information to its proper
destination. The reason for this is obviously adding
more anonymity to the transmission of traffic and
also hiding the identity of the sender. However this
also adds more latency and overhead on the
communication stream. The worst case scenario
could be taken geographically whereby regional
traffic may travel to remote hops and then come
back to the receiver that happens to be close to the
sender’s region. Hence, dynamic, geographically
distributed hops need to exist in order to predict
sender and receiver locations and therefore select a
certain number of hops that is ideal for
communication.
3 SYSTEM DESCRIPTION
1. NetCamo which stands for Network
Camouflage is a system designed to provide both
security and efficiency for real time systems while
avoiding traffic analysis (see figure 1). Traffic
analysis avoidance is ensured by two different
requirements: Traffic padding and traffic rerouting.
In the traffic padding, encrypted data is padded with
meaningless data. In other words, in order to
camouflage the packets sent over the network,
additional packets are inserted into the payload. In
the traffic rerouting, unlike the default behavior of
transferring the data from source to destination
through a single path, data is transmitted through
different routes (camouflaged traffic pattern) to
reach its destination.
According to the authors in (Guan, 2001), the
main challenge in NetCamo is to ensure that the
traffic analysis prevention is performed in a realistic
time (suiting the nature of the interactive
applications). This objective becomes hard to
achieve when the network becomes full of padded
traffic. Actually, the communication over the system
passes through three phases: The system
configuration phase where the traffic pattern is
determined, the admission control phase where new
communication streams are accepted or rejected and
the runtime phase where the actual camouflaging of
data is done (Haraty, 2015).
2. TOR, as predicted in figure 2, is a low-latency
anonymous communication system. It is considered
an improved version of the traditional Onion
Routing that includes new integrated features. There
are three main entities participating in the proposed
approach: Tor client which is the sender that wishes
to establish an anonymous communication, the Tor
servers that are the Onion routers responsible for
routing streams to next nodes and the recipient. In
A Systematic Review of Anonymous Communication Systems
213
summary, Tor is made up of a collection of onion
routers where each one sends information in a secure
way to the next hop. Any client can become a server
acting as a Tor onion router. To ensure
camouflaging the type of data being transferred, data
is sent in encrypted format with fixed size packets
called cells which are relayed without revealing their
content or their complete route. This is achieved
through cell encapsulation and multilevel
encryption.
To start communication, the client contacts a Tor
management node which maintains the list of bridge
nodes which accept connections for which a specific
handshake occurs. When the client determines the
participating nodes, it sends a “create” cell to each
of them without allowing any of them to know the
presence of the other.
One of the disadvantages of Tor resulting from
its architecture is that the central directory
containing the list of servers is often a target for the
attackers. Another problem with Tor is that it serves
a large number of users, therefore the use of a
limited number of servers to build an anonymous
path will which leads to performance issues. In
addition, users must keep track of all the available
servers, especially when the number of servers
becomes large as this may cause a deterioration of
performance due to bandwidth contention.
3. I2P, known as Invisible Internet Project is a
low latency anonymizing mix network. Its main
scope is anonymous file-sharing and web hosting
(Timpanaro, 2014). I2P presents some similarities
with tor however, a major difference between the
two can lies in the fact that tor focuses on hiding the
anonymity of the sender, while I2P also hides the
identity of the receiver (Erdin, 2015). It operates on
the network layer. One main characteristic of this
system is that it distinguishes between user online
identity and its geographical location. In fact, the
user is not identified by its IP address and port
number but by another identifier independent of its
location. I2P is message based instead of circuit
based. It is a fully distributed system for anonymous
P2P communications and not browsing which does
not rely on centralized directory servers. Previously
encrypted onion cells are grouped together with
extra padding as well as delay/no-delay instructions
to other I2P nodes and then packaged in so called
garlic cloves which are passed in encrypted format.
The system uses different types of cryptographic
algorithms. I2P is portrayed in figure 3.
4. Crowds is a proposed anonymity system for
web transactions. It allows users to surf the web
anonymously as shown in figure 4. It is named for
the notion of “blending into a crowd” where users
are grouped into a large and geographically diverse
group. Requests are then issued by the group on
behalf of its members, thus web servers will not be
able to know the exact source of a request as it might
have been issued from any of the crowd members.
Even members of the same crowd cannot
differentiate between the originator of a request and
a member that is only forwarding the request. Each
participant in the group is simultaneously protected
and protects other user as well. Therefore each
participant is playing the role of a proxy (Sui, 2003).
5. Based on Chaum’s mix approach, Tarzan is yet
another low-latency anonymous communication
system (see figure 5). Its main goal is to offer
anonymity for different applications including
instant messaging and web applications. It is a
decentralized P2P anonymous network that provides
IP service which makes it general purpose and
transparent to applications. Each node in the Tarzan
model can be both the client and the relay. It
supports layered encryption and multilevel routing
where a client chooses a path of peers in a restricted
way that protects communication against
adversaries. It also uses a protocol to ensure
unbiased selection of peers. Its cover trafc
mechanism offers protection against trafc analysis
of message volume or content, against message
ooding, and against DoS attacks.
Nodes participating in the communication run
software that discovers other participating nodes,
intercepts packets generated by local applications
that should be anonymized, manages tunnels through
chains of other nodes to anonymize these packets,
forwards packets to implement other nodes’ tunnels,
and operates a NAT (network address translator) to
forward other participants’ packets onto the ordinary
Internet. Therefore, the receiver is not necessary
belongs to the Tarzan network.
6. FreeNet is decentralized P2P network
application for storage and retrieval of files in such a
way that protects the anonymity of both senders and
receivers. The system works as a location-
independent distributed file system where many
individual computers that cooperate in routing the
requests.
Nodes participating in the Freenet network
provide their data storage to the network. Each node
maintains its own data store that it makes available
to the network and a routing table which contains the
addresses of other nodes. Requests are passed
through a series of proxys where every node locally
ICEIS 2017 - 19th International Conference on Enterprise Information Systems
214
decides the next node to receive the request. Hence,
each node in the proxy chain is only aware of the
node it received the request from and the node it
forwards it to which makes infeasible to determine if
a particular node is the actual originator of the
requests or a forwarder. This is revealed in figure 6.
In addition, files are dynamically replicated in
locations close to requesters and removed from
locations where no requests for them are made.
Some disadvantages of Freenet system related to
the performance are that it sometimes poorly locates
files and it presents a low speed in downloading
found files (Skogh 2006).
7. Mixmaster is the most widely deployed and
used remailer system. Remailers are servers whose
main goal is to send email without providing any
information about the source. Messages are
encrypted while keeping their size is kept constant
by appending random noise at the end of the
message. This noise is generated using a secret
shared between the remailer and the sender which
makes allows protecting the integrity of the header
and content of the message. Besides protecting
anonymity, Mixmaster allows sending large emails
without the need to use special software. It also
protects against traffic analysis. Mixmaster is
illustrated in figure 7.
4 ANONYMOUS SYSTEMS
COMPARISON
Each of the mentioned anonymous communication
systems has its own characteristics and architecture.
Some follow the client-server architecture, some
other adopt the peer-to-peer design. Different
systems are suitable for different type of
applications. Real-time application requires low-
latency communication systems whereas high-
latency systems can fit to other applications.
Moreover, these anonymous communication models
are susceptible to various attacks (Haraty, 2015).
Table 1 presents the advantages and the
disadvantages of the discussed approaches and help
deciding which model is best proper to each
application.
5 CONCLUSIONS
In this paper, we discussed how protecting the
information shared by users over the Internet, and
the identity of the users themselves are very
important. Adopting anonymous systems is one way
to achieve secrecy of the user’s identity. Each of the
mentioned technologies has its own limitations. In
fact, some are not applicable for real time
applications and some others might not be able to
detect malicious attacks. As each system has its
challenges, we can explain the conflict in each
system between its advantages and its drawbacks.
Besides the various implemented systems hiding the
identity of the users, studies in the field of
anonymous networks continue to grow.
ACKNOWLEDGEMENTS
This work was funded by the Lebanese American
University in Beirut, Lebanon.
REFERENCES
A. A. M Direct. Retrieved on February 11, 2017 from
http://aamdirect.sourceforge.net/.
Atlassian Documentation. Retrieved on February 11, 2017
from https://confluence.atlassian.com/display/
CROWD/Crowd+2.2.2+Release+Notes.
Chaum, D. 1981. Untraceable electronic mail, return
addresses, and digital pseudonyms. Communications
of the ACM, pp. 84-90.
Erdin, E., Zachor, C., and Gunes, M. 2015. How to find
hidden users: A survey of attacks on anonymity
networks. IEEE Communications Surveys & Tutorials
pp. 2296-2316.
Guan, Y., Fu, X., Xuan, D., Shenoy, P., Bettati, R., and
Zhao, W. (n.d.). 2001. NetCamo: Camouflaging
network traffic for QoS-guaranteed mission critical
applications. IEEE Transactions on Systems, Man, and
Cybernetics - Part A, pp. 253-265.
Guide (n.d). Retrieved on February 11, 2017 from https://
trac.i2p2.de/attachment/wiki/Content/i2prouting.png.
Haraty, R. and Zantout, B. 2014. The TOR data
communication system – a survey. In Proceedings of
the Sixth IEEE International Workshop on
Performance Evaluation of Communications in
Distributed Systems and Web based Service
Architectures (PEDISWESA’2014). Madeira, Portugal.
Haraty, R. and Zantout, B. 2014. The TOR data
communication system. Journal of Communications
and Networks, 16 (4) (2014) ISSN 1229-2370.
Haraty, R. and Zantout, B. 2015. A Collaborative-based
Approach to Avoiding Traffic Analysis and Assuring
Data Integrity in Anonymous Systems. Computers in
Human Behavior Journal. Volume 51, Part B, pp.
780–791.
Haraty, R. and Zantout, B. 2015. The NetCamo data
communication system. Advanced Science Letters.
Volume 21, Number 3, pp. 472-477.
A Systematic Review of Anonymous Communication Systems
215
Mittal, P., and Borisov, N. 2012. Information leaks in
structured peer-to-peer anonymous communication
systems. ACM Transactions on Information and
System Security, pp. 1-28.
Skogh, H., Haeggstrom, J., Ghodsi, A., and Ayani, R.
2006. Fast Freenet: Improving Freenet performance by
preferential partition routing and file mesh
propagation. In Proceedings of the Sixth IEEE
International Symposium on Cluster Computing and
the Grid.
Sui, H., Wang, J., Chen, J., and Chen, S. 2003. An
analysis of forwarding mechanism in crowds. In
Proceedings of the IEEE International Conference on
Communications.
Sysmagazine. Anonymous networks and timing attacks:
Tarzan and MorphMix. Retrieved on February 11,
2017 from: http://sysmagazine.com/posts/117586/.
Timpanaro, J., Chrisment, I., and Festor, O. 2014. Group-
based characterization for the I2P anonymous file-
sharing environment. In Proceedings of the 6th
International Conference on New Technologies,
Mobility and Security (NTMS).
Tor - The Onion HTTP Router. Retrieved on February 11,
2017from http://tohr.sourceforge.net/.
Wiangsripanawan, R., Susilo. W., Safavi-Naini, R. (2007)
Design principles for low latency anonymous network
systems secure against timing attacks. In Proceedings
of the Fifth Australasian Symposium on ACSW
Frontiers, Balart, Australia.
Zantout, B. and Haraty R. 2011. I2P Data Communication
System. In Proceedings of the Tenth International
Conference on Networks (ICN 2011), pp. 401-409, St.
Maarten, The Netherlands Antilles.
Zeinalipour-Yazti, D., Kalogeraki, V., and Gunopulos, D.
2003. Information Retrieval in Peer-to-Peer Systems.
University of California at Riverside. Available: http://
www.cs.ucr.edu/~csyiazti/papers/cise2003/cise2003.pdf.
Zhang, J., Duan, H., Liu, W., and Wu, J. 2011.
Anonymity analysis of P2P anonymous
communication systems. Computer
Communications, pp. 358-366.
Figure 1: Architecture of NetCamo (Guan, 2001).
Figure 2: Tor architecture (TOR - The Onion HTTP Router, 2017).
ICEIS 2017 - 19th International Conference on Enterprise Information Systems
216
Table 1: Comparison of the different anonymous systems Advantages.
Disadvantages
BitTorent
Excellent peer-to-peer communication technique for
utilizing peer bandwidth.
In case file transfer is interrupted, only the missing
pieces are re-downloaded and hence saving time and
bandwidth.
Bandwidth chocking and throttling are efficient to
assure bandwidth reliability and availability for both
existing and new peers.
Blocked by most ISPs based on fingerprinting due to the excessive
traffic it generates.
To enhance traffic, a P2P caching appliance was created which still
constitutes a security risk as many portions of downloaded files are
cached for new coming peers, so spoofing techniques may allow
any user to stream existing cached information.
Strictly used for file transfer and does not contain any substantial
security measures.
Cannot be used for real time communication like secure shell,
Telnet, VoIP…
NetCamo
The system is able to distribute traffic through
different predetermined paths and routes in order to
prevent against traffic analysis.
It does not consider geographical distribution of hosts and routes.
Dependent on hosts and routers where a single managing
component takes control of all routers and hosts to determine
acceptance or rejection of a traffic stream which is hard to
implement in the real world.
Dependent on routers to set the rate of traffic and control the
network flow.
Since the system is centralized, attacks on the main node can
render the system useless.
Traffic padding is performed at a rate of 1/α where α can be
determined if kept constant.
Un-trusted hosts can join. They can drop packets or interfere in
malicious ways to make the system unreliable.
End-to-end prevention for data sniffing and traffic analysis is
impossible when it comes attacks occurring at the LAN level
TOR
Protects against strong and weak attackers based on Tor
design as well as encryption techniques.
Protects anonymity of sensitive published content.
Tor nodes are not aware of the complete plan of
communication, so even if one node is acting
maliciously, it can only know little.
The addition of more Tor nodes adds more anonymity.
Contacting nodes gradually adds more security as
building the path is based on a list of bridge nodes.
Directory information server can be blocked.
Blocking based on fingerprinting Tor`s connection: handshakes are
clear to authorities thus any intelligent firewall can detect them and
block them.
Centralization of directory servers for managing the Tor network:
attackers can fake the identity of the Tor directory by redirecting
traffic to a local server, the attacker can maliciously modify a Tor
client and then repackage it for users to download.
Single path for a data stream moving inside a circuit: saving the
traffic for latter analysis may reveal the identity of sender and
receiver.
Malicious attackers and relayers may not be identified.
Slow performance: more load on Tor dedicated nodes which
decreases anonymity, security and reliability.
Successes or failures in data integrity checks may render a circuit
useless.
Website fingerprinting and backtrack attack due to lack of packet
camouflaging.
I2P
Message based instead of circuit based: which leads to
the decrease in overhead and odds randomness while
allowing hops to control data delivery.
Various protocols support: offers a wide range of
internal services, anonymous hidden services...
New P2P infrastructure over the Internet: P2P activities
become anonymous to all participating parties.
Different encryption techniques: good set of algorithms
(symmetric, asymmetric..
Decentralized System: protected against attacks on its
directory serves.
Different types of un-directional tunnels which
enhances the amount of peers participating in
communication and therefore increasing number of
hops.
End user Node Participating in communication:
encourages every node joining I2P to use part of its
bandwidth as relay node which adds more hops and
thus leads to randomness.
Vulnerable to partitioning attacks: may disconnect targets in the
system and reveal identities of all parties involved.
Possible intersection attacks: attacker may eliminate nodes that
have not participated in communication with target until target's
paths are narrowed down which makes nodes participating exposed
for monitoring.
Lack of node and bandwidth monitoring, participating peers
variably change their relay capabilities and random joins and
departures may allow wide distributed attacks resource
consumption attacks without being detected.
NetDB conflicts and resolution
DOS attacks:
1. Greedy user attack: users willing to download more than upload
which decreases traffic replay, less anonymity since the number of
hopes is less.
2. Starvation attack: bad and intermittent communication for end
users.
A Systematic Review of Anonymous Communication Systems
217
Table 1: Comparison of the different anonymous systems Advantages (cont.).
Crowds
It offers the user some degree of deniability for her
observed browsing behavior, if it is possible that she
was using Crowds.
No single failure discontinues all ongoing web
transactions
Privacy is enhanced by increasing the average number
of times a request is forwarded among members before
being submitted to the end server with less impact on
the performance because of the use public private key
operations…
Probability of receiver anonymity increases as size of
crowd.
Sender anonymity against end servers
Protection against timing attacking the crowd
Good performance due to load balancing
A user may be incorrectly suspected of originating a request
It cannot protect user`s anonymity if the content of her web
transaction reveals her identity to the webserver
It can be undermined by executable web content that, if
downloaded into the user`s browser, can open network connections
directly from the browser bypassing Crowds and exposing the user
to the end server.
No effort to defend against DOS attacks by crowd members
No sender anonymity against local eavesdropper and no receiver
anonymity against end servers
Firewalls represent a barrier to wide-scale inter-corporation
adoption of Crowds
Tarzan
Reduces effort needed to incorporate anonymity into
existing designs without the need to change them.
Self-organizing and fully-decentralized.
Scales to much larger networks.
Better protection against static adversaries than Crowds
and Onion Routing
Effective and efficient peer-discovery mechanism
Its cover trafc mechanism offers protection against
trafc analysis of message volume or content, against
message ooding, and against DoS attacks of slowing
incoming rates.
Prevents information leakage at exit points by using
integrity checks
Fast packet forwarding rate, high throughput, and
reasonable tunnel-setup latency.
No sender protection against malicious routers.
It may be susceptible to some attacks that use additional
application-layer information.
No real protection against information that leaks through
application-layer interaction.
A new tunnel is unlikely to traverse the same small path thus long-
term node observation or time-intensive attacks are less effective.
Freenet
It keeps information available while remaining highly
scalable.
Sender`s anonymity is preserved beyond suspicion
against collaboration of malicious nodes.
Key anonymity and stronger sender anonymity are
achieved by pre-routing of messages.
Protection of data source by the occasional resetting of
the data source field.
Since routing depends on knowledge of the key, key anonymity is
not possible in the basic FreeNet scheme.
Against local eavesdropper, there is no protection of messages
between user and the first node contacted.
Requested files can be modified by malicious nodes in addition to
data being vulnerable to dictionary attacks
Subject to DoS attacks.
Mixmaster
Supports sender anonymity.
Allows large emails to be transmitted without the use of
special software and messages to be transmitted
multiple times using different paths.
Protects against email spammers.
Protects content by including a hash of the message in
the header.
No analysis on the impact of its anonymity features has ever been
performed.
No filtering of content.
Figure 3: I2P architecture (Guide, 2017).
ICEIS 2017 - 19th International Conference on Enterprise Information Systems
218
Figure 4: Architecture of Crowds (Altassian, 2017).
Figure 5: Tarzan Architecture based on simulators (Anonymous networks, 2017).
Figure 6: The Freenet architecture (Zeinalipour-Yazti, 2003).
A Systematic Review of Anonymous Communication Systems
219
Figure 7: The Mixmaster remailer model (AAM Direct, 2017).
ICEIS 2017 - 19th International Conference on Enterprise Information Systems
220