Reputation based Secure Routing Protocols

Some routing attacks like selfish behaviour, cannot be detected using cryptography and in an open large scale MANET, where any node is allowed to join or leave at any time, trust in a node based on cryptography is difficult to realise in practice. Another approach for detecting malicious and selfish nodes in a MANET is to analyse the behaviour of the nodes and based on that create lists where the trust against other nodes are weighted. In this subsection, a few examples of reputation based secure MANET routing protocols are examined.

Cooperation of Nodes: Fairness in Dynamic Ad hoc NeTworks (CONFIDANT) (Buchegger & Le Boudec, 2002)

The main idea of CONFIDANT is to make non-cooperative nodes unattractive for other nodes to communicate with. A node chooses a route based on trust relationships built up from experienced, observed or reported routing and forwarding behaviour of other nodes. Each node observes the behaviour of all nodes located within the radio range. When a node discovers a misbehaving node, it informs all other nodes in the network by flooding an alarm message. As a result, all nodes in the network can avoid the detected misbehaving node when choosing a route.

CONFIDANT consists of the monitor, reputation system, path manager and trust manager components. The monitor component listens to its neighbours and inspects if they forward a routing packet that has been sent to them and thus detects non-cooperative nodes such as selfish nodes. The monitor can also check whether a forwarded packet is modified according to the routing protocol, if not then a modification attack is suspected. The trust manager is responsible for sending and receiving alarm messages which are sent by nodes suspecting malicious behaviour in a certain node. The reputation system maintains a table with node ratings and the path manager component manages route information according to feedback from the reputation system.

A major weakness of CONFIDANT is that an attacker is able to send false alarm messages, and as a consequence the attacker can claim that a node is misbehaving even if that is not true. No wormhole is detected due to the fact that they do not either drop or falsely modify routing packets during the route discovery procedure. Furthermore, CONFIDANT has no capability for detecting impersonation, replay, rushing or fabrication attacks.

Collaborative Reputation Mechanism to Enforce Node Cooperation in MANETs (CORE) (Michiardi & Molva, 2002)

CORE is similar to CONFIDANT but employs a more complicated reputation exchange mechanism. Reputation is divided into three distinct components; subjective reputation,

indirect and functional. Subjective reputation is created through their own observations, indirect reputation is built based on reports from other nodes, and functional reputation is

based on behaviour monitored during a specific task. All these reputations together are weighted for a combined reputation value. The major difference between CORE and CONFIDANT is that CORE only allows positive reports while CONFIDANT also accepts negative reports. As a result, in CORE it is not possible to decrease the popularity of a certain node by sending false reports.

Friendship-based AODV (FrAODV) (Eissa et al., 2013)

FrAODV is a trust based security extension for AODV where each node maintains a list of friends and a friendship value for each friend. The friendship value can range from 0 to 100 where 100 refers to the highest level of trust. Three features, i.e. packet precision, blacklists and trust value metric are used to assess the level of trust for each node (Samian et al., 2008).

Packet precision means the accuracy of a routing packet forwarded by a neighbouring node,

which can be used for example, to detect malicious modifications of routing packets, while a node will be listed in a blacklist if it does not forward a routing packet it has received and can thus be suspected as a selfish node. Trust value metric means the use of discrete values to define the trust level of a node.

Trusted routes are built by two algorithms, i.e. RvEvaluate and FwEvaluate. The RvEvaluate algorithm builds up a trusted reverse route from the destination and intermediate nodes to the source node. When an intermediate node receives a RREQ its previous and next hop node’s friendship value is evaluated and the RREQ is rejected if either of these values are less than a threshold for friendship (TF) value. If the friendship value is greater than TF the friendship value of the whole reverse route, i.e. the average friendship value of all nodes on the reverse path between the current and source nodes, is calculated. If the friendship value of the reverse route is greater than a possible previous reverse route to the source, the intermediate node updates the previous route with the new route in its routing table. When a destination node receives a RREQ the procedure is the same, except that the friendship value is only evaluated for the previous hop. Similarly, the FwEvaluate algorithm builds up a trusted forward route from the source and from an intermediate node to the destination node upon receiving a RREP.

FrAODV has no measures for detecting routing attacks that do not modify or drop routing packets such as malicious nodes distributing false RERR messages (fabrication), rushing,

replay and wormhole attacks. Node authentication mechanisms are also missing, which means that spoofing is possible.

2.4.3. Secure Routing Protocols based on a Combination of Cryptography and