Attacks in Sensor Networks Team Members: Subramanian Madhanagopal Sivasankaran Rahul Poondy Mukundan
Sensor Networks Wireless sensor networks enable wide range of applications in both military and civilian domains Consists small, low-cost, resource limited nodes. Forward data in a multi-hop fashion This lack of infrastructure makes them susceptible to numerous attacks
Typical Attacks ATTACKS ON CONTROL TRAFFIC Wormhole Sybil Attack Used to attack data traffic attacks ATTACKS ON DATA TRAFFIC Blackhole Selective forwarding Artificial delaying of packets
Existing Countermeasures HMAC and digital signatures Intermediate node authentication Hash trees U(Mu) Tesla The drawbacks of these measures are, Highly complex High communication overhead Require infrastructure Not feasible for Sensor networks
DICAS - Framework DICAS is a lightweight framework, which mitigates the earlier mentioned attacks. Achieved by detection and isolation of malicious nodes. DICAS provides the following, Primitives: Neighbor Discovery One-Hop Authentication Modules: Local Monitoring Local Response
System Model and Assumptions Model Attacker can control both external and/or internal nodes A malicious node can perform any of the attack individually or by colluding with other nodes Assumptions Attacker can’t compromise more than an application defined threshold of guards in a certain transmission range in a given amount of time Key management protocol is used to pre distribute pair wise keys for secure communication Static Topology
Primitives Neighbor discovery Every node joining the network find its immediate two hops by secure communication between its neighbors. The communication is carried out using the shared secret keys (Authentication) One Hop Source Authentication Commitment key for neighbor verification along with message authentication Undisclosed Commitment key piggybacked with response for source authentication
Local Monitoring - Detection Guard Node Can monitor a node Neighbor to both communicating nodes Functions Maintains a watch buffer Contains immediate and original Source/Destination pairs Packet ID Packet Information Drop, Delay Detection – Packet header Modification Detection – Entire Payload Malicious Counter (incremented with malicious activity)
Local Response – Isolation of Nodes Node deemed malicious if Malicious counter exceeds threshold value Guard Node (say M) revokes malicious node (say A) from neighbor list M alerts A’s neighbor (say D) D stores A in Alert Buffer Number of messages per isolation = number of neighbors for guard Light weight property
Lightweight Source Routing (LSR) Routing protocol similar to AODV More resilient and secure Appropriate for Sensor Networks Working Route Request Route Reply
Route Request
Route Response
Analysis Collision Probability increases with increase in nodes Detection rate equals zero for number nodes > 24 ADVANTAGE Lightweight Secure Negligible False Alarm Rate DISADVANTAGE Not Feasible for large number of nodes Works only for static topology Requires pairwise keys to be distributed among the nodes (N*N-1 Keys)
Conclusion Can be extended to mobile networks in future Might require Neighbor Discovery throughout the communication
Reference DICAS: Detection, Diagnosis and Isolation of Control Attacks in Sensor Networks, Issa Khalil, Saurabh Bagchi, Cristina Nina-Rotaru, IEEE Conference on Security and Privacy for Emerging Areas in Communication Networks (SecureComm), Athens, Greece from September, 2005 DICAS: Detection, Diagnosis and Isolation of Control Attacks in Sensor Networks, Issa Khalil, Saurabh Bagchi, Cristina Nina-Rotaru, IEEE Conference on Security and Privacy for Emerging Areas in Communication Networks (SecureComm), Athens, Greece from September, 2005