1. Clock Synchronization in Sensor Networks for Civil Security Farnaz Moradi Asrin Javaheri

2. Outline  Wireless sensor networks  Surveillance Applications  Clock synchronization  Synchronization algorithms  RBS  FTSP  Implementation  Performance Evaluation  Fault tolerance  Security  Demonstration  Conclusion

3. Wireless Sensor Networks  Consist a set of small sensor devices  Sensor nodes are deployed in an ad hoc fashion  Used for sensing a physical phenomenon  Communication using wireless radio channels  Nodes are constrained in memory, computational power, battery lifetime, …

4. Surveillance Applications  Transport Applications:  Airports, Harbours, Borders, Railways and Roadways  Public Places:  Banks, Supermarkets, Homes, Parking Lots, Hospitals, Bridges  Law Enforcement and Military:  Forensic Applications, Intrusion detection, Target tracking, Remote Surveillance  Environmental Monitoring:  Habitat Monitoring, Forrest Fire Monitoring

5. Surveillance Scenario

6. Clock and Time  System clock counts oscillations of a quartz crystal  Clock offset: difference between the time reported by a clock and the real time  Nodes started at different times  Clock Skew: difference in the frequencies of the clock and the perfect clock  Different frequency of the oscillators  Frequency change of the clocks over time due to temperature, aging, …

7. Why do we need Time Synchronization? “A man with a watch knows what time it is. A man with two watches is never sure.” -Segal’s Law  Time synchronization is a basic middleware service of wireless sensor networks  Time synchronization is required for:  Events with timestamps: Mobile object tracking  Ordering of collected sensor data/events  Delay measurements for distance/location estimation  TDMA radio scheduling  Detection of duplicate events  Coordination of wake-up and sleep times ( for energy efficiency )

8. Errors in Clock Synchronization 0 1 t=2 3 4 5 6 0 1 2 3 t=4 5 6 0 1 2 3 4 5 t=6 6 7 t=8 9 10 11 3 4 5 6 t=7 8 9 0 1 2 3 4 5 t=6 Result

9. Clock Synchronization Algorithms  Design approaches:  Leader-based clock synchronization  Reference broadcast clock synchronization  Averaging-based clock synchronization  Converge to max clock synchronization  …  Protocol classifications:  Continuous  On-demand (post-facto synchronization)  Trade-off between energy efficiency and fine-grained synchronization

10. Flooding Time Synchronization Protocol (FTSP)  Leader-based synchronization (sender-receiver)  MAC Layer time-stamping  Fine-grained time synchronization Receiver iReceiver jSender Synchronization Packet T—Timestamp Packet t2— Timestamp arrival t1—Packet arrival Calculate Global Time t2— Timestamp arrival t1—Packet arrival

11. Reference Broadcast Synchronization (RBS)  Receiver – Receiver based approach  Fine-grained time synchronization  Can be implemented in a distributed manner Receiver iReceiver jSender Beacon Packet t2—Packet reception interrupt t1—Packet arrival t3—Timestamp packet t2—Packet reception interrupt t1—Packet arrival t3—Timestamp packet Asynchronous Exchange of Timestamps

12. Implementation  MSB-430  Texas Instrument MSP430 microcontroller  Chipcon CC1020 transceiver  32.768 kHz quartz clock  4 nodes used for experiments in a single-hop network  Contiki Operating System  Hybrid model of even-driven systems and preemptive multi- threading systems (using proto-threads)  Rime communication stack  Designed for low-power radios wireless sensor networks

13. Experiments FTSP  Synchronizer node periodically broadcasts a message containing the global time  Other nodes timestamp message’s arrival  Nodes calculate their offset and skew with the synchronizer using least square linear regression  An external node periodically broadcasts a query message  All nodes report their estimated global time

14. Experiments RBS  Each node periodically broadcasts a beacon  Other nodes timestamp the beacon’s arrival  Receiver nodes exchange the stored beacon arrival time with each other  Each node calculates the offset and skew with every neighbor using linear regression  An external node broadcasts a query message  All nodes report their estimated global time

15. Performance Evaluation

16. Fault Tolerance  Nodes gets lost (Unattained environments)  Harsh environments  Batteries run out  Adversarial attacks (Sensor nodes can be physically captured or destroyed)  Important requirement for time synchronization services:  Robustness to node and communication failures

17. FTSP  No message collisions (in single-hop network)  Single point of failure  Synchronizer node  Distributed Leader Election  Designating a single node as the synchronizer  After start-up  In case of leader failure  Node with smallest ID will be selected as the leader  Guarantees that there will be only one leader in the network at any time

18. RBS  No single point of failure (completely distributed)  Message collision  Exchange messages (42~70%)  Collision Avoidance  Random back-off (19~38%)  TDMA-based scheduling (0~0.004% )  Nodes transmit one after the other using their dedicated time slot 123456789… BeaconSlot 1Slot 2Slot 3BeaconSlot 1Slot 2Slot 3Beacon… Round 1Round 2 …

19. Comparison Protocol Node Failure Tolerance Message Collision Overall Complexity Basic RBS (with one beacon sender) LowHighMedium RBS (with all nodes sending beacons) High RBS with random back-offHighMediumHigh RBS with TDMA using local timeHighLowHigh RBS with TDMA using global timeHighVery LowHigh Basic FTSPLowNoneLow FTSP with leader electionHighMedium

20. Security Threats to Time Synchronization  Main Goal: convincing nodes that their neighbours' clocks are at a different time than they actually are  Forging/modifying messages  Denial of Service  Pulse-delay attacks  Sybil attacks  Compromising nodes Time = 2 Time = 5 Delay… Replay message Modify message Jam signals Time = 5 I am node 3 Time = 10 I am node 6 Time = 4 ATTACK Time = x

21. Secure Clock Synchronization Algorithm  Secure synchronization protocol  Masks attacks by adversaries  Guarantees automatic recovery after arbitrary failures  Tolerates message collisions and message losses  Self-stabilizing algorithm for secure clock synchronization  Masks pulse-delay attacks in presence of captured nodes  Guarantees efficient communication overheads with high probability

22. Demonstration MSB-430 Core Module PIR Sensor Power Supply

23. Conclusion  Fine-grained clock synchronization is crucial for applications that depend on global notation of time  Surveillance applications: Target tracking, …  The required precision can be achieved by employing distributed approaches  Reference broadcast synchronization (RBS)  Synchronization should be robust and fault tolerance  Leader election to tolerate node failures  TDMA-based scheduling to tolerate message collisions  Attacks against clock synchronization can lead to erroneous application outputs  Secure and self-stabilizing synchronization

24. Future Work  Optimizing the period of sending synchronization messages  Comparing precision against the bandwidth used for synchronization messages and energy consumption  Extending the robust algorithms to larger networks  Testing the secure algorithm in presence of attacks  …

25. Thank You Questions