1. An Evaluation of Routing Reliability in Non-Collaborative Opportunistic Networks Ling-Jyh Chen, Che-Liang Chiou, and Yi-Chao Chen Institute of Information Science, Academia Sinica {cclljj, clchiou, yichao}@iis.sinica.edu.tw

2. Motivation 1/2 Opportunistic Networks: –Network contacts are intermittent –There is rarely an e2e path between the source and the destination –Disconnection and reconnection are common –Link performance is highly variable or extreme Potential Applications –Interconnect mobile search and rescue nodes in disaster areas –Allow message exchange in underdeveloped areas –Permit scientific monitoring of wilderness areas 2

3. Motivation 2/2 An implicit assumption is usually made in opportunistic networks: all participating peers are collaborative. Many schemes proposed for data dissemination are based on the assumption. However, there may be uncooperative or malicious peers in the network, and these schemes may be vulnerable. 3

4. Our Contribution We identify five types of non-cooperative behaviors: Free Rider, Black Hole, Supernova, Hypernova, and Wormhole We evaluate the impacts of non-cooperative behaviors on data transmission performance of three popular opportunistic network routing schemes. 4

5. Type 1: Free Rider A type of selfish behavior Use the network to forward data, but refuse to serve as a relay for others Effects: –Free riders require less memory and energy than others –Data transmission performance of the system degrades due to the reduced level of collaboration. 5

6. Type 2: Black Hole Drop all relayed data without forwarding to other peers Dropping may be: –Intentional –Due to a lack of capability, e.g. limited battery power or buffer size Black holes cause data loss and may significantly degrade the transmission performance 6

7. Type 3: Supernova A type of malicious attack that propagates random messages destined to other network peers Similar to –Email spamming –Network worms –Denial of service attacks The malicious traffic –consume network resources –interfere with the transmission of regular messages 7

8. Type 4: Hypernova A type of malicious behavior that propagates random messages intended for virtual peers that may or may not exist The network keeps random messages until –destination nodes are found or –they are dropped due to buffer overflow Random messages initiated by hypernova peers may exist longer than those by supernova peers. 8

9. Type 5: Wormhole Composed of one black hole and one white hole –Black holes ‘absorb’ data from others –White hole ‘radiate’ data as much as they can Effects: –likely to be overloaded –single-point-of-failure –security and privacy issues 9

10. Evaluation Settings Evaluate reliability of three opportunistic network routing schemes: –Epidemic –PRoPHET –HEC-BI Simulator: DTNSIM –A Java-based opportunistic simulator 10

11. Evaluation Settings (cont.) Messages: –generated in the first 10% of the simulation time –with a Poisson rate of 1,800 seconds/message –are 1M Bytes Data rate: 2Mbps Buffer size: –1G Bytes for evaluations of free riders and black holes –100 Bytes for evaluations of supernova, hypernova, and wormholes 11

12. Evaluation Scenarios Use two realistic wireless network traces: –iMote: collected from 2005 Infocom conference –UCSD: collected from UCSD campus Trace NameiMoteUCSD DeviceiMotePDA Network TypeBluetoothWiFi Duration(days)377 Devices participating274273 Number of contacts28,217195,364 Avg # Contacts/pair/day0.251480.06834 12

13. Evaluation I: Free Riders - The results indicates that free riders are very harmful to data transmission in opportunistic networks. 13

14. Evaluation II: Black Hole Peers - Similar to free riders, the results indicates that black holes are very harmful to data transmission in opportunistic networks. 14

15. Evaluation III: Supernova Peers - The degradation rates in the supernova scenario are much slower than those in the free rider and black hole scenarios. - The three schemes are more robust against supernova behavior than free rider and black hole behavior. 15

16. Evaluation IV: Hypernova Peers - The effects of supernova and hypernova are similar. - Hypernova, similar to supernova, has less impact on the data transmission performance than free riders and black holes. 16

17. Evaluation V: Wormhole Peers - Surprisingly, the delivery performance does not degrade as the percentage of wormholes increases. - The results indicate that the three schemes are robust against wormholes, and they can even benefit substantially from wormholes when the network connectivity is poor. 17

18. Conclusion We identify five types of non-cooperative behaviors, namely free rider, black hole, supernova, hypernova, and wormhole. We evaluate their impacts on Epidemic, ProPHET, and HEC-BI. Data transmission performance degrades significantly as free rider, black hole, supernova, or hypernova behavior increases. All three routing schemes are robust against wormhole behavior, and can even benefit from it – especially when the network connectivity is poor. 18

19. Thanks!http://www.iis.sinica.edu.tw/~cclljj/http://nrl.iis.sinica.edu.tw/ 19