View
3
Download
0
Category
Preview:
Citation preview
Effect of Social Relations on Cooperative Sensing in
Cognitive Radio Networks
Can Güven, Suzan Bayhan*, Fatih Alagöz
Department of Computer Engineering, Boğaziçi University, Turkey * Helsinki Institute for Information Technology HIIT, Aalto University, Finland
Can Güven – BlackSeaCom 2013 Effect of Social Rela;ons on Coopera;ve Sensing in Cogni;ve Radio Networks
Slide 2 of 22
Agenda
¨ Introduc)on ¤ Coopera)ve Sensing ¤ Social Network Analysis
¨ Key Contribu)ons ¨ Social-‐Aware Coopera)ve Spectrum Sensing (SAC)
¤ System Model ¤ SAC ¤ RAND (Social-‐Oblivious Scheme)
¨ Performance Analysis ¨ Conclusion and Future Direc)ons
Can Güven – BlackSeaCom 2013 Effect of Social Rela;ons on Coopera;ve Sensing in Cogni;ve Radio Networks
Slide 3 of 22
Cooperative Sensing
¨ Coopera)ve vs. Local Sensing ¤ Improvements
n Combined decision accuracy n Spectrum capacity gain
¤ Messaging overhead – Repor)ng delay ¤ Common effec)ng factor: User selec)on
n Independent, non-‐malicious users n Coopera)on investment must have posi)ve returns
n Clever selec)on of coopera)ng CRs
Can Güven – BlackSeaCom 2013 Effect of Social Rela;ons on Coopera;ve Sensing in Cogni;ve Radio Networks
Slide 4 of 22
Social Network Analysis
¨ Social Network Analysis (SNA) ¤ Rela)ons, paOerns between en))es ¤ Social Characteris)cs
n Understanding the Implica)ons
¤ Popularity of online social networks
n Need to study social rela)onships and )es n SNA: Important tool
n Crea)ng a new design approach n New protocols, policies or applica)ons
Can Güven – BlackSeaCom 2013 Effect of Social Rela;ons on Coopera;ve Sensing in Cogni;ve Radio Networks
Slide 5 of 22
Social Network Analysis
¨ Friendship and Community in SNA ¤ Posi)ve Social Characteris)cs
¨ Selfishness in SNA ¤ Nega)ve Social Characteris)cs (ra)onality)
¨ Long term characteris)cs, less vola)le.
Can Güven – BlackSeaCom 2013 Effect of Social Rela;ons on Coopera;ve Sensing in Cogni;ve Radio Networks
Slide 6 of 22
Social Network Analysis
¨ Homophily ¤ Similar interests – befriending, behavior, contact
¨ In our work ¤ To assign friendship )es ¤ Use friendship and community rela)ons
n Select coopera)on set n Evalua)ng if a node should respond
Can Güven – BlackSeaCom 2013 Effect of Social Rela;ons on Coopera;ve Sensing in Cogni;ve Radio Networks
Slide 7 of 22
Key Contributions
¨ New CRN Model ¤ CRs’ opera)on mimic human socie)es ¤ No assump)on of altruis)c coopera)on ¤ Social Aware Scheme
n Clever choice of coopera)on candidates n Exploit mutual sympathy n Select accep)ng and trustworthy CRs (RAND?)
¨ Bringing two emerging research domains together ¤ Social Networks & Cogni)ve Radio Networks
Can Güven – BlackSeaCom 2013 Effect of Social Rela;ons on Coopera;ve Sensing in Cogni;ve Radio Networks
Slide 8 of 22
System Model
¨ CR in the system ¤ C – Community ¤ Lf – Friend list ¤ Ls – Sympathy list ¤ Lc – Coopera)on score list ¤ β – Coopera)on tendency / selfishness
¨ Wireless device with a social context
Can Güven – BlackSeaCom 2013 Effect of Social Rela;ons on Coopera;ve Sensing in Cogni;ve Radio Networks
Slide 9 of 22
System Model
¨ Community Membership ¤ Interests, educa)on etc.
¨ Assigning of communi)es to users ¤ Random
¨ Assignment of friends ¤ Homophily
Can Güven – BlackSeaCom 2013 Effect of Social Rela;ons on Coopera;ve Sensing in Cogni;ve Radio Networks
Slide 10 of 22
System Model
¨ Friend list Lf ¤ Sta)c vs. Dynamic
¨ Dynamic ¤ Sympathy list – changes in community percep)on ¤ Coopera)on score list -‐ Lc
n Memory n Previous and most recent
Can Güven – BlackSeaCom 2013 Effect of Social Rela;ons on Coopera;ve Sensing in Cogni;ve Radio Networks
Slide 11 of 22
System Model
¨ Coopera)on Tendency (Selfishness) – β ¤ Dominant in a CR’s response ¤ Social )e – selfishness (can be altered by social )es)
n Selfishness high – friends only n Average selfishness – same community may be enough
Can Güven – BlackSeaCom 2013 Effect of Social Rela;ons on Coopera;ve Sensing in Cogni;ve Radio Networks
Slide 12 of 22
Social-‐Aware Cooperative Spectrum Sensing (SAC)
¨ Operates in three steps: ¤ STEP 1: Coopera;on set selec;on
n Friend score, community score, coopera)on score n Coopera)on receiving CRs respond (social )es and tendency)
¤ STEP 2: Coopera;ve sensing n Majority Logic from accep)ng CRs
¤ STEP 3: Upda;ng scores n Community and coopera)on scores
Can Güven – BlackSeaCom 2013 Effect of Social Rela;ons on Coopera;ve Sensing in Cogni;ve Radio Networks
Slide 13 of 22
Social-‐Unaware Cooperative Spectrum Sensing (RAND)
¨ RAND operates in two steps ¤ Coopera)on set selec)on:
n Randomly select CRs from the grid n Coopera)on receiving CRs respond according to tendency
¤ Coopera)ve Sensing
¤ No contact memory or social-‐)e based knowledge.
Can Güven – BlackSeaCom 2013 Effect of Social Rela;ons on Coopera;ve Sensing in Cogni;ve Radio Networks
Slide 14 of 22
Malicious Users
¨ Realis)c model ¤ Exploit coopera)ve sensing
¨ Malicious user ¤ Also a CR with social informa)on
¨ Opera)on ¤ Accept ¤ No sensing ¤ Return busy channel result
Can Güven – BlackSeaCom 2013 Effect of Social Rela;ons on Coopera;ve Sensing in Cogni;ve Radio Networks
Slide 15 of 22
Performance Analysis
¨ Simula)on Scenarios ¤ Effect of Coopera)on Tendency β ¤ Effect of Maliciousness dm
Can Güven – BlackSeaCom 2013 Effect of Social Rela;ons on Coopera;ve Sensing in Cogni;ve Radio Networks
Slide 16 of 22
Performance Analysis
¨ β vs. Opportunity discovery ra)o
Can Güven – BlackSeaCom 2013 Effect of Social Rela;ons on Coopera;ve Sensing in Cogni;ve Radio Networks
Slide 17 of 22
Performance Analysis
¨ dm vs. probability of malicious user coopera)on
Can Güven – BlackSeaCom 2013 Effect of Social Rela;ons on Coopera;ve Sensing in Cogni;ve Radio Networks
Slide 18 of 22
Performance Analysis
¨ dm vs. opportunity discovery ra)o
Can Güven – BlackSeaCom 2013 Effect of Social Rela;ons on Coopera;ve Sensing in Cogni;ve Radio Networks
Slide 19 of 22
Conclusion
¨ We have presented a coopera)ve sensing scheme ¨ Makes use of social rela)ons (homophily)
¨ Selec)ng cooperators
¨ Eliminates the altruis)c coopera)ve behavior assump)on ¨ Modeling coopera)on tendency according to social )es
¨ Makes use of previous coopera)ons ¨ Sensing performance memory
¨ Trust mechanism between users ¨ Experience, recommenda)on and knowledge
Can Güven – BlackSeaCom 2013 Effect of Social Rela;ons on Coopera;ve Sensing in Cogni;ve Radio Networks
Slide 20 of 22
Conclusion
¨ Strengths ¨ Dis)nguish malicious and selfish users ¨ Prefer eager-‐to-‐cooperate CRs ¨ Local > coopera)ve sensing if coopera)ve sensing isn’t beneficial
¨ Weakness ¨ Prejudice and sympathy for communi)es
¨ SAC outperforms RAND overall. ¨ Benefits of social-‐aware models over social-‐oblivious models.
Can Güven – BlackSeaCom 2013 Effect of Social Rela;ons on Coopera;ve Sensing in Cogni;ve Radio Networks
Slide 21 of 22
Future Directions
¨ Extending the model to various network models ¨ e.g. Erdős–Rényi (ER), Barabási–Albert model
¨ Currently: Individual preven)on mechanisms ¨ Network-‐wide methods of encouraging coopera)on
¨ And discouraging selfishness/malicious ac)vity ¨ Quicker discovery of malicious nodes
¨ Random mobility –> Homophily
Can Güven – BlackSeaCom 2013 Effect of Social Rela;ons on Coopera;ve Sensing in Cogni;ve Radio Networks
Slide 22 of 22
Thank you.
Any ques)ons?
Contact Info: can.guven@boun.edu.tr
Recommended