Source-linked AI summary
Socially-Aware Networking: A Survey
Feng Xia, Li Liu, Jie Li, Jianhua Ma, Athanasios V. Vasilakos
TL;DR
Mobile-network mobility and connectivity are shaped by users’ social relationships, creating a need for human-centric networking solutions. This survey introduces socially-aware networking, synthesizes social properties and protocols, and identifies open challenges; reviewed work generally reports advantages over socially-oblivious routing, while congestion and other practical challenges remain.
Problem
Mobile-device mobility and intermittent connectivity complicate networking, while existing mobile networks face communication overload and limited resources.
Method
The paper surveys SAN by introducing its architecture and reviewing social properties, routing and forwarding, incentive mechanisms, and data dissemination.
Results
Socially-aware routing and forwarding protocols generally outperform socially-oblivious protocols, while SimBetTS improves delivery performance and distributes load across central nodes.
Takeaways & Limitations
Social properties are a powerful source for designing networks, addressing selfishness, and supporting socially-aware data dissemination.
Abstract
from arXiv · showhide
The widespread proliferation of handheld devices enables mobile carriers to be connected at anytime and anywhere. Meanwhile, the mobility patterns of mobile devices strongly depend on the users' movements, which are closely related to their social relationships and behaviors. Consequently, today's mobile networks are becoming increasingly human centric. This leads to the emergence of a new field which we call socially-aware networking (SAN). One of the major features of SAN is that social awareness becomes indispensable information for the design of networking solutions. This emerging paradigm is applicable to various types of networks (e.g. opportunistic networks, mobile social networks, delay tolerant networks, ad hoc networks, etc) where the users have social relationships and interactions. By exploiting social properties of nodes, SAN can provide better networking support to innovative applications and services. In addition, it facilitates the convergence of human society and cyber physical systems. In this paper, for the first time, to the best of our knowledge, we present a survey of this emerging field. Basic concepts of SAN are introduced. We intend to generalize the widely-used social properties in this regard. The state-of-the-art research on SAN is reviewed with focus on three aspects: routing and forwarding, incentive mechanisms and data dissemination. Some important open issues with respect to mobile social sensing and learning, privacy, node selfishness and scalability are discussed.
I. INTRODUCTION
SAN emerges as a human-centric networking paradigm that exploits users’ social properties and mobility to design networking solutions for mobile applications. The survey introduces SAN’s architecture and reviews routing, incentive mechanisms, and data dissemination.
- Motivation: Mobile-device mobility and social relationships motivate networking solutions that account for human behavior and contact patterns.Social properties can help identify mobility regularities and predict contact opportunities.
- SAN Scope: SAN targets wireless ad hoc networks characterized by intermittent disconnection and absent infrastructure.Its scope includes applications such as pocket switched, vehicular ad hoc, and cyber-physical systems.
- Survey Scope: The survey reviews socially-aware routing and forwarding, incentive mechanisms, and data dissemination algorithms.These components support data delivery while addressing node selfishness and mobile-network constraints.
- Architecture: The proposed SAN architecture senses and analyzes personal, social-contact, and environmental data to obtain social awareness.The paper identifies mobile sensing, data mining, and machine learning as relevant technologies.
- Open Direction: SAN is presented as an emerging field whose architecture and terminology remain subject to further development.The paper describes its architecture as a starting point for broader study.
III. SOCIAL PROPERTIES
Social-network analysis provides concepts and representations for describing relationships among people. SAN concentrates on social properties that can inform networking protocols and algorithms.
- Small-World Structure: The small-world phenomenon describes how people can connect through short sequences of social relationships.The paper notes that this phenomenon has been repeatedly reaffirmed by subsequent work.
- Social-Network Foundations: Social-network research examines relationship dynamics and structure, including relationship strength and intensity.Social network analysis studies relationships among people and their patterns and implications.
- Graph Representation: A social graph represents individuals as nodes and their social relationships as links.The paper treats social networks and social graphs as interchangeable representations in this context.
B. Community
Community and centrality describe complementary aspects of social structure relevant to networking: dense local groups and strategically prominent nodes. These properties support relay selection and information dissemination.
- Community: A community is a subunit with dense internal links, whose members generally connect more often with one another than with outsiders.Communities may arise from family, friendship, common location, or common interest.
- Centrality: Centrality and community are important choices in socially-aware forwarding and data dissemination algorithms.Experiments cited by the paper show that selecting appropriate measures affects algorithm performance.
- Centrality Measures: Degree centrality measures a node’s normalized number of one-hop neighbors and identifies highly active nodes.A high-degree node can serve as a conduit for information exchange.
- Centrality Measures: Betweenness centrality measures a node’s participation in shortest paths and helps identify bridges between communities.Nodes with high betweenness can influence information flow between otherwise nonadjacent nodes.
- Centrality Measures: Closeness centrality captures the efficiency of propagation from one node to others through network distances.It can be used to select relay nodes for faster or successful message delivery.
- Empirical Evidence: Analysis of ACM CoNEXT’07 data found centrality to be the primary factor in deciding whether a node is a good next hop.The best performance trade-off was obtained by combining several complementary rules.
D. Similarity
Similarity and tie strength capture social closeness through common contacts, interests, interaction intensity, and related indicators. These properties complement mobility and contact information in socially-aware networking.
- Similarity: Similarity measures how strongly nodes share contacts or interests, commonly through ratios of common links.It reflects clustering and the likelihood of shared social context.
- Tie Strength: Tie strength quantifies the relationship between two nodes using factors such as frequency, intimacy, longevity, reciprocity, and recency.Strong ties are more likely than weak ties to be activated for information flow.
- Mobility Regularity: Human mobility exhibits spatial and temporal regularity, making repeated movement patterns relevant to networking decisions.Spatial regularity concerns recurring locations, while temporal regularity concerns repetition over time.
- Networking Use: Routing in intermittently connected networks uses store-carry-and-forward, making relay selection and forwarding decisions critical.Social properties such as similarity and tie strength are among the contextual information used by socially-aware protocols.
A. Unicast Routing
Unicast socially-aware routing uses community structure, centrality, tie strength, and mobility context to select forwarders, while community-independent methods trade easier deployment for weaker robustness to mobility deviations.
- Community-Based Routing: Community-based routing forwards messages through inter- and intra-community strategies, using communities as the basis for selecting relays near the destination.Nodes are grouped first; forwarding then differs depending on whether relays are outside or inside the destination’s community.
- Community-Based Routing: BUBBLE RAP combines global and local centrality rankings, first moving messages toward the destination community and then within it.Its hierarchical process uses global ranking until a node reaches the destination community, followed by local ranking until delivery or expiration.
- Community-Based Routing: Distributed community detection uses contact graphs and algorithms including SIMPLE, K-CLIQUE, and MODULARITY, whereas LocalCom forms communities from limited local information.LocalCom uses encounter frequency, length, and separation history to construct a similarity-based neighboring graph and adapt intra- and inter-community forwarding.
- Community-Based Routing: Friendship communities use direct or indirect social ties, while gateways and inter-community relationships support forwarding between otherwise separate communities.The survey identifies friendship among communities as a potential way to improve inter-community forwarding efficiency.
- Community-Independent Routing: Community-independent routing avoids community detection by using locally available context, centrality, similarity, tie strength, or mobility history to estimate suitable forwarders.SimBet compares betweenness and destination similarity, while SimBetTS additionally uses frequency, intimacy, and recency-based tie strength.
- Community-Independent Routing: Community-independent methods are simple to implement but can produce large prediction errors when nodes deviate from their regular mobility patterns.The survey identifies accounting for mobility deviation as a critical future issue; Diverse Routing addresses deviations from habitual activities.
B. Multicast Routing
Socially-aware multicast routing exploits community membership, centrality, and cumulative contact probability to deliver data to multiple destinations in disrupted networks.
- Multicast Routing: Multicast routing in frequently disrupted networks delivers information to multiple destinations whose group membership may change over time.Existing DTN strategies include flooding, tree-based, probability-based, and intelligent multicast routing, but these categories do not consider social properties.
- Multicast Routing: Community-based multicast forwards a message to a community member and can replicate it among members when recipients share a community.Community-based unicast protocols can support multicast when recipients belong to the same community.
- Multicast Routing: Gao et al. proposed community- and centrality-based multicast methods for single and multiple data items, using cumulative contact probability to select relays.For single-data multicast, CCP represents the average probability that a node meets a random node within a time constraint, and relay selection is formulated as a unified knapsack problem.
- Multicast Routing: Compared with unicast, multicast chooses relays using cumulative probabilities rather than generally relying on future meeting probability, making its context more complicated.The survey characterizes multicast as more challenging to design despite its broad application scope.
C. Congestion
Socially-aware routing faces a delivery-ratio, delay, and overhead trade-off, with congestion driven by duplication and concentration of traffic on central nodes.
- Congestion: Routing protocols must balance improving delivery ratio against reducing delivery delay and overhead, while limited buffers make excessive message exchange prone to congestion.Congestion can deteriorate network QoS when nodes are overloaded with too many messages.
- Congestion: 94% of duplicate packets reach the destination in the worst case, creating substantial bandwidth, energy, and memory overhead.The survey describes duplication control as a strategy for achieving a more efficient delivery trade-off.
- Congestion: SimBet’s top 10% of nodes carry 54% of forwards and 85% of handovers, concentrating load and increasing local storage congestion.The resulting unfair distribution can increase packet discarding and decrease delivery rate.
- Congestion: Congestion-control methods mitigate central-node overload or limit message exchange through social-aware forwarding and queue control.Fair routing classifies nodes into levels, forwards using stronger social relations, and applies assortative-based queue control.
- Congestion: Experiments and simulations generally report that socially-aware routing and forwarding protocols outperform socially-oblivious protocols.The surveyed protocols commonly use real trace data, and community-based approaches support both unicast and multicast.
- Node Cooperation: These routing approaches commonly assume cooperative nodes, although selfish or malicious behavior can arise from resource constraints or attempts to increase individual benefits.Selfish nodes may conserve power or buffer, whereas malicious nodes may drop or forge packets or jam channels.
- Node Cooperation: Incentive mechanisms address selfishness through reputation, credit, Tit-For-Tat, or trust-based strategies, but intermittently connected networks make existing mechanisms difficult to apply effectively.Trust-based strategies specifically incorporate social selfishness, while traditional categories generally focus on individual selfishness.
A. Reputation-Based Incentive Mechanisms
Reputation-based mechanisms identify and respond to misbehaving nodes while using forwarding behavior, social information, or routing conditions to regulate cooperation and delivery. The surveyed protocols differ in operational simplicity, reliability, security handling, and dependence on social information.
- Reputation mechanisms: Reputation systems reward forwarding, reduce scores for misbehavior, and exclude nodes whose reputation falls below a threshold.This approach uses reputation to prioritize cooperative nodes and identify misbehaving ones.
- Protocol examples: RADON combines forwarding reputation with destination-meeting probability and uses positive feedback to monitor forwarding behavior.It is designed to prevent deliberate dropping and arbitrary forwarding in malign environments.
- Protocol examples: MobiID combines self-maintained reputation tickets with community-based consensus and a social metric derived from forwarding history.It also addresses attacks involving altered forwarding histories and node isolation.
- Protocol examples: IRONMAN uses pre-existing social information to detect and punish selfish nodes, including information obtained from interviews or online social networks.The paper identifies integration between online social and mobile networks as a future research problem.
- Protocol examples: SRed switches between reputation-based and probabilistic routing generation to balance efficiency and security under benign or adverse conditions.Its dynamic window mechanism supports this mode selection while addressing several routing-layer attacks.
- Comparison: The surveyed protocols achieve high delivery efficiency, with RADON, Give2Get, SRed, and Li and Cao’s scheme emphasizing operational simplicity while MobiID and IROMAN emphasize reliability.The reliability advantage is attributed to exploiting social community and group strength.
B. Credit-Based Incentive Mechanisms
Credit-based mechanisms regulate forwarding through virtual currency, while TFT, barter, and trust-based approaches use reciprocity or social relationships to encourage cooperation. The survey highlights management, cheating, infrastructure, and social-awareness challenges across these mechanisms.
- Credit-based mechanisms: Credit-based schemes pay nodes for forwarding and let them spend earned credits to obtain forwarding service.Successful delivery distributes credits among intermediate relays.
- Credit-based mechanisms: Credit-based schemes face difficulties managing nodes and credit distribution, often requiring trusted parties that may be unavailable in delay-tolerant networks.Selfish nodes may also inject or delete relay nodes to obtain more credits.
- Credit-based mechanisms: SMART and Pi use an Offline Security Manager and Virtual Bank for key distribution and credit management, while SMART employs layered coins.MobiCent similarly uses a Trusted Third Party for key storage, verification, and payment services.
- Reciprocity mechanisms: TFT reciprocates service according to prior contributions but can suffer from bootstrapping problems and fake-message injection.Incentive-Aware Routing combines generosity and contrition to address bootstrapping, transient asymmetries, and retaliation.
- Reciprocity mechanisms: Barter mechanisms exchange message volumes between users, using digital signatures and reputation to help prevent fake-message injection.RELICS ranks nodes by transit behavior, accounting for both forwarding participation and message sending.
- Trust-based mechanisms: Traditional credit-based and TFT mechanisms do not directly incorporate social properties, while trust-based strategies use trusted relationships and social ties for relay selection.Trust establishment and social-information propagation remain difficult in infrastructure-less networks.
VI. SOCIALLY-AWARE DATA DISSEMINATION
Socially-aware data dissemination addresses large-scale, asynchronous content delivery by exploiting user communities, interests, and behavioral regularities. The survey organizes approaches around solicitation and caching versus forwarding, with buffer limits and content selection as central concerns.
- Motivation: Social properties can improve multi-point asynchronous dissemination because people form communities and exhibit regular behaviors.The setting includes users who both consume and produce user-generated content.
- Publish/Subscribe: Publishers, subscribers, and brokers support decoupled content dissemination, with brokers providing routing, event matching, and filtering.Decoupling removes static dependencies in dynamic and decentralized environments.
- Approach categories: Solicitation-and-cache approaches select uninterested content for future distribution, whereas forwarding approaches select relays to reach destination communities quickly.The two categories differ in whether decisions are viewed from the node or content perspective.
- Buffer constraints: Limited buffers make it necessary to choose which unsubscribed content to solicit and which cached content to replace when full.Poor choices can increase power consumption and reduce dissemination efficiency.
- Solicitation and caching: PodNet uses feed-channel subscriptions, pairwise encounters, receiver-driven broadcasting, and separate private and public caches.The public cache stores unsubscribed content intended for redistribution, while the private cache stores subscribed channels.
- Solicitation and caching: Caching strategies use encounter history, subscriber preferences, content popularity, metadata exchange, or cooperative decisions based on mobility and subscriptions.Other approaches account for future encounter interests or optimize local buffer utility through a multi-constrained 0-1 knapsack problem.
B. Forwarding Based Approaches
Forwarding-based dissemination selects nodes using social information such as communities, centrality, interests, and mobility regularity. These methods can improve delivery targeting but incur overhead for maintaining social structure and may not generalize across multiple interests.
- Forwarding-based approaches: Forwarding-based approaches select appropriate nodes to disseminate data quickly by predicting destination information and meeting probability.Community and centrality are the two commonly used social properties.
- Community and centrality: Socio-aware overlays dynamically detect communities and select high-closeness-centrality nodes as brokers for multi-point Pub/Sub dissemination.The design uses community membership and centrality to facilitate event dissemination.
- Interest-based forwarding: SocialCast uses an interest-based utility function and predicted co-location and connectivity changes to select message carriers.When a better carrier appears, one copy is moved, maintaining at most r message copies in the network.
- Interest-based forwarding: SocialCast performs well when a community shares one content type, but its behavior in more general settings is unclear.This limits the supported scenario relative to broader multi-interest environments.
- Community-based forwarding: MOPS defines closeness from temporal and spatial encounters and uses thresholded local communities to support content-based services.Its community model is based on neighboring relationships represented by encounter information.
- Mobility and interest: Habit builds regularity and interest graphs to represent familiar-stranger relationships, movement regularity, and interest connections.These graphs support forwarding decisions based on temporal, spatial, and interest information.
- Limitations: Forwarding methods consume resources to detect and maintain dynamic communities and do not adequately address multi-interest scenarios.The survey identifies this as a challenge for forwarding-based approaches.
VII. OPEN ISSUES
SAN faces open issues in sensing and learning, privacy, and the scope and quality of information needed to derive social properties. Existing research also leaves challenges in applying these capabilities effectively and efficiently.
- Mobile Social Sensing and Learning: Mobile social sensing must move beyond time and location data toward collecting users’ words and actions for analyzing social behavior.Existing datasets provide substantial value, but these common data types are insufficient for further social-behavior analysis.
- Mobile Social Sensing and Learning: Different applications and social properties require different data types and scales, complicating mobile social learning.Individual mobility depends on time and location, whereas group-level community and centrality depend on interests or interactions.
- Mobile Social Sensing and Learning: Mobile social sensing and learning require further research to improve effectiveness and efficiency in ways matched to applications and social properties.The survey identifies both sensing and learning as needing continued development rather than treating data collection alone as sufficient.
- Privacy: SAN privacy is difficult because protocols require sharing location, preferences, and social relations, while privacy concerns vary by relationship and granularity.Encryption and access control are existing approaches, but access control becomes difficult when many relationships are involved.
C. Node Selfishness
SAN must address selfish behavior and scalability across routing, data dissemination, and application levels. These challenges become harder as networks grow, because cooperation, route management, memory, and system integration all become more demanding.
- Node Selfishness: Node selfishness includes social and individual forms, which require different responses based on contextual and social information.Existing incentive mechanisms mainly target individual selfishness, while social selfishness is more often used for trusted-node selection than incentives.
- Scalability: Scalability remains an open problem that previous SAN work has seldom considered across the architecture.The survey highlights routing and data dissemination alongside application-level scalability.
- Scalability: Growing networks make route discovery, effective route finding, encounter-information storage, and route selection increasingly difficult.Dynamic topologies, higher memory requirements, and multiple feasible dissemination routes create distinct scaling pressures.
- Scalability: Application-level scalability requires rapidly integrating new devices and extending systems with new features as requirements change.The survey treats quick device integration and adaptability as essential aspects of robust systems.
- Conclusion: The survey identifies social properties as useful for routing, forwarding, and selfishness-related problems, while noting that many SAN challenges remain before practical realization.Its broader conclusion presents SAN as a promising paradigm for next-generation network engineering, not as a completed solution.