TL;DR
This paper presents a comprehensive systematic mapping and literature review of context modeling for adaptive authentication systems (CM4AA). It identifies key contextual features, modeling approaches, and desired properties, highlighting a significant lack of standardization and a strong focus on authentication-driven research rather than modeling itself.
What Problem Does This Paper Address?
Adaptive authentication systems aim to dynamically adjust authentication mechanisms based on context to balance security and usability. However, effectively modeling this context is a significant challenge. Existing research often uses "vague descriptions and grounds" for context, making it difficult to reuse, extend, or standardize these systems. This paper addresses the need for a clearer understanding of how context is modeled for adaptive authentication, what information is crucial, and what properties these models should possess to facilitate better design and implementation.
Methods
The authors employed a two-pronged approach: a Systematic Mapping Study (SMS) to structure the research area and a Systematic Literature Review (SLR) to gather and synthesize evidence. They developed a logical search clause combining terms related to "authentication system," "context modelling," and various expressions of adaptation capability. After an initial search and duplicate filtering, they analyzed 40 relevant publications based on metrics like keywords, contribution types (concepts, methods, tools, studies, reviews), and application domains. For the SLR, they focused on context information, its modeling formalisms, and its use across the authentication system lifecycle (design, deployment, runtime).
Datasets
Not discussed in the paper. The paper focuses on analyzing existing literature rather than training or evaluating models on specific datasets.
Key Insights and Ideas
- Authentication-Centric, Not Model-Centric: The research field is heavily driven by the authentication community, with less emphasis on the modeling aspects themselves. This leads to a lack of standardized context modeling approaches.
- Generic Concepts Dominate: While a common set of concepts and relationships for CM4AA is identifiable, the majority of modeling concepts are generic (75%), indicating a potential for a unified framework, yet no such standard exists.
- Mobile-First, But Don't Forget Non-Mobile: Mobile devices are a primary source of context information due to their sensors, but non-mobile devices are often disregarded, posing a challenge for broader applicability.
- Privacy is a Major Concern, Yet Abstract: Privacy is a frequently mentioned desired property, but its implementation remains abstract, with no clear consensus on what constitutes private data or how to handle it.
- Runtime is Underserved: Most context models are used during the design and deployment phases, with a significant gap in research addressing their use at runtime, which is crucial for actual adaptation.
Key Results
- Publication Trends: A continuous interest in CM4AA has been observed over the last decade, with a peak in 2017.
- Key Contextual Features: Behavior (29%) and biometrics (22%) are the most frequently used contextual features, often analyzed in conjunction with time.
- Modeling Concepts: Generic concepts (75%) are prevalent, followed by authentication-specific (17%) and domain-specific (8%) concepts.
- Modeling Objectives: Formalizing complex relationships mathematically (54%) is the primary objective, followed by visualizing system organization (21%) and representing processes (17%).
- Lifecycle Stage Usage: Context models are primarily used during the design (50%) and deployment (29%) stages, with limited use at runtime (8%).
- Desired Properties: Accuracy (75%), temporality (71%), and dynamicity (58%) are the most frequently cited desired properties of context models.
Limitations and Future Work
The authors acknowledge that their study is a first step and that the research field is still abstract, with a lack of standardization and benchmark solutions. They highlight the need for more research on:
- Runtime Application: Developing context models that are effectively used at runtime for adaptive authentication.
- Standardization: Establishing standards for context modeling in CM4AA to improve reusability and interoperability.
- Data Availability: Addressing the lack of public data, which hinders the development of tools and benchmarking.
- Heterogeneity: Better handling of the heterogeneity of mobile and non-mobile devices and their context acquisition.
- Model-Based Framework: Developing a model-based framework to support practitioners in context modeling for adaptive authentication systems.
- Recommendation Tool: Creating a tool to assist in modeling context information for adaptive authentication.
Summary updated