American Journal of Applied Science and Technology

Hybrid Machine Learning Architecture for SAP Production Planning Optimization and Lean Manufacturing Performance

Huda Al-Shehri — 2026-09-01

Modern production planning increasingly requires intelligent decision-support mechanisms capable of integrating demand variability, production constraints, inventory conditions, and operational performance objectives. SAP-based production planning provides an enterprise-level environment for coordinating these activities, but conventional planning logic may be insufficient when manufacturing systems exhibit nonlinear demand patterns, fluctuating capacities, bottlenecks, and complex interactions among planning variables. This research proposes a Hybrid Machine Learning Architecture for SAP Production Planning Optimization and Lean Manufacturing Performance that combines predictive learning, feature-aware representation, residual learning, attention mechanisms, and optimization-oriented decision support. Because the supplied literature primarily concerns machine learning architectures for medical-image segmentation rather than SAP or manufacturing, these studies are treated as methodological foundations rather than direct evidence of SAP performance. In particular, residual learning, attention mechanisms, recurrent contextual learning, and encoder–decoder architectures provide transferable principles for constructing a hybrid analytical architecture (He et al., 2016; Hu et al., 2018; Cai et al., 2019; Ronneberger et al., 2015). The proposed framework uses SAP planning data to generate forecasts, estimate production risks, identify operational constraints, and recommend planning adjustments. A conceptual evaluation demonstrates how the architecture can be assessed through forecasting accuracy, inventory efficiency, throughput, schedule stability, and lean-performance indicators. The study contributes a structured research framework for integrating machine learning with enterprise production planning while emphasizing interpretability, data quality, scalability, and operational feasibility.

Modeling Security Threats in Enterprise Federated Identity Systems: Attack Vectors, Vulnerabilities, and Mitigation Strategies

Nguyen Minh Anh — 2026-09-01

Enterprise information environments increasingly depend on federated identity architectures to provide centralized authentication and authorization across heterogeneous applications, organizational domains, and cloud services. Although federation and multi-factor authentication (MFA) reduce several weaknesses associated with isolated credential management, they also create concentrated trust relationships and technically complex attack surfaces. This research and review paper develops a structured threat-modeling perspective for enterprise federated identity systems by examining authentication flows, federation trust relationships, token-processing components, identity providers, service providers, and MFA mechanisms as interconnected security assets. A STRIDE-oriented analytical model is employed to classify major threat categories, including spoofing, tampering, repudiation, information disclosure, denial of service, and elevation of privilege. The methodology combines architectural decomposition, attack-surface identification, threat classification, vulnerability assessment, and mitigation mapping. The theoretical foundation is strengthened through the supplied literature on robust estimation, nonsmooth optimization, piecewise-affine modeling, machine-learning behavior, and statistical learning. These works provide useful methodological perspectives for analyzing uncertain, nonlinear, and heterogeneous security environments, although they do not directly investigate federated identity. The analysis indicates that the most consequential risks arise at trust boundaries, token issuance and validation points, MFA recovery processes, administrative interfaces, and identity-provider dependencies. The study further demonstrates that security controls should be evaluated as an interconnected system rather than as independent authentication mechanisms. The resulting framework provides a systematic basis for identifying attack vectors, prioritizing vulnerabilities, and designing layered mitigation strategies for enterprise federated identity infrastructures.

Contextual Risk-Based Verification for Strengthening JWT-Based Authentication

Arben Kola — 2026-09-03

JSON Web Token (JWT)-based authentication provides a scalable mechanism for representing authenticated identity and authorization claims across distributed applications, APIs, and microservices. However, conventional JWT validation generally emphasizes cryptographic validity, issuer, audience, expiration, and token structure while providing limited consideration of the contextual circumstances in which a valid token is presented. This creates an important security limitation: possession of a correctly signed and unexpired token may be treated as sufficient evidence of legitimacy even when the surrounding request context is anomalous. This paper proposes a contextual risk-based verification framework that supplements conventional JWT validation with adaptive assessment of contextual signals, including device characteristics, request origin, temporal behavior, access patterns, session consistency, and resource sensitivity. The methodology conceptualizes authentication as a continuous risk evaluation process rather than a one-time token-validation event. A multi-stage architecture is developed consisting of cryptographic validation, contextual feature extraction, risk scoring, policy evaluation, and adaptive response. The analysis indicates that contextual verification can strengthen JWT-based authentication by distinguishing technically valid tokens from potentially suspicious token usage. The approach also introduces trade-offs involving latency, privacy, false positives, implementation complexity, and policy calibration. The proposed framework therefore positions JWT as one component of a broader adaptive authentication architecture rather than as an independently sufficient security control.