- The paper integrates RFID with weight sensors to validate physical presence and block proxy attendance using statistical thresholds.
- It employs a three-layer IoT design using Arduino and Bluetooth to synchronize sensor data and streamline student management.
- Experimental evaluation confirms low error rates and strong regulatory compliance, despite current prototype limitations.
RFID-Based, Non-Biometric Attendance with Proxy Detection via Weight Sensor Integration
Context and Motivation
Attendance management is crucial in educational environments for pedagogical accountability and institutional operations. Traditional attendance methods (oral roll call, signature sheets) incur significant instructional overhead (up to ten minutes in higher education) and introduce integrity vulnerabilities such as proxy (or “buddy”) attendance—where students sign in for absentees. Despite the proliferation of electronic attendance solutions using biometric modalities (fingerprint, face, iris, voice), such approaches are increasingly limited by regulatory frameworks (GDPR, KVKK, FERPA), given the immutability and sensitivity of biometric data. Meanwhile, RFID-based systems, although scalable and low-cost, are fundamentally susceptible to proxy attendance due to unrestricted card transfer.
This paper introduces an IoT attendance system specifically engineered to address both the regulatory constraints on biometrics and the proxy attendance loophole in RFID-only approaches. The system employs a fusion of RFID authentication with weight-based physical presence verification, thereby disambiguating identity and presence without the retention of biometric or sensitive physical data (2604.22697).
System Architecture and Methodology
The architecture adopts a three-layer IoT paradigm:
- Sensing Layer: Each participation event is mediated by the RC522 RFID module (MIFARE Classic 1K cards, ISO/IEC 14443) and four half-bridge load cells (50 kg capacity, HX711 ADC) installed under seating. The RFID card provides putative identity, while seat occupancy and weight measurement cross-validate physical presence.
- Processing Layer: An Arduino UNO orchestrates real-time data acquisition from both RFID and weight sensors, synchronizing identity and seating events.
- Management Layer: Via HC-06 Bluetooth, raw data streams to a Python-based desktop GUI implementing student management, course assignment, and attendance CSV reporting. Notably, age and gender metadata are prepopulated from standard institutional records; instantaneous weights are never persistently stored.
Proxy detection is realized through a dual-stage algorithm. After confirming a valid RFID, the system compares the sensor value against an empirical threshold statistically modeled from three merged Kaggle datasets (total n=350, ages 18–22). Gender- and age-adaptive mean weights are used as non-biometric priors to distinguish legitimate seat occupancy, thus structurally blocking proxying by individuals with substantially divergent body mass. No biometric database is maintained, and verification hinges solely on cross-session statistical matching.
Experimental Evaluation
Qualitative tests were conducted in near-classroom conditions, focusing on integration correctness, user feedback, and technical stability (RFID reading, weight conversion accuracy, Bluetooth transmission, GUI functionality). Key verified claims include:
- Integrated functionality: RFID, weight sensors, Bluetooth connectivity, and GUI management all operated as intended.
- Low error rate: Scenarios including invalid cards, out-of-slot attempts, and unauthorized course matches reliably triggered correct rejections or warning messages.
- Repeatability and modularity: Despite the current prototype’s limitation to single-seat hardware, the architectural design anticipates scalable extension to multiple seats and aggregate “class weight pool” validation.
- Regulatory compliance: No biometric or persistent physical characteristic data is collected, minimizing GDPR/KVKK/FERPA compliance overhead.
Quantitative performance metrics (e.g., false accept/reject rates, latency) were not systematically reported in this prototype-phase work; such analysis remains a critical open area.
Comparative Analysis and Limitations
The proposed approach clearly delineates itself from biometric-based systems, which, despite high accuracy (face: 82–95%, fingerprint: 95%+ [Koçak & Yeleç 2017; Soewito et al. 2016]), embed inherent privacy and legal liabilities and require costly sensory infrastructure or mobile device management. Pure RFID and wireless-based (BLE, Wi-Fi, NFC) systems lack native proxy attendance defenses, while hybrid RFID+biometric schemes continue to inherit data protection burdens.
Primary limitations:
- The system is currently restricted to a single sensorized chair; multi-seat, real-world deployments have not yet been implemented or validated.
- The demographic reference model leverages synthetic elements (BMI-derived weights, fixed heights) and may not be fully robust to global demographic variance or outlier students (e.g., extreme body mass).
- Proxy prevention remains fallible if an individual physically similar to the target participant proxies attendance.
- Systematic validation under environmental perturbations (vibration, temperature, EM noise) was not addressed, nor was the system’s long-term operational reliability.
- Institutional integration (e.g., automatic fetching of age/gender) was not part of the current prototype.
Practical and Theoretical Implications
This work introduces a procedurally privacy-preserving attendance solution that advances the state of the art in non-biometric identification and proxy attendance detection for classroom settings. By removing biometric enrollment and storage, administrative onboarding is streamlined and legal exposure is reduced. The use of physical sensors for presence detection, while less granular than biometrics, offers an effective deterrent against card transfer and dual seating. The statistical reference model represents a pragmatic compromise between per-person calibration (high privacy cost) and generic sensor thresholds (prone to error), although edge cases require further study.
Theoretically, this multi-modal fusion approach outlines a new axis in privacy-preserving authentication: leveraging population statistics and soft physical correlates to enhance identity-verification integrity while remaining regulatory-compliant. The architecture is well-aligned with principles of data minimization and edge-computing privacy.
Future Directions
Of highest priority is the scaling and validation of the multi-seat “weight pool” implementation, enabling not just individual but aggregate class-level anomaly detection. Systematic benchmarking in operational educational environments (including robustness to environmental factors and adversarial behavior) is needed. Seamless interoperability with institutional student information systems (SIS) and the integration of advanced anomaly-detection algorithms or ML-driven outlier filtering could further refine authenticity checks, especially as sensor density increases. Application extensions include mobile notifications, adaptive access control, and incorporation into broader campus security workflows.
Conclusion
The paper presents a technically and procedurally innovative solution for classroom attendance verification by fusing low-cost RFID-based authentication with non-biometric, statistically-driven physical presence validation. The approach delivers meaningful enhancements over both biometric and RFID-only systems in terms of privacy, security, and compliance, albeit with current practical constraints in prototype scale, population representativeness, and validation scope. Future work on scaling, robustness, and data-driven refinement will be crucial for institutional adoption and broader impact in privacy-conscious educational technology deployment (2604.22697).