CUMAD is a framework developed to detect compromised IoT devices with enhanced security measures. Current IoT devices generally lack sufficient built-in security, leading to frequent false alerts in existing anomaly detection schemes. CUMAD addresses these issues by integrating an autoencoder-based subsystem with a sequential probability ratio test subsystem, shifting the focus from individual anomalous events to a cumulative evidence approach. Evaluation based on the N-BaIoT dataset reveals that CUMAD lowers the false positive rate from 3.57% to 0.5%, and detects compromised devices efficiently, averaging less than 5 observations for detection.
CUMAD integrates an autoencoder-based anomaly detection subsystem with a sequential probability ratio test (SPRT) subsystem, enabling effective detection of compromised IoT devices with reduced false positives.
The evaluation of CUMAD using the N-BaIoT dataset demonstrates a reduction in false positive rate from approximately 3.57% to about 0.5%, enhancing the reliability of IoT security.
By accumulating evidence rather than relying on isolated anomalous events, CUMAD significantly improves the effectiveness and efficiency of anomaly detection in IoT devices.
CUMAD manages to detect compromised IoT devices with an average of less than 5 observations, providing a quick response to security threats.
Collection
[
|
...
]