PESAN DAN TANGGAPAN :
Disertasi S3: Deteksi Dini Kondisi Anomali pada Jaringan Sensor berbasis Metode Entropi
Oleh : A. A. Waskita
Jumat, 19 Agustus 2016 (06:13 WIB) dari IP 188.8.131.52
This study has successfully develop an early anomaly detection system (EADS) based on the entropy method especially for the safety critical systems to to ensure the operation of the system safely. The safety critical system are generally have defined the boundary of the normal operation condition for every component in- volved since it was designed. The deviation from this boundary should be detected soon in order to conduct the correct anticipation. To observe the system condition element, a number of sensors, either homogeneous or heterogeneous, are deployed. The more critical the system, the more sensors are involved. However, the number of sensors involved should not burdening the EADS. The entropy method can successfully identify the anomaly on a dummy data set, either in full or half mode. Some specific characteristics were also identified. These characteristics are related to the sensitivity of the detection result, sensors and coding scheme fragmentation, and the factor of the anomaly on the edge of the sensor set. The fragmentation scheme can prevent the burdening of the EADS due to the increasing number of the sensors. The anomaly detection was also conducted on a real data with a reference to a previous research based on the elliptical method. For the same data set, the entropy method detects more anomaly condition than the elliptical one. This difference due to the use of normal boundary approach. The entropy method uses numeric and partial approach, while the elliptical method uses visual and collective approach. Moreover, no interaction between sensors is considered in the entropy method yet. Then, the next research can be conducted on exploring the interaction between sensors, either deterministic with the use of exhaustive approach or probabilistic with the use of mutual information.
| revisi terakhir : 19 Agustus 2016