Intrusion Detection in A Smart Forest-Fire Early Warning Sensory System
Download as PPTX, PDF1 like215 views
The document describes a methodology for an early detection system for forest fires, which utilizes neural networks to identify intrusions and malfunctions. It includes data analysis of node behaviors and comparisons with existing work while detailing the use of specific communication technologies and classifiers in data processing. Future work involves improving the system by adding virtual nodes and enhancing the weight of various classes.
![INTRUSION DETECTION IN A
SMART FOREST-FIRE EARLY
WARNING SENSORY SYSTEM
Group[2]
Ratul ,Ravali,Suraj and Amani](https://image.slidesharecdn.com/wsn-firedetection-180704214218/85/Intrusion-Detection-in-A-Smart-Forest-Fire-Early-Warning-Sensory-System-1-320.jpg)
Intrusion Detection in A Smart Forest-Fire Early Warning Sensory System
- 1. INTRUSION DETECTION IN A SMART FOREST-FIRE EARLY WARNING SENSORY SYSTEM Group[2] Ratul ,Ravali,Suraj and Amani
- 2. OBJECTIVE Finding an approach for early smart detection of forest fires focusing on intrusions and malfunctions using neural networks. 2
- 3. METHODOLOGY Field Experiment Data Analysis Determination of Classes Data Preprocessing Data Processing : Slow motion the data Data Processing : Working on data with WEKA Comparing Results Compare existing work and papers Future Work 3
- 4. THE EXPERIMENT 4 ● Radio communication on 2.4-2.48GHz ISM band, IEEE 802.15.4 compliant. ● 250kbps data rate. ● Range up to 50m indoors and 500m outdoors. ● Atmel ATmega1281 based. ● 128KB program flash, 512KB measurement flash, 8KB RAM
- 6. DATA ANALYSIS : SINGLE NODE BEHAVIOUR 6
- 7. DATA ANALYSIS : NEIGHBOUR NODE BEHAVIOUR 7
- 8. PRIMARY CLASSES Malfunctioning: Malfunctioning for While Malfunctioning for Long Intrusion: Intrusion detection because of neighbourhood nodes 8
- 9. DATA ANALYSIS : BEHAVIOUR OF DEAD NODE 9
- 10. ALL CLASSES Malfunctioning: Malfunctioning for While Malfunctioning for Long Intrusion: Intrusion detection because of neighbourhood nodes Dead Node behave like Working Node. Secondary Classes: Fire-No Fire 10
- 11. DATA PREPROCESSING Data preprocessing include two stages : Stage 1: Filling missing data to track each node if its not working and for how long. Stage 2: Adding neighbors for each node to do comparison 11
- 12. DATA PROCESSING Stage 1: Slow motion the data using a script. Stage 2: Creating the super vector. Stage 3: Feeding the data into WEKA. Stage 4:Using different classifiers in WEKA. Stage 5:Analyze and compare the different results. 12
- 14. COMPARING WITH NEIGHBOR DATA for all node nodei do for all neighbor neij do for specific amount of time tim do if ( diff of reading between neighbor > threshold_value) occourance ++ else continue endif endfor if( occourance > target_number_of_occourance ) number_of_node_with_difference ++ else number_of_node_with_difference = 0 endif endfor if(number_of_node_with_difference == number of node nodei has) nodei is in potential threat endif endfor 14
- 21. INTRUSION IMPLEMENTATION 21 WE IMPLEMENT INTRUSION BY ALTERING NODE NUMBER 13. AND THE RESULT WITH SAME CLASSIFIERS SHOWS AS FOLLOWING IN NEXT SLIDES.
- 29. CONCLUSION This work included spatial & temporal components at each time stamp and weka results for each class. The problem was modeled as event detection task, where the aim was the prediction of the Asynchronous event. 29
- 30. FUTURE WORK Adding virtual nodes to implement intrusion.( Done) Adding weight to the classes. 30