- sudo apt-get install linux-tools-6.2.0-32-generic\n
- sudo echo -1 > /proc/sys/kernel/perf_event_paranoid
- sudo echo 0 > /proc/sys/kernel/nmi_watchdog
perf stat --post /mnt/B/sem3/mss/project/post.sh -ebranch-misses,cache-misses,LLC-loads,node-stores -p 266791 -I 5000 --interval-count 2 -o /mnt/B/sem3/mss/project/temp.log
perf stat --post /mnt/B/sem3/mss/project/post.sh -ebranch-misses,cache-misses,LLC-loads,node-stores -p 266791 -I 5000 --interval-count 5
sudo echo -1 > /proc/sys/kernel/perf_event_paranoid sudo echo 0 > /proc/sys/kernel/nmi_watchdog
naming style
condition[wi/wo].version[v1,v2..].metric/event[temp1/temp2].log ex. without DDoS --> wo.v1.temp1.log and wo.v1.temp2.log
- main.py: collect metrics
- read.py: make sure no files other than generated from main.py present in o/p folder. output folder path, flag=0(no attack) It does all cleaning, concatening
- gmm.py: single gmm, run from output
- batch_para_gmm.py: batched gmm 100 parameter per gmm model
- online.py: daemon process collect, train (from output/all_df.log) and test (from folder_name/ as it generated temp files)
all_df.log training samples on 11 Nov a1 -> whole night no activity 1000 samples a2 -> while normal usage 1000 samples a3 -> 60 samples of attack
attack bin for i in $(yes | sed 10q); do ./a.out & done