Professional Documents
Culture Documents
ASiffer Presentation 2017
ASiffer Presentation 2017
Context
1
Context
General motivations
2
General motivations
2
General motivations
2
General motivations
2
General motivations
⇒ Network security :
a major concern
2
A Solution
3
A Solution
3
A Solution
3
Overview
⊸ Basic scheme
4
Overview
⊸ Basic scheme
4
Overview
⊸ Basic scheme
alerts
data
4
Inherent problem
5
Inherent problem
5
Inherent problem
5
Inherent problem
5
Providing better thresholds
My problem
0.025
0.020
Frequency
0.015
0.010
0.005
0.000
0 25 50 75 100 125 150 175 200
Daily payment by credit card ( )
6
My problem
0.025
0.020
Frequency
0.015
0.010
0.005
0.000
0 25 50 75 100 125 150 175 200
Daily payment by credit card ( )
0.025
0.020
Frequency
0.015
0.010
0.005
0.000
0 25 50 75 100 125 150 175 200
Daily payment by credit card ( )
7
Solution 1: empirical approach
0.025 1−q q
0.020
Frequency
0.015
0.010
zq
0.005
0.000
0 25 50 75 100 125 150 175 200
Daily payment by credit card ( )
7
Solution 1: empirical approach
0.025 1−q q
0.020
Frequency
0.015
0.010
zq
0.005
0.000
0 25 50 75 100 125 150 175 200
Daily payment by credit card ( )
0.025
0.020
Frequency
0.015
0.010
0.005
0.000
0 25 50 75 100 125 150 175 200
Daily payment by credit card ( )
8
Solution 2: Standard Model
0.025
0.020
Frequency
0.015
0.010
0.005
0.000
0 25 50 75 100 125 150 175 200
Daily payment by credit card ( )
8
Solution 2: Standard Model
0.025
0.020
Frequency
0.015
0.010
zq
0.005
0.000
0 25 50 75 100 125 150 175 200
Daily payment by credit card ( )
8
Solution 2: Standard Model
0.025
0.020
Frequency
0.015
0.010
zq
0.005
0.000
0 25 50 75 100 125 150 175 200
Daily payment by credit card ( )
0.015
0.20
0.010 0.15
0.10
0.005
0.05
0.000 0.00
0 100 200 300 16 18 20 22 24
0.020 0.03
0.015
0.02
0.010
0.005 0.01
0.000 0.00
20 40 60 80 100 0 25 50 75
9
Results
10
Inspection of extreme events
0.025
0.020
Frequency
0.015
0.010
Probability estimation ?
0.005
0.000
0 25 50 75 100 125 150 175 200
Daily payment by credit card ( )
11
Inspection of extreme events
0.025
0.020
Frequency
0.015
0.010
Probability estimation ?
0.005
0.000
0 25 50 75 100 125 150 175 200
Daily payment by credit card ( )
11
Extreme Value Theory
12
Extreme Value Theory
12
Extreme Value Theory
0.3
0.2
0.1
0.0
0 1 2 3 4 5 6
x
12
An impressive analogy
∑
n
Sn = Xi Mn = max (Xi )
1≤i≤n
i=1
13
An impressive analogy
∑
n
Sn = Xi Mn = max (Xi )
1≤i≤n
i=1
13
An impressive analogy
∑
n
Sn = Xi Mn = max (Xi )
1≤i≤n
i=1
⊸ FTG Theorem
M n − an d
−→ EVD(γ)
bn
13
A more practical result
14
A more practical result
14
How to use EVT
15
How to use EVT
15
How to use EVT
15
How to use EVT
X1 , X2 . . . Xn EVT zq
15
Finding anomalies in streams
Streaming Peaks-Over-Threshold (SPOT) algorithm
16
Streaming Peaks-Over-Threshold (SPOT) algorithm
q
0.30
(initial batch)
0.20 zq
X1 , X2 . . . Xn Calibration t
0.10
0
20 40 60 80 100 120
16
Streaming Peaks-Over-Threshold (SPOT) algorithm
q
0.30
(initial batch)
0.20 zq
X1 , X2 . . . Xn Calibration t
0.10
0
20 40 60 80 100 120
16
Streaming Peaks-Over-Threshold (SPOT) algorithm
q
0.30
(initial batch)
0.20 zq
X1 , X2 . . . Xn Calibration t
0.10
0
20 40 60 80 100 120
16
Streaming Peaks-Over-Threshold (SPOT) algorithm
q
0.30
(initial batch)
0.20 zq
X1 , X2 . . . Xn Calibration t
0.10
0
20 40 60 80 100 120
16
Streaming Peaks-Over-Threshold (SPOT) algorithm
q
0.30
(initial batch)
0.20 zq
X1 , X2 . . . Xn Calibration t
0.10
0
20 40 60 80 100 120
16
Streaming Peaks-Over-Threshold (SPOT) algorithm
q
0.30
(initial batch)
0.20 zq
X1 , X2 . . . Xn Calibration t
0.10
0
20 40 60 80 100 120
16
Streaming Peaks-Over-Threshold (SPOT) algorithm
q
0.30
(initial batch)
0.20 zq
X1 , X2 . . . Xn Calibration t
0.10
0
20 40 60 80 100 120
16
Streaming Peaks-Over-Threshold (SPOT) algorithm
q
0.30
(initial batch)
0.20 zq
X1 , X2 . . . Xn Calibration t
0.10
0
20 40 60 80 100 120
16
Streaming Peaks-Over-Threshold (SPOT) algorithm
q
0.30
(initial batch)
0.20 zq
X1 , X2 . . . Xn Calibration t
0.10
0
20 40 60 80 100 120
16
Streaming Peaks-Over-Threshold (SPOT) algorithm
q
0.30
(initial batch)
0.20 zq
X1 , X2 . . . Xn Calibration t
0.10
0
20 40 60 80 100 120
16
Streaming Peaks-Over-Threshold (SPOT) algorithm
q
0.30
(initial batch)
0.20 zq
X1 , X2 . . . Xn Calibration t
0.10
0
20 40 60 80 100 120
16
Can we trust that threshold zq ?
17
Can we trust that threshold zq ?
0.04
Relative error
0.03
0.02
0.01
18
About the data
18
About the data
18
About the data
18
An example to detect network syn scan
19
An example to detect network syn scan
0.6
0.4
0.2
0.0
0.0 100.0 200.0 300.0 400.0 500.0 600.0 700.0 800.0
Time (s)
19
An example to detect network syn scan
0.6
0.4
0.2
0.0
0.0 100.0 200.0 300.0 400.0 500.0 600.0 700.0 800.0
Time (s)
20
SPOT results
0.6
0.4
0.2
0.0
0.0 100.0 200.0 300.0 400.0 500.0 600.0 700.0 800.0
Time (s)
20
Do we really flag scan attacks ?
21
Do we really flag scan attacks ?
100.0
8. 10 −5
1. 10 −5 5. 10
−5
80.0 6. 10 −3
60.0
TPr (%)
40.0
9. 10 −6
20.0
8. 10 −6
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5
FPr (%)
21
Do we really flag scan attacks ?
100.0
8. 10 −5
1. 10 −5 5. 10
−5
80.0 6. 10 −3
60.0
TPr (%)
40.0
9. 10 −6
20.0
8. 10 −6
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5
FPr (%)
22
SPOT Specifications
22
Other things ?
⊸ SPOT
• performs dynamic thresholding without distribution assumption
• uses it to detect network anomalies
23
Other things ?
⊸ SPOT
• performs dynamic thresholding without distribution assumption
• uses it to detect network anomalies
⊸ But it could be adapted to
23
Other things ?
⊸ SPOT
• performs dynamic thresholding without distribution assumption
• uses it to detect network anomalies
⊸ But it could be adapted to
• compute upper and lower thresholds
23
Other things ?
⊸ SPOT
• performs dynamic thresholding without distribution assumption
• uses it to detect network anomalies
⊸ But it could be adapted to
• compute upper and lower thresholds
• other fields
23
Other things ?
⊸ SPOT
• performs dynamic thresholding without distribution assumption
• uses it to detect network anomalies
⊸ But it could be adapted to
• compute upper and lower thresholds
• other fields
• drifting contexts (with an additional parameter) → DSPOT
23
A recent example
24
A recent example
24
A recent example
24
A recent example
24
A recent example
24
EDF stock price ( )
9.05
9.10
9.15
9.20
9.25
9.30
9.35
9.40
09
:0 2
EDF stock prices
09
:4 2
10
:3 4
11
:3 2
12
:1 9
13
Time
:3 0
14
:4 3
15
:3 4
16
:2 1
17
:1 4
25
EDF stock price ( )
9.05
9.10
9.15
9.20
9.25
9.30
9.35
9.40
09
:0 2
EDF stock prices
09
:4 2
10
:3 4
11
:3 2
12
:1 9
13
Time
:3 0
14
:4 3
15
:3 4
16
:2 1
17
:1 4
25
Conclusion
26
Conclusion
26
Conclusion
26
Conclusion
26
Conclusion
26
Conclusion
26
Conclusion
26