Network anomaly detection based on customers light level signal

1. A method for detecting an anomaly in a Passive Optical Network (PON) using light level detection, the method comprising:

generating an optical profile for a PON based on one or more characteristics of network light level signals collected over a first time period and delivered over the PON, the PON including an optical line terminal (OLT) optically connected to a plurality of optical network terminals (ONTs) via a plurality of optical fibers;

generating an optical profile for a customer based on one or more characteristics of customer light level signals collected over the first time period and delivered over the PON to or from an ONT for the customer, wherein the one or more characteristics of the customer or network light level signals includes at least one of: a mean light level signal indicating a moving average over the first time period, a standard deviation in the light level signal, or a slope of the light level signal;

obtaining one or more customer light level signals collected over a second time period and delivered over the PON to or from the ONT for the customer;

detecting an anomaly in the PON for the customer in response to determining that the one or more customer light level signals collected over the second time period deviate from the optical profile for the customer or the optical profile for the PON by more than a threshold amount; and

providing an indication of the anomaly to at least one of a user interface or another computing device.

2. The method of claim 1 , further comprising:

transmitting a control signal to automatically reboot a component of the PON in response to detecting the anomaly.

3. The method of claim 1 , further comprising:

performing a root cause analysis to detect a cause of the anomaly based on locations of customers with detected anomalies during the second time period.

4. The method of claim 1 , further comprising:

detecting, via an optical detector, the customer light level signals or the network light level signals at an upstream or downstream endpoint of at least one of the plurality of optical fibers.

5. The method of claim 1 , wherein determining that the one or more customer light level signals deviate from the optical profile for the customer or the optical profile for the PON by more than the threshold amount includes:

determining a customer light level over the second time period;

determining a Z-score for the customer light level by comparing the customer light level to the mean light level signal and the standard deviation in the light level signal for the network light level signals collected over the first time period; and

determining that an absolute value of the Z-score exceeds a threshold Z-score.

6. The method of claim 5 , wherein the Z-score is a population Z-score, the threshold Z-score is a first threshold Z-score, and further comprising:

determining a customer-specific Z-score for the customer light level by comparing the customer light level to the mean light level signal and the standard deviation in the light level signal for the customer light level signals collected over the first time period;

determining that the customer-specific Z-score exceeds a second threshold Z-score; and

determining that the one or more customer light level signals deviate from the optical profile for the customer and the optical profile for the PON by more than the threshold amount in response to determining that an absolute value of the population Z-score exceeds the first threshold Z-score and an absolute value of the customer specific Z-score exceeds the second threshold Z-score.

7. The method of claim 1 , wherein determining that the one or more customer light level signals deviate from the optical profile for the customer or the optical profile for the PON by more than the threshold amount includes:

detecting an abnormal trend in the one or more customer light level signals collected over the second time period.

8. The method of claim 7 , wherein detecting the abnormal trend includes:

generating a trend component of customer light level signals collected over a third time period by filtering out frequency components which exceed a threshold frequency;

generating a seasonality component of the customer light level signals collected over the third time period by computing a Fourier transform of the trend component;

generating a residual component of the customer light level signals collected over the third time period by subtracting the trend component and the seasonality component from the customer light level signals collected over the third time period;

comparing a slope of the trend component and the residual component to a slope of a trend component and a residual component for the customer light level signals collected over the first time period;

comparing the slope of the trend component and the residual component to a slope of a trend component and a residual component for the network light level signals collected over the first time period;

detecting the abnormal trend in response to determining that the slope of the trend component and the residual component for the customer light level signals collected over the third time period deviate from the slope of the trend component and the residual component for the customer light level signals and the network light level signals collected over the first time period by more than a threshold amount.

9. A computing device for detecting an anomaly in a Passive Optical Network (PON) using light level detection, the computing device comprising:

one or more processors; and

a non-transitory computer-readable memory storing instructions thereon that, when executed by the one or more processors, cause the computing device to:

generate an optical profile for a PON based on one or more characteristics of network light level signals collected over a first time period and delivered over the PON, the PON including an optical line terminal (OLT) optically connected to a plurality of optical network terminals (ONTs) via a plurality of optical fibers;

generate an optical profile for a customer based on one or more characteristics of customer light level signals collected over the first time period and delivered over the PON to or from an ONT for the customer;

obtain one or more customer light level signals collected over a second time period and delivered over the PON to or from the ONT for the customer;

detect an anomaly in the PON for the customer in response to determining that the one or more customer light level signals collected over the second time period deviate from the optical profile for the customer or the optical profile for the PON by more than a threshold amount by detecting an abnormal trend in the one or more customer light level signals collected over the second time period; and

provide an indication of the anomaly to at least one of a user interface or another computing device.

10. The computing device of claim 9 , wherein the instructions further cause the computing device to:

transmit a control signal to automatically reboot a component of the PON in response to detecting the anomaly.

11. The computing device of claim 9 , wherein the instructions further cause the computing device to:

perform a root cause analysis to detect a cause of the anomaly based on locations of customers with detected anomalies during the second time period.

12. The computing device of claim 9 , wherein the instructions further cause the computing device to:

detect, via an optical detector, the customer light level signals or the network light level signals at an upstream or downstream endpoint of at least one of the plurality of optical fibers.

13. The computing device of claim 9 , wherein the one or more characteristics of the customer or network light level signals includes at least one of: a mean light level signal, a standard deviation in the light level signal, or a slope of the light level signal.

14. The computing device of claim 13 , wherein the mean light level signal is a moving average over the first time period.

15. The computing device of claim 14 , wherein to determine that the one or more customer light level signals deviate from the optical profile for the customer or the optical profile for the PON by more than the threshold amount, the instructions cause the computing device to:

determine a customer light level over the second time period;

determine a Z-score for the customer light level by comparing the customer light level to the mean light level signal and the standard deviation in the light level signal for the network light level signals collected over the first time period; and

determine that an absolute value of the Z-score exceeds a threshold Z-score.

16. The computing device of claim 15 , wherein the Z-score is a population Z-score, the threshold Z-score is a first threshold Z-score, and the instructions further cause the computing device to:

determine a customer-specific Z-score for the customer light level by comparing the customer light level to the mean light level signal and the standard deviation in the light level signal for the customer light level signals collected over the first time period;

determine that the customer-specific Z-score exceeds a second threshold Z-score; and

determine that the one or more customer light level signals deviate from the optical profile for the customer or the optical profile for the PON by more than the threshold amount in response to determining that an absolute value of the population Z-score exceeds the first threshold Z-score and an absolute value of the customer specific Z-score exceeds the second threshold Z-score.

17. The computing device of claim 9 , wherein to detect the abnormal trend, the instructions cause the computing device to:

generate a trend component of customer light level signals collected over a third time period by filtering out frequency components which exceed a threshold frequency;

generate a seasonality component of the customer light level signals collected over the third time period by computing a Fourier transform of the trend component;

generate a residual component of the customer light level signals collected over the third time period by subtracting the trend component and the seasonality component from the customer light level signals collected over the third time period;

compare a slope of the trend component and the residual component to a slope of a trend component and a residual component for the customer light level signals collected over the first time period;

compare the slope of the trend component and the residual component to a slope of a trend component and a residual component for the network light level signals collected over the first time period; and

detect the abnormal trend in response to determining that the slope of the trend component and the residual component for the customer light level signals collected over the third time period deviate from the slope of the trend component and the residual component for the customer light level signals and the network light level signals collected over the first time period by more than a threshold amount.