Scribd
Upload
2 views36 pages

M 6 Anomaly Detection in Networks

Module 6 focuses on anomaly detection in networks, covering both static and dynamic networks, and emphasizes the role of Recurrent Neural Networks (RNNs) in temporal and time-series anomaly detection. RNNs are effective for modeling sequential data, allowing for the identification of deviations from normal patterns through techniques such as forecasting and autoencoders. The module also discusses tools and libraries for implementing these methods in practice.

Uploaded by

Raj Shah
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
2 views36 pages

M 6 Anomaly Detection in Networks

Module 6 focuses on anomaly detection in networks, covering both static and dynamic networks, and emphasizes the role of Recurrent Neural Networks (RNNs) in temporal and time-series anomaly detection. RNNs are effective for modeling sequential data, allowing for the identification of deviations from normal patterns through techniques such as forecasting and autoencoders. The module also discusses tools and libraries for implementing these methods in practice.

Uploaded by

Raj Shah
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 36

Module 6: Anomaly

Detection in Networks:

1. Anomaly in Static Networks: Plain and attributed networks


2. relational learning
3. Anomaly in Dynamic Networks: Preliminaries, feature and decomposition-based approaches.
4. Recurrent Neural Networks (RNNs) for Temporal Anomaly Detection
5. Time-Series Anomaly Detection
Anomaly Detection in
Networks
Recurrent Neural Networks (RNNs) for Temporal Anomaly Detection and
Time-Series Anomaly Detection
Recurrent Neural Networks (RNNs)
for Temporal Anomaly Detection
• Recurrent Neural Networks (RNNs) are well-suited for temporal
anomaly detection and time-series anomaly detection due to their
ability to model sequential data and capture temporal dependencies.
• RNNs process input sequences one element at a time, maintaining a
hidden state that captures information about previous elements. This
makes them ideal for time-series tasks where context and temporal
order matter.
Time-Series & Temporal Anomaly
Detection
•Time-Series Anomaly Detection involves identifying data points or patterns
that do not conform to expected behavior in sequential data.
•Temporal Anomaly Detection can be broader and may apply to events occurring over time,
including logs, user behavior, or sensor readings.

Use of RNN
1. Sequence Modeling
RNNs learn patterns across time — for example, they can learn what a "normal" sequence
looks like and identify when something deviates.
2. Forecasting-Based Anomaly Detection
•Train an RNN (or LSTM/GRU) to predict the next value(s) in a time series.
•If the prediction error at a time point exceeds a certain threshold → mark as anomaly.
3. Autoencoder-Based Detection
•Use an RNN Autoencoder: input a sequence, and the RNN encodes and then decodes it.
•The reconstruction error is used as an anomaly score.
•High error → sequence is anomalous.
• TensorFlow/Keras, PyTorch for modelling
• scikit-learn for preprocessing
• Prophet or ARIMA can complement RNNs for comparison
• TSAD libraries like Merlion or River for broader toolkits
• Anomaly:
• A data point or sequence that deviates significantly from the model’s
expectation or learned distribution.
• Two main RNN-based approaches are:
Recurrent Neural Networks (RNNs)
for Temporal Anomaly Detection
• What is Temporal Anomaly Detection?
• Temporal Anomaly Detection refers to identifying unexpected or
abnormal patterns over time in sequential or time-dependent data.
• Examples:
• Unusual spikes in network traffic logs
• Irregular heartbeat patterns in ECG data
• Unexpected machinery vibration patterns
• These anomalies may not be obvious in isolated data points, but
become clear only in the context of temporal patterns.
• Recurrent Neural Networks (RNNs) are designed to process
sequences. They learn temporal dependencies by maintaining a
hidden state that is updated at every time step based on:
• Current input
• Previous hidden state
• This allows them to remember what happened in the past and detect
deviations from normal sequences — making them perfect for
temporal anomaly detection.
Prediction Based Anomaly Detection
RNN Autoencoder Based Anomaly
Detection

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy