Scribd
Upload
2 views

Autoencoder IoT Threat Detection Report

This project develops an Autoencoder-based system for anomaly detection in IoT networks using the CICIDS 2018 dataset. The model is trained on normal traffic and evaluates anomalies through reconstruction errors, achieving high accuracy compared to traditional methods. The study also includes data preprocessing, model training, and performance comparison with other models like Isolation Forest and One-Class SVM.

Uploaded by

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

Autoencoder IoT Threat Detection Report

This project develops an Autoencoder-based system for anomaly detection in IoT networks using the CICIDS 2018 dataset. The model is trained on normal traffic and evaluates anomalies through reconstruction errors, achieving high accuracy compared to traditional methods. The study also includes data preprocessing, model training, and performance comparison with other models like Isolation Forest and One-Class SVM.

Uploaded by

Shubh
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 2

Anomaly Detection in IoT Networks

using Autoencoder
1. Introduction
This project aims to design an Autoencoder-based anomaly detection system for identifying
potential security threats in IoT networks. The system is trained on normal traffic data from
the CICIDS 2018 dataset and uses reconstruction errors to detect anomalies.

2. Dataset Description
The CICIDS 2018 dataset is a comprehensive collection of labeled network traffic data,
including both normal and various malicious traffic types. For this project, normal traffic is
used to train the Autoencoder, while both normal and attack data are used for testing.

3. Data Preprocessing
Preprocessing steps include handling missing values, encoding categorical features, and
normalizing the dataset using StandardScaler. These steps ensure that the data is clean and
ready for training.

4. Autoencoder Model Design


The Autoencoder model consists of an encoder and a decoder. The encoder compresses the
input data into a lower-dimensional representation (bottleneck), and the decoder
reconstructs the original input from this compressed form. Mean Squared Error (MSE) is
used as the loss function.
Figure 1: Autoencoder Neural Network Architecture

5. Model Training and Evaluation


The Autoencoder is trained exclusively on normal traffic. After training, reconstruction
errors are computed. A threshold is set (95th percentile of normal errors) to detect
anomalies. Performance is evaluated using confusion matrix, accuracy, precision, recall, F1-
score, and AUC.

6. Python Code (Key Snippets)


The following code demonstrates the data preprocessing, model building, training, and
evaluation process.

# Load and preprocess dataset


data = pd.read_csv('CICIDS2018_sample.csv')
data.dropna(inplace=True)
...

7. Comparison with Other Models


The performance of the Autoencoder is compared with Isolation Forest and One-Class SVM.
Additional classification models like Random Forest and SVM were also used for
comparison. Hyperparameter tuning was applied using Grid Search to improve model
performance.

8. Conclusion
The Autoencoder effectively detected anomalous network traffic with high accuracy. Its
performance was comparable or superior to traditional anomaly detection methods, making
it a suitable choice for securing IoT networks.

You might also like