This research aims to design and develop a secure mechanism for cloud data protection and attack detection. The project consists of two main modules: Data Protection Scheme and User Attack Detection Module.
1. Data Protection Scheme
Data is split into chunks and encrypted before being uploaded to the cloud.
Encryption is based on a modified Huff scheme with an optimal key generation approach inspired by coati and coyote characteristics.
Privacy Preservation is ensured through data sanitization.
2. User Attack Detection Module
Deep Learning Model for attack detection utilizing the BoT-IoT dataset.
Features a Bidirectional LSTM (BiLSTM) with attention layers encapsulated in encoder-decoder modules.
The model focuses on identifying potential attackers through attention mechanisms and optimally tuned parameters.
Confidentiality:
Asymmetric encryption (1:1) and ABE model (1
).
Includes byte substitution, row shifting, column mixing, and round key generation.
Authentication:
Modified hashing for generating 512-bit and 1024-bit keys.
Incorporates byte substitution, row mixing, and column shifting.
Deep Learning for Attack Detection:
BiLSTM processes input sequences in both forward and backward directions.
An attention mechanism highlights crucial input sequence elements for better attack detection.
A context vector is used to improve attack detection accuracy.
The project is implemented in Python and evaluated based on data privacy, detection accuracy, precision, and recall. A comparative analysis with existing methods will be provided.