3. z
Literature Review
1. DoS/DDoS attacks are done on Websites
which mostly runs on FIFO/Drop-Tail
algorithms
2. Almost 65% of web cyber attacks
happens with malicious ware of DoS/DDoS
4. z
Litreature Review
3. DoS/DDoS attacks are done for gaining
personal benefits or for ransom money by an
Individual or a group of attackers.
4. Each system has its own methodologies.
A system always contains some weak-points
to host an attack or loop holes which the
attackers use to attack the system
5. z
Prevention
DDoS/DoS attacks can be prevented using a secure network of
connection and with firewalls made to prevent DoS/DDoS
attacks
On a private cloud network, these attacks can be prevented by
securing the private network established in the cloud
The best way to prevent the attacks is to update the fire-walls
and strengthen the system against such kind of attacks from
time-to-time and regular updates
6. z
Progress Report
Found the dataset on Kaggle
Did data pre-processing and reduced number of features
Used algorithms like K-NN and DCT
Got accuracies and improved them
Feature selection is done with plasma graph and observations
7. z
Findings
65% of network attacks are done with DoS/DDoS attacks
Used for personal gains/money
Attacks usually occurs when the website is growing popular on
networks/ when some new updates are applied on website.
Best algorithm to detect the attacks is RF and Naïve bais algorithm
Without outliers reduction and feature reduction, the accuracies of
models DCT and KNN were around 70-75%
With data pre-processing and feature selection accuracy is
increased to over 95%
8. z
Future Work and Conclusion
Applying more data pre processing, reducing the feature
dependency
Exploration of more models and finding the best one out
Trying to provide a better solution then used in current systems
Create a model to predict various DoS/DDoS attacks with a
single algorithm
9. z
Future Work and Conclusion
Current Best model with highest accuracy is Decision Tree