Distributed denial-of-service (DDoS) is a cybersecurity menace which disrupts online services by sending an overwhelming amount of network traffic. These attacks are manually started with botnets that flood the target network. These attacks could have either of the following characteristics:

• The botnet sends a massive number of requests to the hosting servers.
• The botnet sends a high volume of random data packets, thus incapacitating the network.

Time series analysis helps identify network patterns with respect to time. Such pattern detection is done with the historic monitoring of network traffic data. It helps to identify attacks like DDoS. These attacks can be very critical if implemented. Baselining the regular traffic of a network and then overlaying the network with a compromised activity on top of it will help to detect deviations from the normal.

We will be analyzing this use case and will choose a machine learning model that will help detect such DDoS attacks before they crash the entire network. 

We will work with a dataset that compromises traffic received by a website, say, donotddos.com. We will analyze 30 days of historic network traffic data from this website and detect whether the current traffic being received by the website is a part of any DDoS attack or not.

Before we go into the details of this use case we will analyze the datatime data type of Python since it will form the building block of any time series model.