Small office home office networks have become a target for many threat actors, hackers and cyber attackers and hence there is an urgent need to secure the network from such attackers. Most small office home office network users do not see the need to provide enough security to their networks because they assume no one is going to hack them forgetting that the biggest threat of our small home networks today comes from the outside. The challenge of misconfiguration of routers, firewalls and default configurations in our small home networks renders the network vulnerable to attacks such as DDos , phishing attacks , virus and other network attacks hence the need to implement a detection algorithm to help identify flaws in the pattern of the small office network. It turns out that about 75% of existing approaches focused on intrusion detection in 802.11 wireless networks of a SOHO and not the entire network. These approaches do not efficiently secure the network entirely leaving the rest prone to attacks can occur with or without the internet. This paper proposes to add another layer of security to the other preventive measures in a SOHO network by designing, implementing and testing a supervised neural network algorithm to identify attacks on the small home network and also to send a notification to users to keep them informed of the activities on their network. The supervised neural network algorithm will have a dataset representing both attacks and non-attacks which will be used in the training phase. The system should be able to detect and identify the various attacks and anomalies when they occur on the network and help keep the users informed.
Signature-based intrusion detection methods report high accuracy with a low false alarm rate. However, they do not perform well when faced with new or emerging threats. This work focuses on anomaly-based data driven methods to identify potential zero-day-attacks using a specific class of neural networks known as the autoencoder.