Multitier topology

Multitier topology gives you flexibility to segment resources based on role and access policies. In a typical three-layer application, architecture that has web, app, and DB servers can be distributed based on location. Since web/app zone is something always exposed to end users, Demilitarized Zone (DMZ) IP space is always public. Subnet and database servers should not be directly accessible, hence why we should always allocate private IP space from RFC 1918.

This offers gradual access to control, based on IPs and resource locations. When designing a network, you can introduce a multi-layer firewall approach. In a multiple layer design approach, the basic idea is to isolate resources from each other, considering the fact that if one layer is compromised then others are not impacted.

Cross-premises IPsec tunneling provides you with a way to establish secure connections between two networks and multiple on-premises sites, or other virtual networks in Azure/AWS. This can secure data transfer by encrypting your data via the IPsec encryption using the IPsec framework. Virtual networks in AWS are called VPC and, in Azure, VNET.

Distributed Denial of Service: A Denial-of-Service (DoS) attack or Distributed Denial-of-Service (DDoS) attack  is an attempt to make a network resource out of service to its targeted users.

The real-world target would be online services such as e-commerce and the gaming industry, preventing the shop from doing any business by making front resources unavailable for end users. Just think about a situation during big billion-day sales hours if someone launches a DDOS attack and makes your e-commerce portal shut down.

The two most basic types of DDoS attacks are as follows:

• WAN attacks: WAN DDoS attacks utilize available bandwidth on physical links with a high volume of packets with bigger payloads, or a high volume of packets with smaller payloads. Bigger payload network resources such as router or firewalls will process packets and consume all the bandwidth. With smaller payload network resources like routers, firewalls will try to process all the packets. However, due to limited CPU, cycle hardware resources won't be able to process genuine packets from end users and can fail under the load.

• For example, let's assume you have a 10 Mbps WAN link and during attack BW, utilization is just 5 Mbps. However, a number of small packets can reach one million packets per second. In this case, assume that your network gear has no CPU cycle to process all tiny packets

  Another example would be if someone launched a DDOS attack using a large ICMP packet. This can choke your bandwidth and leave no space for the rest of the application.

• The most common form of bandwidth attack is a packet-flooding attack, in which a large number of legitimate TCP, User Datagram Protocol (UDP), or Internet Control Message Protocol (ICMP) packets are directed to a targeted or aimed destination. Such attacks become more difficult to detect if attackers use techniques such as spoofing source addresses.

• Application attacks: These DDoS attacks use the expected behavior of protocols such as TCP and HTTP. Application attacks are disruptive but small and silent in nature and extremely hard to detect since they use expected behavior. Application-layer attacks are easy to generate and require fewer packets with a small payload to achieve out of services for targeted applications. Application attacks are focused on web-application layers. For a small HTTP request, the actual server has to execute a lot of resources on the web server to fetch the content or resources. Every such server resource will have limited CPU and memory and can be easily targeted. In this example, I am not considering cloud-based web applications, where you have elasticity features enabled and with growth in the number of requests, server resources are automatically created to accommodate such requests.

Let us understand more about this with the help of an example:  

• HTTP Floods: These are simple attacks in nature that try to access the same web page again and again in an automated fashion. They typically use the same range of IP addresses. Based on the trend, as this is being originated from the same source, the source pool can be blocked to mitigate attacks.
• Randomized HTTP Floods: These are complex attacks that use a large pool of IP addresses from multiple locations and randomize the URLs. Since these kind of attacks originate from multiple locations, it is not easy to block the source IP. However, the rate limit can be fixed on server resources.

To simplify, DDoS is a form of attack where multiple compromised networks/hosts are used to target a single system. This is like a zombie attack and it is very tough to identify genuine users. Once infected, the internet-connected devices become part of a botnet army, driving malicious traffic toward a given target.