This is the second installment in a two-part series about distributed denial-of-service (DDoS) attacks and mitigation on cloud. Be sure to read part one for an overview of denial-of-service (DoS) and DDoS attack variants and potential consequences for cloud service providers (CSPs) and their clients.
In the first installment of this series, we demonstrated how cybercriminals could circumvent DoS defenses by launching distributed DDoS attacks. The three major types of DDoS variants are:
- Volume-based attacks
- Protocol attacks
- Application-layer attacks
We can demonstrate how these attacks work in a simulated environment using Graphical Network Simulator-3 (GNS3), a network simulation tool.
To understand this, first let’s break down the network diagram below:
Figure 1: A corporate network configured with OSPF and BGP
The diagram shows a network designed with routers and configured with Open Shortest Path First (OSPF), the company’s internal network, Border Gateway Protocol (BGP), the edge router that reveals the internet service provider (ISP) to the end users and clients and other network devices.
Now let’s examine how threat actors can exploit these systems to launch various types of DoS and DDoS attacks.
Volume-Based DDoS Attacks
Cybercriminals typically leverage tools, such as Low Orbit Ion Cannon (LOIC) and Wireshark to facilitate volume-based attacks through techniques like Transmission Control Protocol (TCP) and User Datagram Protocol (UDP) flooding. Let’s take a closer look at how these attacks work.
In a TCP flooding attack, threat actors generate a large quantity of traffic to block access to the end resource. The magnitude of this type of attack is commonly measured in bits or packets per second. The diagrams below show a TCP flood attack in which the File Transfer Protocol (FTP) service is flooded with huge volumes of TCP traffic, which eventually brings down the service.
Figure 2: A user connecting to an FTP server hosted on a corporate network
Figure 3: An attacker using bots to send malicious traffic to the target port using the LOIC tool
Figure 4: A client unable to access the FTP service after an attacker has flooded it with corrupt FTP packets
UDP flooding means overwhelming the target network with packets to random UDP ports with a forged IP address. It is easy to use a forged IP address in this type of attack since UDP does not require a three-way handshake to establish a connection. These requests force the host to look for the application that is running on those random ports (which may or may not exist) and flood the network with Internet Control Message Protocol (ICMP) destination unreachable packets, thereby blocking legitimate requests.
There are other variations of UDP flooding, such as reflection and amplification attacks. In a reflection attack, a threat actor uses publicly available services, such as the Domain Name System (DNS), to attack the target networks. An amplification attack, on the other hand, targets a protocol in an attempt to amplify the response. For example, an attacker might submit a single query of *.ibm.com to the DNS, which will then gather a massive volume of information related to subdomains of IBM.com.
Figure 5 shows a similar attack using the Network Time Protocol (NTP). This protocol enables network-connected devices to communicate and synchronize time information, which is communicated over UDP. An attacker can forge the source IP address and then use a publicly available NTP application to send queries to the target. Common tools used in this type of attack include Nmap, Metasploit and Wireshark.
Figure 5: An attacker using Nmap to discover hosted NTP servers
Figure 6: An attacker using Metasploit to determine that the target NTP server is vulnerable to a MOD6 UNSETTRAP distributed, reflected denial-of-service (DRDoS) attack, an amplification of 2X packets
In this case, the victim’s response packet would be twice the size of the packet the NTP request sent. By repeatedly sending the request, an attacker could flood the target network with a huge number of responses.
In the scenario shown below, an attacker sends multiple SYN request from several spoofed Internet Protocol (IP) addresses to a corporate network’s Secure Shell (SSH) jump server to disrupt the service. Tools such as Hping3 and Wireshark are commonly used in this type of attack.
Figure 7: A client (Ubuntu Machine) connecting to a company’s jump server (IP: 22.214.171.124) for remote administration
Figure 8: An attacker performing a protocol DDoS attack on a jump server (target IP: 126.96.36.199), preventing the client from accessing the jump server
Figure 9 shows a real-world exploit of a TCP SYNC flood attack performed on a web application as part of a penetration testing (PT) engagement.
Figure 9: A web application becomes unresponsive after a TCP SYNC flood attack
In addition to volume-based and protocol attacks, cybercriminals can also launch DDoS campaigns by targeting the application layer. Below are some variations of this attack type.
Slowloris is a very prominent attack in which the connection is never idle but, as the name suggests, it is slow. The client connects gradually by sending data and connection requests to the server. This keeps the connections open indefinitely and, as a result, the server cannot process any new connections. Threat actors typically use Slowhttptest and Wireshark to facilitate this attack.
Figure 10: A client accessing a web server hosted on a company’s cloud network
Figure 11: A legitimate user unable to access a webpage due to a Slowloris attack
Shown below is a real-world exploit of Slowloris performed on a web application as part of a penetration testing exercise.
Figure 12: A web application becomes unresponsive after a Slowloris attack
In an HTTP flood DDoS attack, the attacker sends an HTTP GET/POST request, which seems to be legitimate, to infiltrate a web server or application. Instead of using a forged IP address, this attack leverages botnets, which require less bandwidth. An HTTP flood attack is most effective when it forces the server or application to allocate the maximum resources possible in response to every single request.
Shown here is a real-world HTTP flood attack performed using a Session Initiation Protocol (SIP) INVITE message flood on port 5060, rendering the phone unresponsive.
Figure 13: An attacker performing a SIP INVITE flood attack on an IP phone
Figure 14: The IP phone becomes unresponsive after the attack
DDoS Mitigation On Cloud
To mitigate DDoS attacks on the cloud, security teams must establish a secure perimeter around the cloud infrastructure and allow or drop packets based on specified rules. Below are some key steps organizations can take to harden their security environments to withstand DDoS attempts.
A next-generation firewall is capable of performing intrusion prevention and inline deep packet inspection. It can also detect and block sophisticated attacks, including DDoS, by enforcing security policies at the application, network and session layers. Next-generation firewalls give security teams granular control to define custom security rules pertaining to network traffic. They also provide myriad security features, such as secure sockets layer (SSL) inspection, web filtering and zero-day attack protection.
Content Delivery Network
A content delivery network (CDN) is a geographically distributed network of proxy servers and their data centers that accelerates the delivery of web content and rich media to users. Although CDNs are not built for DDoS mitigation, they are capable of deflecting network-layer threats and absorbing application-layer attacks at the network edge. A CDN leverages this massive scaling capacity to offer unsurpassed protection against volume-based and protocol DDoS attacks.
DDoS Traffic Scrubbing
A DDoS traffic scrubbing service is a dedicated mitigation platform operated by a third-party vendor. This vendor analyzes incoming traffic to detect and eliminate threats with the least possible downtime for the target network. When a DDoS attack is detected, all incoming traffic to the target network is rerouted to one or more of the globally distributed scrubbing data centers. Malicious traffic is then scrubbed and the remaining clean traffic is redirected to the target network.
An anomaly, such as an unusually high volume of traffic from different IP addresses for the same application, should trigger an alarm. But anomaly detection is not quite that simple since attackers often craft packets to mimic real user transactions. Therefore, detection tools must be based on mathematical algorithms and statistics. This works well for both application-based and protocol attacks.
Source Rate Limiting
As the name suggests, source rate limiting blocks any excess traffic based on the source IP from where the attack originates. This is mainly used to limit volume-based traffic by configuring the thresholds and customizing responses when an attack happens. Source rate limiting provides insights into particular websites or applications on a granular level. The drawback is that this method only works for nonspoofed attacks.
Protocol Rate Limiting
This technique blocks suspicious protocols from any source. For example, the Internet Control Message Protocol (ICMP) can be blocked after a fixed rate — say, 5 megabits per second (Mbps) — thereby blocking bad traffic and allowing legitimate traffic. While it works well for volume-based attacks, the limitation of protocol rate limiting is that sometimes even legitimate traffic will be dropped, requiring security teams to manually analyze logs.
Cloud Security Is More Crucial Than Ever
With more and more applications now migrating to the cloud, it is more crucial than ever to secure cloud infrastructure and the applications hosted therein. The DDoS attacks described above can put CSPs and their clients at great risk of data compromise. By employing various defense mechanisms, such as advanced firewalls, traffic scrubbing and anomaly detection, organizations can take major steps toward securing their cloud environments from DDoS attacks.