Abstract
By allocating and dynamically changing IP addresses in source and destination hosts, network address space randomization (NASR) aims to construct a dynamic and heterogeneous network to decrease attacking possibility and predictability. This research leverages the features of OpenFlow networks—including data and control plane decoupling, centralized control, and dynamic updating of forwarding rules—to design a novel resolution for IP conversion within the Floodlight controller. The proposed system improves unpredictability and decreases the success rate of worm attacks and IP sniffing through automated IP allocation.
Introduction
Traditional network architectures are inherently static. Once an attacker identifies a target's IP address through sniffing, that target remains vulnerable until the address is manually reconfigured. This static nature is particularly exploited by "hit list" worms, which utilize pre-compiled lists of vulnerable IP addresses to spread rapidly across a network.
Existing Network Address Space Randomization (NASR) solutions often struggle with high complexity, requiring synchronized changes between clients and servers, or high costs associated with modifying host operating systems. Furthermore, changing IPs in traditional architectures frequently interrupts active connections. This paper proposes a Moving Target Defense (MTD) strategy that utilizes Software-Defined Networking (SDN) to manage IP randomization transparently at the network level.
System Architecture and Implementation
The researchers developed a modular extension for the Floodlight SDN controller. The architecture leverages the controller's global view to rewrite packet headers (IP conversion) on the fly, ensuring end-hosts remain unaware that their "public" identity is rotating.
The system architecture introduces three primary components:
- vIP-rIP Conversion Module: This core module maintains a mapping table between "virtual" IPs (vIPs) exposed to the network and "real" IPs (rIPs) used by actual hosts.
- Authentication Module: This verifies devices based on MAC addresses upon initial connection to prevent unauthorized access to the randomized network.
- Virtual IP Generating Algorithm: A hash-based algorithm generates unpredictable vIPs using the formula $a_{(H(K,j) \text{ mod } m) + 1}$, where $H$ is a hash function, $K$ is a shared key, and $m$ is the available address space.
Operational Logic and Flow Management
When a packet enters the network, the OpenFlow switch performs a flow lookup. If no matching rule exists for the current virtual IP mapping, the packet is sent to the controller (the "secure channel") for processing.
To manage the overhead of dynamic rules and prevent "flow table explosion," the system employs a two-timer approach:
- Timer 1: Updates the mapping table to rotate identities at a set frequency.
- Timer 2: Periodically purges obsolete flow table entries in the switches, ensuring the table size remains manageable even as mappings change rapidly.
Experimental Results
The system was evaluated using Mininet and Open vSwitch (OVS), with Nmap simulating attacker behavior. Key findings include:
- Scanning Resistance: Nmap scans successfully discovered less than 5% of virtual IPs, demonstrating high unpredictability.
- Worm Suppression: By reducing the "time-to-live" of valid IP mappings, the system significantly slowed worm propagation compared to static networks.
- Resource Efficiency: The two-timer system successfully maintained a stable, linear flow table length, preventing the exponential growth typically seen in dynamic SDN environments.
Conclusion
By leveraging the centralized control of OpenFlow, this research demonstrates an effective MTD strategy that hinders IP sniffing and worm propagation without requiring host-side modifications. Future work will focus on mitigating controller bottlenecks in high-traffic environments and integrating "random routing" to further enhance network security.
Summary updated