Networking / Beginners

Kernel configurations

The kernel is a relatively small program that controls the most critical resources on the system, such as the hard drives, memory, and video card. The kernel allows for many applications to run simultaneously by controlling their access to critical resources. Applications access these resources through system calls.

Most of the kernel code consists of device drivers-over 90 percent of which are probably not needed by any one particular workstation. Usually, the installation of UNIX or Linux does not include a compilation of the kernel. As a result, the kernel must be prepared to support a wide variety of architectures and hardware configurations. This leads to a lot of code that is not used. As a general security principle, there is no advantage to keeping unused kernel code around. Note that most of this unused code is not compiled directly into the kernel but is available to be loaded as a module when needed. Kernel modules are discussed later in this tutorial in the "Kernel Modules" section.

UNIX has two modes: supervisor mode and user mode. In user mode, library functions are used. These functions then make system calls, which execute on behalf of the libraries. Because the system calls are part of the kernel itself, they have privileged access to critical system resources. Once the task (system call) is completed, control is returned to user mode.

Kernel options

A typical kernel has many options, perhaps as many as 1300 or more in the Linux 2.4 kernel. Some of the more significant security-related options are as follows:

  • iptables-iptables is a powerful firewall that can be used on UNIX workstations. Because iptables operates at the kernel level, it must be compiled into the kernel. iptables is discussed in greater detail in the "Networking" section of this tutorial.
  • IP forwarding-With forwarding turned on, the workstation can function as a gateway or router. Traffic sent to the workstation but destined for a different IP will be routed according to the workstation's route table. This can be a secuity risk. Certain network safeguards may be circumvented because the traffic will appear to come from the workstation instead of the originator. Additionally, if the workstation is multihomed (two or more NICs on differnet subnets), the workstation may allow traffic onto a different network. This may circumvent security controls for that network, such as a firewall or proxy. If not disabled in the kernel, IP forwarding can also be disabled after a system has booted. In Linux, the file /proc/sys/net/ipv4/ip_forward should contain 0 to disable fowarding.
  • Support for multiprocessors-If multiple processors are detected on your workstation, the installation process may configure your boot loader to load a multiprocessor version of the kernel. In most cases, this will not make a difference in the security of the workstation. However, if the workstation is doing development and testing of kernel modules and system calls, the multiprocessor kernel might introduce unwanted effects.
  • Source-routed frames-The kernel can be configured to drop source-routed frames. A source-routed frame is a packet that contains all the information needed for the packet to traverse the network and reach its destination. This source routing is not normally needed and is most often used as a small part of a larger attack. By configuring the kernel to drop source-routed frames, an added measure of security is gained.

The typical UNIX kernel comes with many features enabled that are not required. By rebuilding the kernel and eliminating these options, you will increase the overall security of the workstation. Any unneeded code is a potential source of vulnerability. Additionally, if the workstation is compromised, these unneeded features may be useful to the attacker. Following is a short list of some options that have been turned on. You can see from this small sample that a wide variety of configuration items are possible.

[Previous] [Contents] [Next]