A host-based intrusion detection system (HIDS) is an intrusion detection system that is capable of monitoring and analyzing the internals of a computing system as well as the network packets on its network interfaces, similar to the way a network-based intrusion detection system (NIDS) operates. This was the first type of intrusion detection software to have been designed, with the original target system being the mainframe computer where outside interaction was infrequent.
Overview
A host-based IDS is capable of monitoring all or parts of the dynamic behavior and the state of a computer system, based on how it is configured. Besides such activities as dynamically inspecting network packets targeted at this specific host (optional component with most software solutions commercially available), a HIDS might detect which program accesses what resources and discover that, for example, a word-processor has suddenly and inexplicably started modifying the system password database. Similarly a HIDS might look at the state of a system, its stored information, whether in RAM, in the file system, log files or elsewhere; and check that the contents of these appear as expected, e.g. have not been changed by intruders.
One can think of a HIDS as an agent that monitors whether anything or anyone, whether internal or external, has circumvented the system's security policy.
Monitoring dynamic behavior
Many computer users have encountered tools that monitor dynamic system behavior in the form of anti-virus (AV) packages. While AV programs often also monitor system state, they do spend a lot of their time looking at who is doing what inside a computer – and whether a given program should or should not have access to particular system resources. The lines become blurred here, as many of the tools overlap in functionality.
Some intrusion prevention systems protect against buffer overflow attacks on system memory and can enforce security policy.
Monitoring state
The principle operation of a HIDS depends on the fact that successful intruders (hackers) will generally leave a trace of their activities. In fact, such intruders often want to own the computer they have attacked, and will establish their "ownership" by installing software that will grant the intruders future access to carry out whatever activity (keystroke logging, identity theft, spamming, botnet activity, spyware-usage etc.) they envisage.
In theory, a computer user has the ability to detect any such modifications, and the HIDS attempts to do just that and reports its findings.
Ideally a HIDS works in conjunction with a NIDS, such that a HIDS finds anything that slips past the NIDS. Commercially available software solutions often do correlate the findings from NIDS and HIDS in order to find out about whether a network intruder has been successful or not at the targeted host.
Most successful intruders, on entering a target machine, immediately apply best-practice security techniques to secure the system which they have infiltrated, leaving only their own backdoor open, so that other intruders can not take over their computers.
Technique
In general a HIDS uses a database (object-database) of system objects it should monitor – usually (but not necessarily) file system objects. A HIDS could also check that appropriate regions of memory have not been modified – for example, the system call table for Linux, and various vtable structures in Microsoft Windows.
For each object in question a HIDS will usually remember its attributes (permissions, size, modifications dates) and create a checksum of some kind (an MD5, SHA1 hash or similar) for the contents, if any. This information gets stored in a secure database for later comparison (checksum database).
An alternate method to HIDS would be to provide NIDS type functionality at the network interface (NIC) level of an end-point (either server, workstation or other end device). Providing HIDS at the network layer has the advantage of providing more detailed logging of the source (IP address) of the attack and attack details, such as packet data, neither of which a dynamic behavioral monitoring approach could see.