False Alarms & How To Provide Them

From Computing and Software Wiki

Current revision as of 01:48, 24 March 2008

Network-based intrusion detection systems (NIDS) perform in-depth packet analysis in order to enumerate attackers who are attempting to expose network and service vulnerabilities. NIDS devices can also aid in identifying misuse patterns and gathering forensic data. By examining network traffic in real time, NIDS devices can alert users to possible attacks and/or take predefined responsive actions to help mitigate the threat. By providing an additional layer of protection above and beyond access control devices such as a firewall, NIDSs can be a valuable addition to the security arsenal. However, network intrusion detection has been criticized for its propensity to generate a perceived large amount of false positives and false negatives. Effective NIDS device management can appreciably reduce these reporting inaccuracies

False Positives

The term false positive is a broad and somewhat vague term that describes a situation in which an NIDS device trigger an alarm in a when there is malicious activity or attack occurring. Other common terms used to describe this condition are "false alarms" and "benign trigger". False alarm is the better term to describe this behavior since "false positive" gives the impression that IDS technology itself is fundamentally flawed and benign trigger gives the impression that there is no possibility for a true false positive to exist. Here I will use the term false alarm to describe the general condition of an alarm being generated without a true security related event. False alarms are the Internet security equivalents of the boy who cried wolf. They are problematic because by triggering unjustified alerts, they diminish the value and urgency of real alerts.

False Negatives

False negative is the term used to describe a network intrusion device's inability to detect true security events under certain circumstances. In other words, malicious activity is not detected and alerted. Fortunately, there are actions that can be taken to reduce the chance of false negative conditions without increasing the number of false positives. The difficulty in creating this "balance" is to create a more manageable NIDS deployment without introducing extra risk. First, however, we need to analyze how network intrusion detection systems detect these attacks so we can understand the consequences associated with our actions.

Common techniques that can be used to reduce false alarms include placing the NIDS device behind a firewall, tuning signatures to only include certain platforms and or services offered on the network, and reduction through thorough network analysis. Each of these techniques has potential benefits and drawbacks.

Placing the NIDS Behind a Firewall

Placing the NIDS behind a firewall is a common technique that is very easy to implement. This technique requires no alteration of the default configuration of the device and very little expertise. When an alarm is received there is a high percentage of certainty that is it a real threat than if the NIDS was not behind the firewall. However, there are still instances where it will trigger benign alarms, such as the web page posted by Joe sys-admin about the Code Red worm from the example in the the first article in this series. There also exists a potential for false negatives such as denial of service events. The DAA formula for this design would be as follows:

Reconnaissance attempts = 3 (host sweeps may be recognized, port scans probably not) DOS recognition = 3 (The IDS will only see DOS attacks on services allowed through the network firewall) Service compromise = 5 DAA = 15

Tuning NIDS Signatures

Another common and relatively easily deployable solution is to tune the NIDS device's signatures to only watch for services or operating system-specific conditions that apply to the network being monitored. This requires more skill and vulnerability knowledge than does placing the device behind a firewall and it is less prone to false alarms. However if we analyze the DAA formula we realize that the fruits of our labor our not really realized.

Reconnaissance attempt = 5 (host sweeps and port scans recognized) DOS recognition = 5 (device will see all DOS attacks on network) Service Compromise = 3 (will allow an attack until the attacker triggers an alarm that actually effects the network being monitored) DAA = 15

Network Analysis

The most laborious solution is to reduce false alarms through network analysis. This requires considerable networking knowledge, expertise, analysis and some imagination. In this design, the NIDS device is placed outside the firewall with a full complement of signatures. Alarms are analyzed and then tuned very specifically to eliminate common false alarms. The engineer must be careful not to introduce false negative conditions through alarm filtering. Done correctly, this method is very effective in mitigating risk. Most of this analysis is done within the first few weeks of implementation, however some ongoing analysis is required. The DAA for this method reflects its effectiveness.

Reconnaissance attempt = 5 (host sweeps and port scans recognized) DOS recognition = 5 (will see all DOS attacks on network) Service Compromise = 5 (will alert on any attack at a service) DAA = 25