Logging and monitoring are two important aspects of any security program. Without logging, we cannot understand how our systems operate, and without monitoring, we cannot detect anomalies and issues before they become problems.

There are many different types of logs available to us today. Some are generated automatically, while others require manual intervention. For instance, network traffic is usually logged automatically. However, application logs are not. We may need to manually create these logs.

Application logs provide valuable information about what happened during the execution of an application. They can show us which parts of the application were executed, what resources were used, and what was returned. Application logs are often stored in databases, allowing us to query them later.

Network logs are also useful. They allow us to see what packets were sent and received, and what responses were made. 

System logs are another type of log that we should consider. System logs record events such as system startup, shutdown, reboots, etc. They are generally stored in files, but can also be recorded in databases.

While logs are very helpful, they do have their limitations:

• First, logs are only as good as the people who generate them. If  something doesn’t save a log, then we likely don’t know what happened. We might be able to get that from some other log, but having multiple layers of logs around an event is often useful.
• Second, logs are static. Once created, they should remain unchanged. Hashing logs, storing them on read only file systems and other forms of log controls are highly suggested.
• Third, logs are not always accurate. Sometimes, logs contain false positives, meaning that something appears to be happening when actually nothing is. False negatives are also possible, meaning we don’t alert on something we should have. Logs are a part of detection solution, not the sole basis of one.
• Fourth, logs are not always actionable. That means that we can’t easily tell from a log whether something bad has occurred or if it is just noise. This is where log familiarity and anomaly detection comes in. Sometimes reviewing logs in aggregate and looking for trends is more helpful than individual line by line analysis. The answer may be in looking for haystacks instead of needles…
• Finally, logs are not always timely. They might be created after the fact, and therefore won’t help us identify a problem until much later. While good log analysis can help create proactive security through threat intelligence, they are more powerful when analyzing events that have happened or as sources for forensic data.

Keep all of these things in mind when considering logging tools, designing monitoring techniques or building logs for your systems and applications.