Amazon CodeGuru is a developer tool that provides intelligent recommendations for identifying security risks in code and improving code quality. To help you find potential issues related to logging of inputs that haven’t been sanitized, Amazon CodeGuru Reviewer now includes additional checks for both Python and Java. In this post, we discuss these updates and show examples of code that relate to these new detectors.
In December 2021, an issue was discovered relating to Apache’s popular Log4j Java-based logging utility (CVE-2021-44228). There are several resources available to help mitigate this issue (some of which are highlighted in a post on the AWS Public Sector blog). This issue has drawn attention to the importance of logging inputs in a way that is safe. To help developers understand where un-sanitized values are being logged, CodeGuru Reviewer can now generate findings that highlight these and make it easier to remediate them.
The new detectors and recommendations in CodeGuru Reviewer can detect findings in Java where Log4j is used, and in Python where the standard logging module is used. The following examples demonstrate how this works and what the recommendations look like.
Findings in Java
Consider the following Java sample that responds to a web request.
This simple example generates a result to the initial request, and it extracts the
userId field from the initial request to do this. Before returning the result, the
userId field is passed to the
log.info statement. This presents a potential security issue, because the value of
userId is not sanitized or changed in any way before it is logged. CodeGuru Reviewer is able to identify that the variable
userId points to a value that needs to be sanitized before it is logged, as it comes from an HTTP request. All user inputs in a request (including query parameters, headers, body and cookie values) should be checked before logging to ensure a malicious user hasn’t passed values that could compromise your logging mechanism.
CodeGuru Reviewer recommends to sanitize user-provided inputs before logging them to ensure log integrity. Let’s take a look at CodeGuru Reviewer’s findings for this issue.
An option to remediate this risk would be to add a
sanitize() method that checks and modifies the value to remove known risks. The specific process of doing this will vary based on the values you expect and what is safe for your application and its processes. By logging the now sanitized value, you have mitigated those risks that could impact on your logging framework. The modified code sample below shows one example of how this could be addressed.
The example now uses the
sanitize() method, which uses a
replaceAll() call that uses a regular expression to remove all non-digit characters. This example assumes the
userId value should only be digit characters, ensuring that any other characters that could be used to expose a vulnerability in the logging framework are removed first.
Findings in Python
Now consider the following python code from a sample Flask project that handles a web request.
In this example, the input variable is assigned the
input query string value from a web request. Then, the Flask logger records its value as an info level message. This has the same challenge as the Java example above. However this time rather than changing the value, we can instead inspect it and choose to log it only when it is in a format we expect. A simple example of this could be where we expect only alphanumeric characters in the
input variable. The
isalnum() function can act as a simple test in this case. Here is an example of what this style of validation could look like.
While log sanitization implementation is a long journey for many, it is a guardrail for maintaining your application’s log integrity. With CodeGuru Reviewer detecting log inputs that are neither sanitized nor validated, developers can use these recommendations as a guide to reduce risks related to log injection attacks. Additionally, you can provide feedback on recommendations in the CodeGuru Reviewer console or by commenting on the code in a pull request. This feedback helps improve the precision of CodeGuru Reviewer, so the recommendations you see get better over time.
To get started with CodeGuru Reviewer, you can leverage AWS Free Tier without any cost. For 90 days, you can review up to 100K lines of code in onboarded repositories per AWS account. For more information, please review the pricing page.