How to Tame SQL Injection

For more than a decade, injection vulnerabilities have literally topped the charts of critically dangerous software flaws, deemed more serious than all other types of vulnerabilities in the 2010, 2013, and 2017 Top 10 lists maintained by the Open Web Application Security Project (OWASP).

Still, the warnings have failed to weed out the issues. Last year, the Cl0p group stole data from companies using a previously unknown SQL injection (SQLi) vulnerability in MOVEit's file-transfer application. In late March, the Cybersecurity and Infrastructure Security Agency (CISA) called for companies to redouble their efforts to eliminate the security issue, which application security experts consider one of 13 different "unforgivable" classes of vulnerabilities that programmers should catch during development.

"Despite widespread knowledge and documentation of SQLi vulnerabilities over the past two decades, along with the availability of effective mitigations, software manufacturers continue to develop products with this defect, which puts many customers at risk," the agency stated in its March 25 advisory. "Vulnerabilities like SQLi have been considered by others an 'unforgivable' vulnerability since at least 2007."

The root of injection vulnerabilities is a lack of input sanitization; when the application receives variable input, there's always the risk of that input being tainted, says Randall Degges, head of developer relations at application security firm Snyk.

"Although this has been an issue since programming existed, the reason it’s still in the top 10 vulnerabilities after all this time is because there are an infinite number of ways to use input, and often time sanitizing input is tricky," he says.

For software developers looking to nix this particular issue, here's how.

1. Educate Yourself and Others

The first step is always education. OWASP offers cheat sheets on SQLi, how to detect the vulnerability, and ways of creating safe code. Some Web application frameworks aim to educate developers while they are programming, using application programming interface (API) names to make the risk of some functions clear, such as React's "dangerouslySetInnerHTML" function, says James Kettle, director of research at PortSwigger, an application security testing firm.

In addition, developers should not necessarily trust the makers of open source software — especially components that have not been well vetted — to use safe code, and online tutorials are often unsafe as well, he says.

"I think the core issue is that there's a lot of unsafe APIs, where anyone using the API is vulnerable by default," Kettle says. "Even when there are more modern secure APIs available, fresh code is written using the unsafe versions, thanks to old unsafe examples in StackOverflow."

2. Harden the DevOps Pipeline Using Automated Tools

Developers should implement unit tests to check code for SQLi flaws — and other common security issues — during development, add static application security testing (SAST) both prior to and after commits, and include scans for SQLi as part of dynamic application security testing (DAST).

Unit tests can be added using frameworks such as tSQLt for testing Microsoft SQL Server, pgTAP for testing applications that use PostgreSQL, and Pytest and SQLAlchemy for unit testing in Python programs. A variety of SQL unit testing best practices should be followed to make the tests more useful, such as isolating the SQL tests from dependencies and avoiding descriptive naming of the tests.

In addition to automated tests in the development pipeline, developers should make sure to use SQL frameworks, such as SQLAlchemy, because many security improvements are already baked in, says Snyk's Degges.

"Pretty much all modern SQL frameworks and tools provide convenience methods to help with this nowadays, so your best bet is to thoroughly read through the relevant framework documentation to ensure you're using it correctly when building queries," he says.

3. Play Around With SQLMap

The open source program SQLMap is a great tool for penetration testers to experiment with SQLi, exploit any potential vulnerabilities, and dump a database to prove that the vulnerability can be exploited. The tool can also educate application developers to the true dangers of SQLi and how vulnerable code can be exploited.

However, the tool is not necessarily the best way to scan for potential vulnerabilities, says PortSwigger's Kettle.

"In my experience, the detection capabilities are slow, heavyweight, and prone to false positives," Kettle says. "Also, it can't explore websites to find the attack surface, which is one of the biggest challenges for finding these vulnerabilities automatically."

4. Consider a DAST Service

Automating SQL injection scanning using DAST as part of the quality assurance stage — and even earlier in the DevOps pipeline, if possible — can help catch any overlooked vulnerabilities. In addition, DAST scanning is a good way to find SQLi in legacy code.

While Web application firewalls (WAFs) can prevent SQLi attacks from reaching an application, they should be used only as part of a defense-in-depth strategy, Kettle says.

"Personally, I've seen runtime protection like WAFs bypassed so many times that I don't have much confidence in them," he says. "I would recommend a bug-bounty program as an effective way to surface undetected vulnerabilities instead, and use WAFs as a last resort for systems that are in such a bad state that known vulnerabilities can't be patched."

5. Expand Beyond SQL

Finally, companies should also look for other types of injection vulnerabilities and make sure their developers recognize risky patterns, since SQLi is only one class of injection vulnerabilities.

OWASP broadened the definition of an injection vulnerability to be any software flaw where user-supplied data is not validated or sanitized by an application and then sent to an interpreter. Cross-site scripting, SQL, operating system scripting, and parsing the Lightweight Directory Access Protocol (LDAP) are all areas that can be vulnerable to injection.

With the advent of artificial intelligence models, for instance, prompt injection is the latest form of an injection attack.