In recent years, there’s been a significant emphasis on securing the software supply chain. Especially concerning is the growing number of risks inherent within open source software distributions. This has ignited much development around cloud-native open source security in the form of software bills of materials (SBOMs), projects intended to verify the provenance of OSS packages and more.

Many organizations loop in large open source packages but only use a tiny sliver of the features, opening up an unnecessary attack surface. OSS is still prone to typosquatting and new zero-day exploits. Not to mention exploits like Log4j are still unpatched in a high percentage of deployments.

At the tail end of 2023, I caught up with Ville Aikas, co-founder of Chainguard, to consider the state of cloud-native open source security. I gathered his thoughts along with some other reports and expert viewpoints to paint a picture of where cloud-native supply chain security stands and what to expect in the immediate future.

Recent Developments In Cloud-Native OSS Security

In 2023, we saw many significant developments in the cloud-native space around security, especially regarding standards for securing the supply chain. For instance, maturity around SBOM usage is growing — 53% of the surveyed engineering professionals reported they are generating an SBOM for every application, according to Sonatype’s 9th Annual State of the Software Supply Chain.

Aikas first pointed to the movement around Sigstore for signing software artifacts to verify trust attestations and source of origin. He also highlighted the trend around SBOMs for solving a critical problem in the supply chain, even though he admits that SBOM quality has room for improvement. “New frameworks like OpenVex ultimately can help make SBOM formats more accurate,” he added.

“I’d say the progress this year has been more in the ‘first step’ phase, which is having something in place to address signatures and SBOMs,” said Aikas. “But the next step I think we’ll see is having more high-quality outputs of these tools for security.”

Highest Severity OSS Vulnerabilities

In the aforementioned study, Sonatype reported discovering 245,000 component downloads with known vulnerabilities in the wild. Impressively, 96% of these have a fixed version already available. They’ve also tracked that a quarter of all Log4j downloads are of the vulnerable version holding the infamous and highly exploitable Log4Shell vulnerability.

Typosquatting is still a predominant method for inserting malicious code into OSS software components, which could be used by attackers to exfiltrate keys or create backdoors into systems. However, the next trend could focus on bad actors finding ways to contribute malicious commits to the actual open source projects, predicted Brian Fox, co-founder & CTO at Sonatype and Governance Board Member at OpenSSF. Tomorrow’s sophisticated attack strategies might look like unintentional accidents, like code submissions to ‘fix a bug in code’ that might actually cause an exploit, he explained.

Instead of focusing on the CVE of the week, Aikas encouraged IT leaders to see the bigger picture. “As an industry, we need to start focusing on how we fix the vulnerability problem more holistically and efficiently,” he said. This involves having deep visibility into your software dependencies and a risk management infrastructure in place to enable fast patches, he said.

Generative AI Complicates Cloud-Native Security

Over the past year, generative AI emerged onto the scene in a fierce wave, carrying with it a heap of new cybersecurity implications. Aikas sees data science workflows at an early evolutionary stage, prone to security defects. “The industry needs to quickly come to see this is a problem, and there needs to be checks and balances in place verifying the trustworthiness of AI/ML models,” he said.

Generative AI can be beneficial for automating cloud-native DevOps, code reviews and for specific defensive strategies. Yet, we need better visibility into the data these models were trained on and chains of custody for these models, he said, to avoid poisoning or tampering with AI data models.

Thankfully, we have seen momentum toward heightening visibility and setting security standards around AI/ML. For instance, CycloneDX, the leading SBOM standard, recently incorporated ways to describe machine learning models to its specification. And the new OWASP LLM security Top 10 can act as a guidepost for addressing some key LLM-related threats.

Ways to Mitigate Cloud-Native Supply Chain Threats

Use SBOMs. “SBOMs can give you the visibility into what software you are running and where, which can help when you need to patch a vulnerability, understanding licensing risks or general software end-of-life policies,” said Aikas.

It’s challenging to secure the entire software supply chain on day one. Therefore, Aikas encouraged IT leaders to begin with the low-hanging fruit and implement more security-aware practices, like choosing more secure base images, replacing insecure OSS components with more mature ones and having automated ways to remove CVEs as they arise.

Another way to mitigate cloud-native supply chain threats is to proactively reduce unnecessary or transitive dependencies (which may hold vulnerabilities). Aikas recommends using minimal container images to reduce the overall surface area. “By not having any unnecessary tools or components in your software, an attacker cannot leverage that to bring in more security risk to your environments,” he said.

Observability will also be critical to providing comprehensive monitoring and analysis of the development pipeline, helping to identify anomalies and potential security breaches, said Giordano Ricci, senior software engineer at Grafana. “Importantly, it helps enhance visibility across the supply chain, from source code repositories to deployment environments,” said Ricci. “This transparency allows teams to trace the flow of code and dependencies to identify any unexpected changes or unauthorized access points.”

The End Goal: Secure-By-Design Software

Fewer than half of CISOs believe their developers are very familiar with the security risks of their development tools and workflows, the 2023 CISO & Developer Trends in Software Supply Chain Security Report found. The report also revealed that a significant number of organizations reported vulnerability scanning false positive fatigue.

Given these findings, we need more awareness and collaboration among teams and more accurate and actionable scans, said Aikas. “The prioritization of rapid updates and patching can contribute to removing much of this fatigue.”

“In 2024, I think we will see more of this ‘first step’ progress toward better software supply chain security practices across the industry,” said Aikas. Looking to the future, he also foresees people moving toward a more advanced state of securing the supply chain through strategies such as SLSA provenance and more trustworthy attestations.

“It’s no longer an option to just pass the buck on who owns the security of open source,” said Aikas. “It’s on everyone to take this seriously and start chipping away at what can be done today, incrementally, to get to that steady state of having secure-by-design software.”