Open-source malware has changed shape.
What once focused on noisy cryptomining has moved toward something far more valuable: access.
Our recent data shows attackers are increasingly targeting credentials and secrets embedded in software dependencies, with UK organizations firmly in scope.
Latest Videos FromTechRadar
This shift marks a move away from opportunistic abuse toward deliberate supply-chain compromise. Instead of draining compute cycles, attackers are positioning themselves inside build pipelines and developer workflows.
The goal is persistence, not disruption.
For organizations that rely heavily on open source software, this fundamentally changes both the threat model and the potential impact.
This is what “shift left” actually means in 2026: controlling what enters the build, not just detecting what runs in production.
Why credential theft has overtaken cryptomining
More than half of malicious open-source packages now focus on stealing credentials and secrets, overtaking cryptomining as the dominant threat type. The reason is straightforward. Credentials offer lasting value. They provide persistent access, broader reach across environments, and a lower risk of detection than resource abuse. A stolen token or API key can unlock entire systems, not just a single machine.
Cryptomining, by contrast, is easy to spot and quick to shut down. It consumes resources and triggers alerts. Credential theft blends in and can be executed in seconds. It exploits the trust placed on developer workflows to operate in a safe environment.
For attackers looking to maximize return while minimizing exposure, this approach maximizes returns whilst doing away with the risk of being discovered.
The implication is clear: protecting runtime infrastructure is no longer enough. The security boundary now starts at dependency intake and at the developer environment.
Multi-stage malware becomes the norm
Modern open-source malware is rarely single-purpose. Our analysis shows dropper and loader behavior increasing by nearly 2,900 percent year over year in Q1 2025, signaling a shift toward engineered, multi-stage attacks.
Around 77 percent of malicious packages distributed through open source ecosystems now combine multiple threat types. Droppers appear in nearly all observed cases, while secret exfiltration features in close to two-thirds.
These packages are designed to evolve after installation, pulling in additional payloads or changing behavior over time. This reflects industrialized campaigns rather than opportunistic experimentation. Attackers are investing in resilience, stealth, and scale.
For defenders, this means signature-based thinking is outdated. If malware is modular and adaptive, controls must focus on provenance, behavior, and prevention before execution. Again, this is what “shift left” actually means: securing the build graph itself, not just the workloads it produces.
Supply chains under direct pressure
The widespread use of open source, particularly within the JavaScript ecosystem, creates systemic exposure. Modern applications routinely depend on hundreds of direct and transitive npm packages. That density of reuse creates efficiency, but also amplifies upstream risk.
Recent activity linked to the Lazarus group illustrates the threat. More than 200 malicious packages were identified, almost all concentrated in npm. When a single ecosystem underpins financial services platforms, government services, and critical national infrastructure, concentration risk becomes a strategic issue.
A compromised dependency does not stay isolated. It propagates through shared frameworks, internal libraries, and CI pipelines. In sectors built on speed and reuse, upstream compromise quickly becomes downstream impact. This is why dependency governance is no longer just a developer hygiene issue; it is a board-level supply-chain concern.
Automation turns one package into thousands of compromises
Today’s malware increasingly targets CI/CD pipelines and developer workflows optimized for automation. When a compromised dependency enters a build, it can quietly extract API keys, certificates, and access tokens without triggering runtime alerts. Automation does the rest.
What starts as a single poisoned package can spread across hundreds or thousands of builds. The very systems designed to accelerate delivery now accelerate compromise.
The practical takeaway is uncomfortable but necessary: if build systems are automated, security controls must be automated at the same level. Manual review cannot scale against automated distribution.
AI coding assistants and the hallucination problem
AI-assisted development introduces an additional layer of risk. Studies and testing have shown that large language models can, in a meaningful percentage of cases, suggest packages or functions that do not exist. Developers under time pressure may attempt to install or rely on these hallucinated dependencies, unknowingly expanding the attack surface.
Hallucinated package names, fabricated examples, and unsafe dependency suggestions can quietly undermine supply-chain integrity. Attackers are already exploiting naming conventions and trust models to seed packages that appear legitimate to both humans and machines.
Each hallucination creates rework, friction, and lost productivity. Much of this waste could be reduced if AI systems were grounded in authoritative, real-time package intelligence rather than pattern prediction alone.
Our recent research reinforces this point. The company found that smaller AI models augmented with live package intelligence significantly outperformed larger standalone models when handling dependency upgrades and package selection tasks. The findings suggest that real-time ecosystem context matters more than model size alone when developers are making security-sensitive decisions. It also helps smaller models are 70x cheaper compared to frontier models.
This has direct implications for software supply-chain defense. If AI coding assistants recommend dependencies without verifying package provenance, maintenance status, or ecosystem trust signals, they risk accelerating the spread of malicious or hallucinated packages into production environments.
In practice, secure AI-assisted development will depend less on increasingly large models and more on whether those models are connected to authoritative, continuously updated software intelligence.
Here too, the lesson is upstream control. Guardrails must sit at the point of dependency selection, not after the code ships.
Why defenders are falling behind
Many UK security controls remain focused on detecting threats after code is deployed. Attackers have moved upstream. They target the build process, the dependency graph, and the trust relationships developers rely on.
This mismatch leaves organizations well prepared for runtime incidents but exposed during development. As long as defenders assume malware announces itself loudly, supply-chain compromise will continue to slip through unnoticed.
“Shift left” is often treated as a slogan. In practice, it means enforcing policy before installation, validating provenance before execution, and blocking malicious packages before they enter the graph.
Stealing the keys, not the cycles
Open-source malware has evolved from stealing compute to stealing access. Credentials unlock ecosystems, not just machines. For UK organisations, this makes supply-chain security a strategic concern rather than a technical afterthought.
Preventing malicious code from entering the build is now more effective than responding after deployment. The quiet shift from coins to credentials has already happened. The question is whether defenses will adapt quickly enough to match it.
If it isn’t automated, it won’t scale.
What organizations should prioritize now
To respond effectively, UK organisations should focus on a small number of structural controls:
● Gate dependency intake with automated policy enforcement before packages enter CI/CD.
● Continuously monitor for secret exposure within build environments and revoke compromised credentials rapidly.
● Enforce provenance and integrity verification for open-source components, including transitive dependencies.
● Ground AI coding tools in authoritative package intelligence to prevent hallucinated or malicious dependency suggestions.
None of these measures eliminate risk. But together, they realign defenses with where attackers are actually operating: upstream, automated, and inside the supply chain.
We’ve ranked and reviewed the best antivirus software available.
This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.
The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit
