Software Supply Chain Attacks in 2025: From Typosquatting to AI-Generated Malware
Published on 2025-05-30 by Security Research Team
The Scale of Package Ecosystem Attacks
Software supply chain attacks targeting package managers surged through 2025. The npm, PyPI, and RubyGems ecosystems each saw thousands of malicious package uploads, with attackers using increasingly sophisticated techniques to evade detection. Typosquatting remains the most common vector: packages with names like 'reqeusts' or 'colurs' that exploit common misspellings of popular libraries. But the attack sophistication has increased substantially.
Dependency confusion attacks exploit the resolution order of package managers that check both public and private registries. An attacker publishes a package to the public registry with the same name as an internal package, using a higher version number. If the build system checks the public registry first or prefers higher versions, the malicious package is installed instead of the legitimate internal one. Microsoft first documented this attack in 2021, but it continues to succeed against organizations that have not properly configured their registry resolution policies.
The newer threat is AI-assisted package generation. Attackers use LLMs to generate packages that include realistic documentation, test suites, and functional code alongside hidden malicious payloads. These packages pass casual human review because they appear to be legitimate open-source contributions rather than hastily assembled attack vectors.
SBOM Adoption and Its Limits
Software Bill of Materials (SBOM) adoption accelerated following executive orders and EU regulations mandating supply chain transparency. Tools like Syft, CycloneDX, and SPDX generate machine-readable inventories of every component in a software product. SBOMs enable organizations to quickly identify whether they are affected when a vulnerability is disclosed in a transitive dependency.
However, SBOMs address visibility, not prevention. Knowing that your application depends on a vulnerable library is useful only if you have a process to act on that information quickly. Many organizations generate SBOMs for compliance purposes but lack the operational maturity to consume them effectively. The gap between SBOM generation and vulnerability remediation remains a significant challenge.
Build Provenance and Signed Artifacts
Sigstore and SLSA (Supply-chain Levels for Software Artifacts) represent the most promising structural defenses against supply chain attacks. Sigstore provides keyless code signing using short-lived certificates tied to developer identities, eliminating the key management burden that historically prevented widespread adoption of package signing. SLSA defines a maturity model for build provenance, from basic source tracking to fully hermetic, reproducible builds.
At SLSA Level 3, build artifacts include a cryptographically signed provenance attestation that records the exact source revision, build configuration, and builder identity. Consumers can verify that a package was built from the claimed source code by a trusted builder, making it significantly harder for attackers to inject malicious code during the build process. Adoption of SLSA Level 3 remains low, but major package registries are beginning to require provenance attestations for new uploads.
Peer-reviewed and accepted for publication. Supplementary data available via linked endpoints.