Supply-chain attack using invisible code hits GitHub and other repositories
· Ars Technica
Researchers have uncovered a sophisticated supply-chain attack that is compromising repositories like GitHub, utilizing malicious packages embedded with invisible code. This alarming discovery was made by security firm Aikido Security, which identified 151 such packages uploaded between March 3 and March 9 of this year. Supply-chain attacks have been a persistent threat for nearly a decade, typically involving the upload of harmful packages that mimic popular code libraries in order to deceive developers into integrating them into their projects. In some instances, these malicious packages have been downloaded thousands of times, posing significant risks to software integrity.
The novel aspect of this attack lies in the use of invisible code, which is undetectable in most editors, terminals, and code review tools. While the visible parts of the code may appear harmless, the malicious functions are concealed using unicode characters that are imperceptible to the human eye. This technique dramatically undermines traditional security defenses and manual code reviews, making it increasingly difficult for developers to identify potential threats. Other platforms, including NPM and Open VSX, have also reported similar attacks, highlighting the widespread nature of this issue.
Aikido's researchers noted that the malicious code is cleverly integrated into seemingly benign updates, such as documentation changes or minor bug fixes, making it even harder to spot. They suspect that the group behind the attacks, which they have dubbed "Glassworm," may be leveraging large language models (LLMs) to generate these convincingly legitimate packages. The sheer volume of bespoke code changes across various codebases suggests that manual crafting is no longer feasible, reinforcing the belief that AI tools are being utilized in this malicious campaign.