Helping the space industry stay aware of
incidents, threats & vulnerabilities

Briefing 43: Adversary use of AI: 2026 Outlook and Impacts on the Space Sector

2/10/2026

Overview

In February 2026, the Department of Science, Innovation & Technology (DSIT) and the AI Security Institute (AISI) released their International AI Safety Report 2026. This report assesses that general-purpose AI systems are increasingly capable of supporting complex, multi-step tasks across technical domains. While much of the report focuses on safety, governance and misuse prevention, its findings are directly relevant to the cyber threat landscape. The report reinforces the notion that cybercriminals and sophisticated adversaries are no longer just experimenting with AI tools but are beginning to operationalize them across meaningful portions of the cyber kill chain.

Recent reporting from government agencies, private-sector researchers and incident responders indicates that artificial intelligence is becoming a persistent force multiplier in cybercrime and state-sponsored cyber operations. Rather than enabling fully autonomous attacks, observed activity reflects a more pragmatic evolution: AI is being used to accelerate development cycles, scale social engineering and reduce technical friction across attack workflows, with greater autonomy emerging as a longer-term trajectory rather than an immediate capability. These shifts carry direct implications for the commercial space sector, which relies heavily on agile software development, distributed engineering teams and cloud-based ground systems that introduce numerous non-traditional entry points. AI-assisted tradecraft enables adversaries to rapidly adapt and iterate on attack flows tailored to these environments, increasing the likelihood of successful initial access and facilitating lateral movement from enterprise IT networks into mission-critical operational systems.

Evolution of AI-Enabled Threats

Early discussions of adversarial AI usage often centered on theoretical risks of isolated proofs of concept. By contrast, activity observed throughout 2025 and into early 2026 shows a gradual but meaningful transition toward operational use. Threat actors continue to demonstrate integration of AI into discrete tasks where it provides immediate value: code generation, vulnerability research, phishing content creation and workflow automation.

Trend Micro’s State of Criminal AI report from January 2026 underscores this shift, noting that most cybercriminals currently rely on jailbroken commercial large language models (LLMs) such as ChatGPT, Claude and Gemini rather than bespoke models. While this dependence introduces constraints, such as API monitoring and key revocation, it has not prevented adoption. Instead, actors are adapting their tooling and operational security to account for these limitations.

Critically, AI systems are not currently conducting end-to-end cyberattacks autonomously. Human operators remain responsible for target selection, campaign timing and key decision points. However, AI is increasingly embedded within attack workflows, enabling faster iteration and improved scalability. Additionally, the growing prevalence of Agentic AI is lowering the technical barrier for malware development.

The Rise of AI-Assisted Malware Development

The most consequential development observed so far in 2026 is the emergence of AI-assisted and, in some cases, AI-authored malware frameworks. In January, researchers disclosed VoidLink, an advanced Linux-focused malware platform reportedly developed almost entirely by Chinese AI agents. VoidLink features a modular, cloud-aware architecture designed to maintain persistent access across Linux environments, with capabilities tailored for long-term operations. This focus on Linux is particularly relevant to the space sector, where operational systems frequently rely on Linux-based servers, embedded systems and containerized workloads to support command-and-control and data processing functions. As commercial space evolves with cloud-native and hybrid operational models, malware optimized for Linux environments aligns closely with the technical realities of space system operations.

While VoidLink does not represent autonomous malware evolution, its significance lies in how AI was used during development. Tasks traditionally requiring experienced malware engineers were largely automated or accelerated through agent-based AI workflows. This suggests that advanced tooling may no longer be constrained by the availability of elite human developers, particularly for well-resourced state-sponsored actors. For space-sector targets, this lowers the barrier for adversaries to tailor malware to specialized operational environments, increasing the likelihood that custom payloads can be adapted to mission-specific configurations, legacy systems or air-gapped support networks.

VoidLink builds on a trajectory observed throughout 2025. Malware families such as MalTerminal, LameHug, and PROMPTLOCK demonstrated earlier approaches to embedding or querying LLMs during execution, often using static prompts or external API calls to generate code, enumerate systems or assist with payload development. Later variants, including PROMPTFLUX and PROMPTSTEAL, showed more dynamic interaction with language models during runtime, signaling a shift toward adaptive malware behavior.

Augmenting the Cyber Kill Chain

Beyond malware development, AI is increasingly supporting multiple phases of the cyber kill chain. Researchers have demonstrated how AI-assisted tools can expedite vulnerability discovery, a finding reinforced when Microsoft used Copilot to identify previously unknown flaws in open-source bootloaders. Adversaries are likely to apply similar techniques, particularly against widely deployed software and cloud-native components.

Social engineering remains one of the most immediately impacted domains. Deepfake technology is already reshaping phishing, business email compromise (BEC) and vishing campaigns. These tools enable threat actors to produce convincing, tailored lures at scale, while rapidly adapting messaging based on victim responses.

AI is also being leveraged to automate operational overhead. Reporting highlights the use of AI agents to create and manage user accounts, rotate infrastructure and assist with reconnaissance. While these capabilities may appear incremental, they collectively reduce friction and allow actors to sustain higher operational tempo with fewer resources.

Conclusion

Observed activity through early 2026 indicates that AI will continue to reshape cyber operations in subtle but consequential ways. Rather than replacing human operators, AI is amplifying their effectiveness by compressing development timelines, enabling rapid experimentation and lowering the barrier to advanced tradecraft benefiting well-resourced actors while gradually diffusing sophisticated capabilities across broader criminal ecosystems.

For the space sector, AI’s integration across the cyber kill chain increases risk to the interconnected systems that support satellite operations, as adversaries become more capable of rapidly tailoring malware to specialized environments and pivoting from enterprise and development networks into mission-critical operational domains. The most significant implication is not the advent of fully autonomous attacks, but the steady erosion of the time, complexity and specialization advantages that once constrained adversary activity, narrowing the gap between cyber intrusion and operational impact on space missions.

---

Briefing 42: Infostealers, Credential Abuse, and the Weaponization of File Transfer Infrastructure in the Satellite Sector

1/13/2026

Overview:

Recent reporting by HudsonRock highlights a growing and relatively underappreciated cyber risk to the aerospace and satellite industries: the exploitation of corporate file transfer and collaboration platforms using credentials harvested by infostealer malware. These incidents demonstrate how financially motivated threat actors can translate seemingly low-level credential theft into high-impact compromises involving sensitive satellite and defense-related data. The campaign attributed to the Sentap-affiliated actor “Zestix” illustrates how weaknesses in identity security, rather than advanced exploitation, continue to enable serious operational exposure.

About the Threat:

On January 5, 2026, cybersecurity firm HudsonRock reported that dozens of global organizations had been compromised through cloud credentials originating from infostealer infections. These compromises were attributed to a threat actor tracked as Zestix, assessed to be affiliated with the financially motivated cybercriminal group Sentap, which operates as an initial access broker (IAB). Open-source reporting links Sentap to approximately 50 high-profile data breaches spanning late 2024 through 2026.

HudsonRock specializes in tracking infostealer malware ecosystems and has previously documented widespread infections affecting high-security environments, including the U.S. Government and the Defense Industrial Base (DIB). In a February 2025 publication, HudsonRock characterized infostealers as a “cybersecurity disaster in the making,” particularly for defense contractors and space-sector organizations that rely heavily on cloud-based collaboration platforms. The January 2026 reporting builds on this assessment by demonstrating how stolen credentials are operationalized in real-world attacks.

Campaign Overview:

According to HudsonRock, Zestix was observed selling data exfiltrated from corporate file sharing portals belonging to approximately 50 major global organizations. Notably, victims included a Turkish aerospace manufacturer and an Indonesian satellite operator. The compromised data sets reportedly contained sensitive military intellectual property and confidential satellite program documentation, including technical materials associated with prominent defense primes.

The intrusions did not rely on zero-day vulnerabilities or advanced exploitation techniques. Instead, attackers leveraged valid credentials obtained from infostealer malware infections to authenticate directly to corporate file sharing platforms such as ShareFile, OwnCloud and Nextcloud. These platforms are widely adopted across the aerospace, satellite and defense supply chain due to their support for large file transfers, external partner access and distributed engineering workflows.

Once authenticated, threat actors were able to enumerate repositories, download sensitive documentation and package the data for resale. In effect, trusted enterprise infrastructure was transformed into an exfiltration mechanism—without triggering many traditional security controls.

Infostealers as an Enabling Capability:

Infostealers are a class of malware designed specifically to harvest credentials, browser session tokens, cookies and stored authentication data from infected systems. Common families such as RedLine, Lumma and Vidar infect both personal and corporate devices, often through phishing, malicious downloads or trojanized software.

The scale of this threat is significant. According to Flashpoint’s 2025 Global Threat Intelligence Report, infostealer malware infected more than 23 million devices and facilitated the theft of over 2.1 billion credentials in 2024 alone. These credentials are frequently aggregated into underground marketplaces and data dumps, where they may remain unused for extended periods. In 2025 so far, Infostealers.com reports over 17,000 compromised machines and 4,000 compromised users.

A key finding from the HudsonRock investigation is the temporal persistence of risk: While some credentials used by Zestix originated from recent infections, others had been exposed years earlier and were only later weaponized. This highlights that credential compromise is not a point-in-time event, but a long-lived vulnerability that can be exploited opportunistically as access needs arise.

Operational and Sector-Specific Impact:

For the satellite industry, these incidents demonstrate how cyber risks extend beyond traditional IT concerns and into operational, programmatic and strategic domains. File transfer platforms may often be used to host satellite design documentation, information about system architectures, ground segment configurations, interface specifications, supplier and partner deliverables and other test data and planning artifacts.

Unauthorized access to this information can enable intellectual property theft, competitive intelligence collection or downstream targeting of satellite networks and supply chains. Importantly, these risks are not confined to nation-state actors as financially motivated groups have increasingly demonstrated the capability and intent to monetize sensitive aerospace data.

---

Briefing 41: Distributed Jamming and the Implications for Commercial Space

12/16/2025

Overview:

Recent academic research has introduced a new theoretical model for disrupting satellite communications using large swarms of airborne jamming platforms. While not an operational capability, the study reflects a growing interest in exploring scalable electronic warfare (EW) concepts tailored to counter modern low Earth orbit (LEO) broadband constellations. As the commercial space sector becomes more deeply integrated into critical services, research of this nature warrants attention, highlighting how adversaries may attempt to offset the resilience advantages enjoyed by satellite mega-constellations today.

China’s strategic community has long studied anti-satellite options through kinetic, cyber and electromagnetic means. Historically, most attention has focused on directed-energy systems, ground-based jammers and kinetic interceptors such as the FY-1C test in 2007. But as commercial LEO networks proliferate and demonstrate high redundancy, researchers are now exploring distributed, persistent and lower-cost EW concepts that could theoretically degrade services across wide areas. This evolution underscores a broader shift in counter-space strategy away from single-point, high-power systems toward networked, adaptive interference architectures—mirroring the very design principles that give modern constellations their resilience.

Summary of the Recent Research:

In November 2025, a peer-reviewed paper published in Systems Engineering and Electronics described simulation-based modeling of a distributed jamming network designed to disrupt mega constellation downlink communications. Conducted by teams from Zhejiang University and the Beijing Institute of Technology, the study examined how a large fleet of airborne jammers could interfere with satellite links over an area approximating the size of Taiwan.

The researchers used publicly available orbital data to simulate Starlink satellite positions, signal behaviors, and link variability across a 12-hour period, capturing the constellation’s dynamic mesh architecture. Their model tested both narrow-beam and wide-beam jammers and evaluated how a synchronized grid of drones could generate a high noise environment. Key findings from the research included countermeasures to Starlink’s decentralized architecture via a large-scale, distributed grid of jammers that could cause meaningful degradation of constellation performance. Theoretical minimum requirements call for at least 935 jammers, but real-world operations would likely require 1,000 to 2,000 nodes to achieve meaningful effect. The researchers did not assess how the PLA might protect such a fleet from air defenses, how long it would need to operate to produce strategic impact or what logistical footprint would be required to sustain operations.

Evolution of Electronic Warfare and Anti-Satellite Applications:

While speculative, the study aligns with several ongoing trends in China’s commercial and defense ecosystem. Over the past three years, China has experienced a surge in space-related companies, largely focused on small satellites, electronic payloads and dual-use technologies.

China’s November 2025 back-to-back orbital missions further demonstrate accelerating experimentation with distributed architectures, autonomy and on-orbit maneuverability. In parallel, Chinese policymakers have openly expressed concern about foreign LEO broadband constellations enabling resilient wartime communication, reinforced by lessons from the Russia-Ukraine conflict. Chinese strategists have also emphasized the need for countermeasures that do not escalate to high-visibility kinetic anti-satellite actions, suggesting that scalable EW approaches may be an increasingly attractive area of inquiry.

The concept of targeting satellite communications via electronic interference is not new. During the Cold War, both the United States and Soviet Union developed uplink and downlink jammers to disrupt military satellites in geostationary orbits. More recent examples include the fielding of ground-based jammers capable of interfering with GPS and sites capable of localized interference against LEO links. These systems, however, traditionally rely on fixed sites or a small number of high-power emitters. The newly published Chinese research departs from that model by exploring a mega-scale, airborne and distributed EW grid mirroring the move to large scale constellations of satellites. This is significant because distributed EW mirrors the broader trend toward resiliency through redundancy; adversaries are now studying how to apply the same principles against LEO systems themselves.

Significance to the Commercial Space Industry:

For LEO broadband operators, the research highlights an emerging class of theoretical threats that differ substantially from legacy jamming models. Satellite constellations are designed to route around localized interference, drop compromised links and maintain user connectivity through multi-path redundancy. But a synchronized, region-scale jamming grid could stress that redundancy by overwhelming multiple downlink paths simultaneously.

While the scale required makes this scenario difficult to execute, its mere consideration within Chinese academic circles suggests the PLA is assessing how distributed EW could offset the inherent advantages of commercial mega-constellations. For operators supporting defense, humanitarian, maritime or government clients, this reinforces the need to invest in adaptive anti-jam waveforms, crosslink prioritization, dynamic power allocation and multi-orbit interoperability.

Ultimately, the study reinforces concerns over the evolution of electronic warfare capabilities, and signals the potential development into redundant systems capable of degrading LEO constellation performance. For the commercial space sector, particularly LEO broadband providers, it underscores that future EW challenges may arise not from isolated high-power systems but from distributed, persistent and unconventional architectures designed to exploit the very resilience features that define today’s satellite networks.

---

Briefing 40: Operation DreamJob Expands Targeting of Europe’s Aerospace Sector

11/18/2025

Overview:

On 23 October 2025, new reporting revealed another coordinated wave of intrusions linked to Lazarus Group’s long-running “Operation DreamJob,” marking the latest escalation in cyber espionage activities against Europe’s aerospace and defense sectors. Lazarus is a cluster of North Korean-based cyber actors and is assessed as one of the most prolific advanced persistent threat (APT) groups. They have sustained a years-long series of cyber operations combining targeted social engineering, supply-chain abuse and stealthy multi-stage malware delivery.

The most recent activity reinforces a trend with direct relevance to the satellite communications industry: Threat actors are increasingly blending professional networking deception with developer-tool compromise to infiltrate sensitive engineering environments.

In the campaign observed this October, Lazarus operators posed as recruiters offering lucrative positions to mid- and senior-level engineers. Victims were contacted through LinkedIn or polished email correspondence mimicking legitimate hiring processes. Once engaged, targets received PDFs posing as job descriptions or interview materials. Opening these files initiated multi-stage loader chains delivering malware families that enabled reconnaissance, data theft and persistent access. According to ESET, victims spanned a metal engineering company in Southeastern Europe, an aircraft components manufacturer in Central Europe and a Central-European defense contractor—industries whose intellectual property holds long-term strategic value for Lazarus actors.

Operation DreamJob and Social Engineering:

Operation DreamJob pairs trust-based social engineering with technically adept malware delivery. Lazarus operators build credible recruiter identities, replicate real interview workflows and distribute professionally formatted documents to establish legitimacy. When victims open the malicious PDFs, embedded scripts initiate loader execution designed to bypass common endpoint defenses. This smooth transition from human manipulation to technical compromise is a defining feature of the campaign.

Once initial execution occurs, DreamJob proceeds through several controlled stages that minimize detection and enable tailored exploitation. Dropper components first fetch additional payloads from attacker infrastructure, followed by secondary implants that conduct reconnaissance, credential capture, and movement within the network. These implants then establish persistent access through registry, service, or scheduled task modification. Finally, data exfiltration occurs through encrypted channels or covert DNS, allowing Lazarus to extract proprietary engineering data and sensitive system information. This modular, multi-stage architecture allows the operators to remain hidden for extended periods and refine payloads to match the technical environment of each victim.

Impact on the Space Sector:

Although the latest reporting centers on terrestrial aerospace entities, the operational tradecraft directly maps to risks affecting satellite communications providers, component manufacturers and space-sector integrators. In 2024, ClearSky cybersecurity reported a DreamJob campaign targeting similar industry verticals in aviation, defense and aerospace. These environments share similar characteristics: globally distributed development teams, reliance on specialized engineering roles, extensive use of open-source tooling and multi-tier supply chains spanning both IT and OT domains. As these ecosystems converge, the attack surface expands accordingly.

State sponsored APT groups have consistently pursued foreign aerospace, sensor and propulsion technologies to advance military and space capabilities, and many groups have adapted social engineering TTPs similar to DreamJob over the past several years. Satellite communications infrastructure, particularly systems supporting imaging, navigation and command-and-control, aligns closely with those priorities. As commercial space assets become increasingly intertwined with defense operations and dual-use applications, the incentive to target upstream space-sector engineering workflows continues to grow.

Conclusion:

Operation DreamJob aligns with a broader trend in which North Korean threat actors exploit global remote-work and hiring practices to gain technical access. Recent analyses from Google’s Threat Analysis Group, the FBI and international CERTs have highlighted related campaigns in which adversaries impersonate recruiters, conduct fake technical interviews and distribute test packages or repositories embedded with malware families such as BeaverTail and InvisibleFerret. These operations are supported by a rapidly evolving toolset, including Python-based variants of previously Go-based implants like GolangGhost, demonstrating sustained adaptation in both programming languages and loader architectures.

A parallel line of activity involves workers embedding themselves directly as remote contractors within Western companies using forged identities and falsified résumés. Once inside, they obtain access to source-code repositories, deployment systems, internal communication platforms and cloud-hosted development environments. Although many of these operations focus on generating revenue for the regime, they routinely create access points that enable broader espionage and supply-chain compromise. DreamJob and related IT-worker campaigns collectively demonstrate how sophisticated state-sponsored actors are actively using a combination of social engineering tactics and multi-tiered malware payloads to infiltrate the lifecycles relevant to the development and production of space systems.

---

Briefing 39: RedNovember Expands Global Espionage Campaign to Target Aerospace and Space Sectors

10/21/2025

Overview:

On September 24, 2025, researchers from Recorded Future’s Insikt Group published a report exposing new activity from a Chinese state-sponsored advanced persistent threat (APT) group tracked as RedNovember. Previously identified as TAG-100, the group has continued to evolve its tradecraft and broaden its targeting scope, conducting sophisticated cyber-espionage operations against both government and private sector entities. Recent analysis indicates that RedNovember’s interests now prominently include the defense, aerospace, and space industries, marking a significant escalation in their global operations.

Expanding Scope and Tactical Evolution:

Initially observed in mid-2024 targeting Asia-Pacific intergovernmental bodies, RedNovember leveraged open-source tools and public exploits to gain access to vulnerable networks. Since then, their campaigns have expanded geographically and strategically, targeting entities across the U.S. Defense Industrial Base (DIB) and European space organizations. The group’s methods closely align with other China-nexus APTs, including Salt Typhoon, Volt Typhoon and Silk Typhoon, which are known to exploit perimeter devices and edge infrastructure to maintain persistence and evade detection.

RedNovember has demonstrated consistent interest in exploiting popular edge devices from major vendors such as Cisco, Palo Alto Networks, SonicWall, Fortinet, F5 and Sophos. These technologies have all been impacted by widely publicized vulnerabilities in recent years, many of which were later weaponized by both state and non-state actors. By exploiting these known weaknesses, RedNovember minimizes development costs while maximizing operational impact.

The group’s exploitation of these perimeter technologies also underscores a persistent challenge in enterprise cybersecurity: the lag between public disclosure of vulnerabilities and widespread patch adoption. Their ability to capitalize on these gaps exemplifies how threat actors are effectively combining weaponized proof-of-concept (PoC) exploits with open-source post-exploitation frameworks such as Pantegana, a Go-based backdoor and Cobalt Strike. This approach reduces technical barriers for operators and enables more advanced actors to conceal their involvement by avoiding the use of bespoke malware.

Attack Pattern and Sophistication:

RedNovember’s preference for open-source and commercially available command-and-control (C2) frameworks provides a layer of deniability and complicates attribution. Their operations often blend into the background noise of legitimate red team or penetration testing activity, making detection and source attribution significantly more difficult.

This tradecraft reflects a broader shift among state-sponsored actors toward leveraging publicly available tooling. Such strategies not only obscure attribution but also reduce operational costs, allowing for sustained campaigns against multiple targets. The use of frameworks like Pantegana and Cobalt Strike further suggests an emphasis on operational agility and flexibility across global infrastructures.

Impact to the Space Sector:

RedNovember’s expanding focus on the space domain adds a critical dimension to their evolving threat profile. In April 2025, Insikt Group identified communications between a RedNovember reconnaissance and exploitation server and infrastructure tied to a European space-focused research center. The group also conducted port scanning and reconnaissance activity targeting prominent U.S. aerospace and defense organizations in July 2024, signaling a deliberate effort to map and probe critical assets within this sector.

While no confirmed exploitation resulted from those early probes, subsequent campaigns in early 2025 suggest that RedNovember transitioned from reconnaissance to active compromise attempts, particularly against organizations associated with aerospace engineering, satellite communications, and defense manufacturing.

In March 2025, for instance, Insikt Group observed a RedNovember-controlled IP address interacting with a SonicWall SSL-VPN instance belonging to a U.K.-based manufacturer specializing in bespoke cable harnessing for aerospace and defense applications. This targeting activity reflects a growing emphasis on compromising vendors integral to the aerospace and space supply chain, a recurring theme among APT operations in recent years.

Operational Overlaps and Strategic Continuity:

RedNovember’s campaigns exhibit operational overlaps with several unnamed clusters previously documented by Proofpoint, including UNK_DropPitch, UNK_FistBump, UNK_SparkyCarp and UNK_ColtCentury. These clusters share common targeting patterns, particularly against Taiwan’s semiconductor industry, and demonstrate a unified strategic objective of gathering intelligence across sectors critical to national and technological development.

Recorded Future’s earlier reporting from May 2024 linked TAG-100 (now RedNovember) to cyber-espionage operations against two prominent Asia-Pacific government bodies, providing continuity between the group’s regional focus and its newer campaigns across Europe and North America. These overlaps reinforce the assessment that RedNovember operates as part of a larger ecosystem of cyber-espionage actors, sharing infrastructure, tooling, and objectives across multiple operational clusters.

Conclusion:

RedNovember’s continued evolution underscores a key strategic trend in sophisticated cyber operations: the use of low-cost, high-efficacy methods to pursue broad intelligence-gathering objectives across sectors that underpin national security and technological dominance. By exploiting readily available exploits and tools, RedNovember achieves both scalability and plausible deniability, which are key attributes of a sophisticated threat.

For the space and defense sectors, this campaign serves as another reminder that edge devices and remote access solutions remain prime attack surfaces, particularly as organizations expand hybrid and distributed operations. The group’s consistent focus on these technologies demonstrates not only their tactical value but also their potential as gateways into highly sensitive networks.

---

Briefing 38: Shai-Hulud Supply Chain Campaign Highlights Vulnerabilities in Open-Source Ecosystems

9/23/2025

Overview:

Throughout 2025, the Node Package Manager (NPM) ecosystem has been repeatedly targeted in fast-moving supply chain attacks. Threat actors have flooded the NPM registry with malicious packages, compromised maintainers, and disguised malware within widely used dependencies. By exploiting the trust placed in open-source repositories, attackers aim to gain access to continuous integration and continuous development (CI/CD) environments. These compromises pose a serious threat to research, development and operational activities across the global space sector, where software reliability and security are critical.

Shai-Hulud Attack:

On 15 September 2025, researchers identified more than 187 malicious packages uploaded to the NPM registry as part of an ongoing supply chain campaign. The attack, dubbed Shai-Hulud, involved a self-replicating worm designed to steal developer and maintainer credentials and publish them to GitHub.

The first wave of compromises began on 14 September, when attackers trojanized the popular @ctrl/tinycolor package alongside over 40 other NPM packages. Subsequent reporting from Socket confirmed additional compromises, including multiple CrowdStrike NPM packages that were later removed. At the time of this writing, Socket is tracking over 500 affected packages.

The worm’s functionality includes harvesting developer and cloud credentials, validating them, injecting malicious GitHub Actions workflows to establish persistence and exfiltrating secrets to attacker-controlled webhooks. These tactics align with a larger trend of open-source malware and targeted maintainer compromises that undermine CI/CD pipelines—workflows critical to the commercial space industry’s ability to develop, test and deploy software.

NPM and the Open-Source Supply Chain:

NPM is both a command-line tool and an online repository for JavaScript packages. Its widespread use across development teams and automated build systems makes it a high-value target for adversaries. A single malicious update can cascade across thousands of downstream projects and CI/CD pipelines.

Attackers therefore focus on maintainers, publishing credentials and CI systems to distribute malicious code at scale. This tactic has grown sharply in recent quarters. According to Sonatype’s 2025 Open Source Malware Index Report, open-source malware increased 188% year-over-year, with exfiltration-focused payloads now the dominant type. This surge means that nearly any organization relying on open-source packages risks encountering trojanized code during its development lifecycle.

Other Recent Examples:

Beyond Shai-Hulud, several other incidents illustrate the breadth of NPM-focused activity. In late August 2025, attackers exploited GitHub Actions to steal an NPM token, which they then used to publish malicious Nx packages. This compromise exposed thousands of secrets before mitigation measures were enacted. In early to mid-September, multiple popular packages, including debug, chalk and ansi-styles, were trojanized following a targeted phishing campaign against a maintainer. The attack enabled a credential- and crypto-stealer payload with the potential to affect millions of downstream developers.

In addition, prior campaigns attributed to foreign IT worker cluster known as Contagious Interviewand the broader Lazarus APT group leveraged typosquatting and custom loaders to distribute more than 60 malicious NPM packages.

These incidents collectively demonstrate adversaries’ reliance on phishing, social engineering and MFA bypasses to compromise maintainers, followed by the abuse of legitimate tools such as TruffleHog for secrets discovery and CI automation frameworks like GitHub Actions. Attribution remains complex, with activity ranging from opportunistic, financially motivated actors to more sophisticated, state-linked operators using NPM as an infrastructure vector.

Conclusion:

Taken together, these incidents highlight recurring characteristics of NPM supply chain compromises. First, they represent a novel but increasingly common avenue for adversaries to penetrate trusted ecosystems. Second, threat actors consistently exploit social engineering and phishing to bypass MFA safeguards and seize maintainer accounts. Third, they disguise malware within widely used packages to opportunistically target CI/CD environments at scale. Finally, the focus on developer workflows underscores a strategic effort to compromise the very processes that underpin software innovation.

The repeated targeting of NPM—alongside other repositories such as GitHub and PyPI—illustrates a repeatable and scalable model for supply chain attacks. For the space sector, which relies on rapid iteration, rigorous testing and secure software deployment, these attacks pose systemic risks. As adversaries continue to refine their techniques, building resilience into CI/CD pipelines and open-source dependencies will be essential to safeguarding mission-critical research and operations.

---

Learn More About Space ISAC

Are you interested in learning more about threats to space systems? Visit our website at spaceisac.org to learn more about security for space and how to become a member.