📊 WEEK IN INTELLIGENCE — 7–13 JUN 2026

BLUF

The week’s defining story is the convergence of three simultaneous supply-chain and authentication-layer compromises — the 400+ Arch Linux AUR package hijacking deploying eBPF rootkits, a China-linked PAM/login backdoor that persisted undetected for nearly a decade, and Handala’s claimed breach of California Water Service with exfiltrated OT credentials — arriving in the same week that internal network telemetry confirmed active lateral movement on at least one monitored environment. The through-line is not coincidence: adversaries at every tier, from nation-state APTs to Iranian hacktivists to opportunistic supply-chain actors, are targeting authentication infrastructure and trusted software delivery mechanisms simultaneously. Defenders who have not audited their software supply chains, Linux authentication stacks, and internal east-west traffic this week are operating blind.

ESCALATIONS

Iranian Hacktivist Activity — Critical Infrastructure Handala, the Iranian-linked hacktivist group with a track record of targeting Israeli and Western infrastructure, claims a successful intrusion into California Water Service (Cal Water), one of the largest investor-owned water utilities in the United States. The group published approximately 5GB of exfiltrated data, including customer PII and — more critically — credentials for RTKBase, an NTRIP/GNSS reference station platform used in OT and surveying contexts. Scope of actual OT network access remains unconfirmed; Cal Water has not publicly validated the breach. Even if OT penetration is limited, credential exposure for any system adjacent to operational technology in water infrastructure represents a material escalation. Southern California water infrastructure should be treated as a target environment until Cal Water provides definitive scope assessment. [SecurityWeek, MODERATE CONFIDENCE]

AUR Supply Chain Compromise — Active Exploitation More than 400 Arch Linux AUR (Arch User Repository) packages were confirmed hijacked this week, with payloads including an eBPF-based rootkit operating at kernel level and an infostealer targeting credentials and session tokens. This is not a theoretical supply-chain risk — active exploitation is confirmed. The eBPF rootkit is particularly dangerous because it operates below the visibility horizon of most commercial EDR solutions; standard endpoint telemetry cannot be trusted on affected hosts. Any environment running Arch Linux, or pulling AUR packages in CI/CD pipelines, must be treated as compromised until verified clean. The breadth of the compromise — 400+ packages — suggests either a coordinated campaign or a compromised maintainer account with broad repository access. [BleepingComputer, The Hacker News, HIGH CONFIDENCE]

China-Linked PAM Backdoor — Nine-Year Persistence A China-linked APT was confirmed to have backdoored Linux PAM (Pluggable Authentication Modules) and login software, maintaining persistence for approximately nine years before discovery. The implant targets the authentication stack directly — harvesting credentials at the point of login, making it invisible to network-layer monitoring and largely invisible to application-layer logging. The full scope of affected distributions has not been enumerated. Nine years of dwell time means this actor had access to whatever those systems touched for nearly a decade; the downstream compromise surface is effectively incalculable. This is the week’s most strategically significant long-term intelligence finding. [The Hacker News, HIGH CONFIDENCE]

Active Lateral Movement — Internal Network Internal IPS telemetry flagged host 192.168.1.65 conducting rapid port scanning against 192.168.1.10 — 8 ports in 60 seconds, classified as lateral movement. The IPS detected but did not block the traffic, meaning connections may have succeeded. Origin of compromise on the source host is unknown. This is a confirmed post-exploitation indicator: an adversary or malware already inside the perimeter is conducting internal reconnaissance. Both hosts must be treated as compromised pending forensic review. [Internal IPS telemetry, HIGH CONFIDENCE]

Splunk Enterprise — Unauthenticated RCE A critical remote code execution vulnerability in Splunk Enterprise was disclosed this week, exploitable without authentication. Splunk is frequently deployed as a security monitoring platform, meaning successful exploitation doesn’t just compromise a server — it potentially compromises the integrity of the security monitoring infrastructure itself. An attacker who owns your SIEM owns your visibility. Patch status and active exploitation status should be confirmed immediately. [The Hacker News]

Google Sues Chinese Smishing Network — AI-Assisted Phishing Google filed suit against a Chinese smishing network accused of leveraging Gemini AI to generate phishing content at scale. This is the first confirmed legal action against an AI-assisted phishing operation and signals that AI-generated social engineering has crossed from theoretical concern to operational reality at sufficient scale to warrant civil litigation. The use of Gemini specifically — a Google product weaponized against Google’s own users — is notable. [The Hacker News]

Silent Ransom Group — Continued Activity Silent Ransom Group (SRG), a threat actor specializing in callback phishing and social engineering-led extortion without traditional ransomware deployment, remained active this week. SRG’s model — impersonating IT support, inducing victims to install remote access tools — continues to evade email security controls because the initial lure is often a phone call, not a malicious link. [Graham Cluley]

RESOLUTIONS

Conti Ransomware — Guilty Plea A Ukrainian national pleaded guilty this week to his role in the Conti ransomware operation, representing continued law enforcement progress against one of the most prolific ransomware groups of the early 2020s. Conti itself has been defunct as an organized entity since 2022, but prosecutions of former members serve both deterrence and intelligence functions — plea agreements typically include cooperation provisions. [BleepingComputer]

Ex-School District Employee — Jailed A former school district employee was sentenced to prison for hacking attacks against their former employer. Insider threat prosecution and sentencing. Noteworthy primarily as a reminder that terminated employee access revocation remains a persistent failure mode across sectors. [BleepingComputer]

phpBB — Auth Bypass Patched phpBB patched an authentication bypass vulnerability that had been present in the codebase for approximately a decade. While phpBB’s deployment footprint is smaller than it was at peak, it remains in use across community forums and some legacy enterprise deployments. Patch is available; apply immediately if phpBB is in your environment. [BleepingComputer]

NPM 12 — Supply Chain Hardening NPM announced that version 12 will change default script execution behavior: npm install will no longer automatically execute scripts from dependencies. This is a meaningful structural improvement to JavaScript supply chain security, addressing a long-standing attack vector used in numerous high-profile supply chain compromises. Not a resolution to an active incident, but a significant defensive architectural change to a critical piece of global software infrastructure. [SecurityWeek]

Maine Data Breach Portal — Disabled After Abuse Maine disabled its data breach notification portal after discovering it was being used to submit fraudulent breach disclosures. This is a process integrity failure rather than a security resolution — the portal was taken offline as a containment measure. The underlying problem (abuse of public-facing regulatory infrastructure for disinformation or competitive harm) remains unresolved. [BleepingComputer]

TRENDS

Authentication Infrastructure as Primary Target Three of this week’s most significant items — the China-linked PAM backdoor, the AUR eBPF rootkit (which harvests credentials and session tokens), and the Handala RTKBase credential exfiltration — all target authentication infrastructure. This is not coincidental. Authentication is the chokepoint: own the auth stack and you own everything downstream without needing to exploit individual applications. Defenders should treat authentication infrastructure (PAM, LDAP, Active Directory, SSO providers, credential stores) as the highest-value target class and audit accordingly.

Supply Chain as Preferred Initial Access Vector The AUR compromise joins a pattern that has been building for years but accelerated sharply in 2025-2026: trusted software repositories as attack vectors. The AUR incident, the NPM ecosystem’s ongoing challenges (hence the NPM 12 behavioral change), and the PAM backdoor (which required access to or trust within the Linux software supply chain) all reflect the same adversary logic — compromise the distribution mechanism rather than the target directly. Defenders who trust package repositories implicitly without integrity verification are operating on a broken assumption.

eBPF as EDR Evasion Technique — Maturing The use of eBPF for rootkit deployment in the AUR compromise represents a maturation of a technique that security researchers have been warning about for several years. eBPF operates at kernel level with broad system visibility and limited oversight from userspace security tools. Its legitimate uses (observability, networking, security tooling itself) make it difficult to block categorically. Expect eBPF-based malware to become more common as the technique proliferates from advanced threat actors to commodity malware authors.

AI-Assisted Offensive Operations — Crossing Thresholds The Google lawsuit against a Chinese smishing network using Gemini AI, combined with ongoing reporting on AI-generated phishing content, marks a threshold crossing: AI-assisted offensive operations are no longer a future concern. They are current operational reality at scale sufficient to generate civil litigation. The same week that the U.S. government ordered Anthropic to restrict foreign national access to Fable 5 and Mythos 5 models underscores that AI export controls and access restrictions are now active policy levers, not theoretical ones.

Long Dwell Times — Detection Failure Pattern The nine-year PAM backdoor and the decade-long phpBB auth bypass both reflect a pattern of extended undetected persistence. These are not outliers — they are representative of a broader detection failure across the industry. Threat actors who achieve initial access and then operate quietly, without triggering behavioral anomalies, can persist for years. Mean time to detection remains catastrophically high for sophisticated implants. This week’s internal lateral movement alert is a reminder that detection is possible — but only if you’re looking.

PATCH STATUS SUMMARY

Note: CISA added one new vulnerability to the Known Exploited Vulnerabilities catalog this week; specific CVE identifier was not fully captured in source material. Verify against KEV catalog directly.

WATCH LIST (NEXT WEEK)

1. Cal Water / Handala — OT Access Confirmation. The critical unknown is whether Handala’s access extended beyond customer PII into actual OT networks via the RTKBase credentials. Cal Water’s official response and any CISA/WaterISAC advisories in the coming days will either confirm a contained data breach or escalate this to an active OT intrusion with implications for Southern California water infrastructure. Watch for ICS-CERT advisories.

2. AUR Compromise Scope Expansion. 400+ packages is a large number, but the full list of compromised packages has not been definitively published. CI/CD pipelines pulling AUR packages in development environments represent a secondary infection vector that may not be apparent until next week’s incident reports start arriving. Watch for enterprise security teams reporting downstream compromise from development environment infection.

3. Anthropic Export Control Fallout. The U.S. government’s order to restrict foreign national access to Fable 5 and Mythos 5 is the first confirmed instance of AI model access controls being imposed via government directive rather than voluntary company policy. Watch for diplomatic responses, legal challenges, and whether other frontier AI providers receive similar orders — this may be the opening move of a broader AI export control regime.

4. Splunk RCE — Active Exploitation. Unauthenticated RCE in a widely deployed SIEM platform is a high-value target for ransomware operators and APTs alike. The window between public disclosure and active exploitation for this class of vulnerability is typically measured in days. Watch for exploitation reports and CISA KEV addition.

5. Internal Lateral Movement — Root Cause. The 192.168.1.65 scanning event remains unattributed. The origin of compromise on the source host is unknown. If forensic analysis of that host reveals a known malware family or TTPs consistent with a tracked threat actor, it changes the scope of the incident significantly. Watch for forensic findings and whether additional hosts show anomalous east-west traffic.

ASSESSMENT

This week represents a convergence of threats that individually would be serious and collectively constitute a significant stress test for any security program. The common thread is trust: trust in software repositories (AUR), trust in authentication infrastructure (PAM), trust in vendor platforms (RTKBase, Splunk), and trust in internal network traffic. Adversaries at every sophistication tier have identified that attacking trusted systems and trusted delivery mechanisms is more efficient than attacking hardened perimeters directly. The nine-year PAM backdoor is the most extreme illustration of this logic — an implant so well-positioned that it didn’t need to do anything aggressive; it simply waited at the authentication chokepoint and collected everything that passed through. Defenders who have not explicitly audited their trust assumptions — which packages they pull, which authentication libraries they run, which internal hosts they consider clean — are carrying unquantified risk.

The lateral movement detection on the internal network is the week’s most operationally urgent item for any organization running a similar flat internal network architecture. The IPS detected but did not block the traffic, which means the question is not whether reconnaissance occurred but whether it succeeded. The origin of compromise on 192.168.1.65 remains unknown, which means the initial access vector — and therefore the full scope of the intrusion — is uncharacterized. Until forensics on that host are complete, the correct posture is to assume the worst: that the host was compromised via one of this week’s active exploitation vectors (AUR packages in a CI/CD pipeline, a vulnerable SDK, or a phishing-led initial access), and that the scanning activity represents an adversary mapping the internal network for further exploitation. The absence of a known root cause is not reassurance — it is the most concerning data point of the week.

Strategically, the AI export control action against Anthropic deserves more attention than it has received in the security press. The U.S. government’s directive to restrict foreign national access to specific frontier AI models is a significant policy escalation — it treats advanced AI models as controlled munitions-adjacent technology rather than commercial software. This has downstream implications for how AI-assisted security tooling is deployed in multinational environments, how AI providers structure access controls, and how adversary nations respond. The same week that Google sued a Chinese network for weaponizing Gemini AI, the U.S. government moved to restrict foreign access to competing models. The AI security threat landscape and the AI policy landscape are now moving in parallel, and security teams that have not begun thinking about AI governance as a security function — not just a compliance function — are behind the curve.

Sources: SecurityWeek, BleepingComputer, The Hacker News, AWS Security Bulletins, CISA Current Activity, Unit 42 (Palo Alto Networks), Graham Cluley, SentinelOne Labs, Rapid7, Internal IPS telemetry. Confidence levels noted inline where applicable. This summary reflects intelligence available as of 13 Jun 2026 17:00 PT.