AI Weekly Malaysia

Rapid7 examined Linux backdoors used against telecom and network appliances in South Korea and Taiwan that disguise themselves as email security products rather than just reusing binary names. The South Korea samples include a new BPFDoor variant and a BPF Rekoobe build that impersonate the PID file of SpamSniper, a Korean anti-spam product, and rotate through ten Linux daemon names; a previously unreported implant called AVERAT was delivered by a dropper to Taiwanese appliances. The activity is linked to the threat group Red Menshen, which has targeted telecom providers across the Middle East and Asia since 2021, and Rapid7 notes the operators retooled after vendors wrote static Suricata/Snort signatures for earlier Layer 4 anomalies.

Why: The concrete lesson is that these implants hide behind process and PID-file names, so any Linux host monitoring that allowlists by process name or treats a familiar-looking PID file as trusted is the exact weakness being exploited. It also shows the signature arms race in practice: once Suricata/Snort static rules caught the older traffic pattern, the operators changed tooling. If you do not run SpamSniper or ShareTech appliances or telecom-grade network gear, this is threat intelligence rather than something you must patch today.

The Hacker News reports (dated 2026-10-08) that a tool called ARTEX, described in the headline as an AI pentesting tool, was used in data theft attacks against South Korean financial firms. The excerpt supplied here is truncated to the article's navigation and header markup, so no technical detail is available: no named victims, no attack timeline, no data volumes, no attribution, and no description of how ARTEX actually works.

Why: There is not enough in this text to justify a decision — no victim names, no IOCs, no tooling detail, no indication of whether ARTEX is a repurposed commercial pentest product or custom malware. The only practical action supported by the source is to treat 'AI pentesting tool' in a headline as an unverified label until the full report is read; if you run offensive-security or AI-agent tooling in a financial-adjacent environment, wait for the primary technical write-up before changing anything.

ANY.RUN's analysis of 'Wazza' describes a phishkit hitting banking, manufacturing, and government organizations in the US, Europe, and Australia, using a multi-stage routing chain instead of a single static lure page. Traffic hits a wildcard landing domain (boegl-krysl[.]eu) which calls /api/wazza-config to check whether the hostname belongs to an active campaign, contacts beacon-surge-sync[...]workers[.]dev to issue a client marker, then mints a short-lived signed session token via /api/mint-token. Only after a token and browser-telemetry check at check[.]boegl-krysl[.]eu passes does the visitor reach an Adobe-themed Device Code phishing page.

Why: Two concrete things worth acting on. First, the final lure is OAuth device-code phishing, which targets the device authorization grant designed for input-limited devices - if your product or your staff SSO tenant (Microsoft, Adobe, Google) permits device code flow, a password plus MFA does not stop this, so the decision is whether to block or restrict that grant in conditional access. Second, the payload sits behind a signed-token anti-bot gate and campaign-prefix check, so a single URL reputation scan of the initial link may never surface the phishing page; only detonation-style analysis does. There is no Malaysia-specific angle in this text.

OpenAI says it banned two influence operations over the past two and a half years — one Russia-origin, one Iran-origin — that combined its models with conventional tactics to build "false front" entities. The Iranian operation ran seven "journalist" personas that pitched long-form articles to small and medium outlets worldwide and also mass-generated social media comments on US-Iran war topics; the Russian operation co-opted unwitting people in Latin America to run a ground-level "think tank" and produced fake "leaked" documents and audio scripts. OpenAI rates the Russia-origin operation as Category 5 on the IO Breakout Scale, which it calls a first, and notes both operations leaned on AI heavily to draft internal reports while using questionable methodologies to exaggerate their own effectiveness.

Why: The distribution channel here is not an exploit — it's a pitch email to small and medium outlets, which means any site you run that accepts contributed or guest articles is a candidate target with zero technical barrier. If your product or community publishes third-party bylines, this is a prompt to decide what verification you require before publishing, because the personas are indistinguishable from normal contributors by design. Note what you cannot do with this post: OpenAI publishes no indicators, no detection heuristics, and no model-level detail, so there is no operational action beyond contributor vetting.