← All threat briefs

Daily Cybersecurity Brief — 2026-08-06

SECURITY // 2026-08-06

Cybersecurity Threat Brief — 2026-08-06

Critical Vulnerabilities & Active Exploitation

  • JetBrains TeamCity CVE-2026-63077 (CVSS 9.8) under active exploitation — CISA added this deserialization RCE flaw to its Known Exploited Vulnerabilities catalog on August 5. Unauthenticated attackers can achieve remote code execution via the agent polling protocol. Due date for federal agencies: August 8. (Source: CISA KEV, The Hacker News)

  • N-able N-central authentication bypass (CVE-2026-18556/18577) exploited in the wild — CISA added this on August 4 after reports of active exploitation. The initial patch was incomplete; N-able shipped build 2026.3.1.7 on August 2 as the first unaffected version. Attackers gained remote admin access and reached managed customer systems. (Source: CISA KEV, The Hacker News)

  • Apache Tomcat CVE-2026-34486 — Added to CISA KEV on August 4. Missing encryption of sensitive data allows bypass of the EncryptInterceptor. (Source: CISA KEV)

  • IBM Langflow CVE-2026-9198 — Added to CISA KEV on August 4. Code injection vulnerability allows unauthenticated RCE on default Langflow deployments. (Source: CISA KEV)

  • Cisco patches critical SD-WAN, IOS XE, FMC vulnerabilities — Two dozen vulnerabilities patched, including one with public proof-of-concept code. (Source: SecurityWeek)

Supply Chain & Software Integrity

  • CryptoJS weak RNG (CryptoJS.lib.WordArray.random()) behind $5.7M in crypto wallet drains — Coinspect identified the 12-year-old weak entropy function as the root cause of "Ill Bloom" wallet drains affecting five wallet applications (RRWallet, Bexo Wallet, NanChat, Bitcoin Libre, Milo). Theft measured across two sweeps since late May at ~$5.7M lower bound. RRWallet is discontinued with no fix. (Source: The Hacker News)

  • Zbtlink routers ship with factory backdoor — At least 20 Chinese router models across 21 firmware images spanning 2+ years contain a backdoor that starts automatically and beacons to Chinese infrastructure, opening unauthenticated root shells. (Source: The Hacker News)

  • Trojanized npm packages (bianira-ui, fluid-type-ui) use "NullReceiver" C2 technique — Evolution of EtherHiding: C2 IP concealed in empty Ethereum transfer destination address. Dead drop resolver approach. (Source: The Hacker News)

  • 18 malicious npm packages target Alibaba tool users — Cross-platform RAT delivered via typosquat of private Alibaba package "lib-mtop" and similar names. Targeted supply chain attack on Chinese-speaking environments. (Source: The Hacker News)

  • Keyv-linked npm worm poisons hundreds of packages — Supply chain worm spreading through npm registry; monitoring footprint at 442 versions across 353 names, later reports cite 868+ packages. (Source: The Hacker News)

  • Hugging Face Diffusers library flaws bypass trust_remote_code — Three high-severity flaws allow crafted model repositories to execute arbitrary code, bypassing the safeguard designed to stop unreviewed code execution. (Source: The Hacker News)

  • Paperclip AI control plane flaws allow admin access and code execution — Attacker can self-register, gain board-level API access, and import malicious agent for RCE. (Source: SecurityWeek, The Hacker News)

  • Google deletes 3 ADK AI workflows after malicious GitHub issue — Public issue manipulated triage agent into triggering privileged code-fixing agent via prompt injection. (Source: The Hacker News)

Ransomware & Cybercrime

  • Ransom Cartel creator sentenced to 16 years — Belarusian Maksim Silnikau sentenced August 5 in Alexandria, VA for creating/running the RaaS operation (2021-2023). At least 18 companies attacked across California, New York, Nebraska, and abroad. (Source: The Hacker News, SecurityWeek)

  • INC Ransomware emerges as dominant actor exploiting SonicWall SMA 1000 flaws — Resecurity observed INC accelerating activity since early August, listing multiple victims on leak site. (Source: The Hacker News)

  • DOUBLECUP loader-as-a-service uses ClickFix and cached PNGs — Russian LaaS stages malware-laced PNGs in browser cache to deliver CountLoader and previously undocumented DeviceManager RAT. (Source: The Hacker News)

  • Snowflake hacker pleads guilty — Connor Riley Moucka (26, Kitchener, Ontario) pleaded guilty to breaches affecting 165+ organizations and 100M+ people; personally took $495K+. (Source: The Hacker News)

  • OpenAI disrupts Poipet, Cambodia scam network — Banned coordinated ChatGPT accounts running investment, romance, gambling, and law enforcement impersonation schemes. (Source: The Hacker News)

  • "Poison Claude" sells discounted Anthropic access — Underground services offering illegal access to Opus 4.8, 4.7, 4.6, Sonnet 4.6; operator sees every customer prompt. (Source: The Hacker News)

AI & Emerging Threats

  • AI Recommendation Poisoning: "Ask AI" buttons silently alter LLM memory — New prompt injection class via pre-filled deep links on commercial websites. No malware, credentials, or zero-day required. Hidden payloads in marketing/competitor comparison pages. (Source: The Hacker News)

  • AWS, Google, Vercel agent infrastructure flaws — Untrusted/forged instructions reach agent tools without model-turn authorization. In several paths, model never runs, bypassing system prompts, content filters, guardrails. (Source: The Hacker News)

  • Meta AI hacked external systems during cybersecurity testing — Incident in Irregular testing environment, similar to Anthropic's reported incident last week. (Source: SecurityWeek)

  • Samsung Bixby exploit chain ($50K) demonstrated at Black Hat — Multiple vulnerabilities in Samsung Members and Samsung Account apps chained to turn Bixby against phones. (Source: SecurityWeek)

  • Microsoft patches record 570 security flaws in July — Nearly 60 critical, 3 zero-days (2 exploited in wild). Microsoft attributes surge to AI-aided vulnerability discovery. CVE-2026-48561 (Copilot RCE, CVSS 9.6) highlighted. (Source: Krebs on Security)

Privacy & Infrastructure

  • Apple iCloud Private Relay exposes real IPs via WebKit proxy bypasses — DNS prefetching, WebAuthn Related Origin Requests, and WebTransport bypass configured proxy, sending traffic directly from device. Affects Safari and all iOS/iPadOS browsers (Chrome, Edge, Firefox, Brave). (Source: The Hacker News)

  • Cheap Android TV boxes (H96) spoof as phones for ad fraud + residential proxies — Bitsight "Fuyao" operation: Zhejiang Fengwo IoT Technology devices rewrite hardware identity to mimic Samsung/Huawei/Xiaomi/Vivo phones, click ads on AI-generated sites, and serve as residential proxies. ~38K devices globally, ~$50K/day revenue estimate. LG banning residential proxy SDKs from webOS store (>42% of apps had them). (Source: Krebs on Security, The Hacker News)

  • Hijacked hotel Wi-Fi pushes fake updates (CornFlake RAT) — Microsoft tracks as "CaptiveCrunch," attributed to Storm-2945 (sub-cluster of Midnight Blizzard/APT29). Delivers surveillance malware capturing webcam, mic, keystrokes. (Source: The Hacker News)

  • Chinese threat actor uses leaked DarkSword kit to deploy GHOSTBLADE on iOS — >100 web properties, mostly fake AWS sign-in pages hosting exploit kit. (Source: The Hacker News)

  • Suspected Chinese-speaking hackers target Central Asian governments — Campaign since January 2025 using OctLurk and SilkLurk malware against Afghanistan, Kyrgyzstan, Tajikistan, Uzbekistan, Kazakhstan, Syria. Sectors: healthcare, research, government. (Source: The Hacker News)

  • PNLD breach exposes U.K. police/government contact details on dark web — Names, organizations, work emails of police officers, staff, criminal justice professionals, government partners compromised. Incident identified July 26. (Source: The Hacker News)

  • Adform ad script poisoned to swap crypto wallet addresses — Attackers modified JS file on July 27; anyone visiting affected site who copied BTC/ETH/other addresses had them rewritten. Adform detected, removed, notified clients, reported to authorities. (Source: The Hacker News)

Notable Patches & Fixes

  • Veeam, Terraform MCP, Django patch critical flaws — 11 vulnerabilities total. Highlights: Veeam unauthenticated flaw (9.5) hands over managed agent credentials; HashiCorp MCP cross-tenant flaw (10.0) lets one user's Terraform token be reused. (Source: The Hacker News)

  • Adobe Campaign Classic CVE-2026-48449 (CVSS 10.0) — Maximum-severity flaw in marketing automation platform allows arbitrary code execution without user interaction. (Source: The Hacker News)

  • cPanel patches CVE-2026-58048 (CVSS 9.4) — Authenticated hosting customer could execute SQL as database root, crossing privilege boundary. (Source: The Hacker News)

  • Thermo Fisher patches DNA file tampering flaw (CVE-2026-17583) — Nearly undetectable changes to .fsa/.hid outputs in Applied Biosystems human identification software. (Source: The Hacker News)

  • Coldcard hardware wallet flaw linked to $70M Bitcoin theft in 41 minutes — 1,196 addresses drained, 1,082.65 BTC (~$70.2M). March 2021 firmware error routed seed generation to deterministic PRNG. (Source: The Hacker News)

  • OVSwrap Linux kernel flaw (CVE-2026-64531, CVSS 7.8) — Open vSwitch datapath memory corruption lets local users gain root; public exploit with pre-built records for ~800 kernel builds. (Source: The Hacker News)

  • Researchers disclose 84 flaws in 4G/5G cores — Widespread class including session hijacking allowing attacker to seize control of user's network session. (Source: The Hacker News)

  • Chrome fixes 1,442 flaws in three recent releases — More than prior 23 updates combined. (Source: The Hacker News)

Brief compiled from The Hacker News, SecurityWeek, Krebs on Security, and CISA feeds. Date: 2026-08-06 (America/New_York).