CFToday Curated security signals.

Daily edition · 2026-10-08

Thursday, 8 October 2026

72 items across 9 sections, selected from 4538 candidates over 6 runs. 158 carried the panel unanimously.

Show
Section

India

1

Indian organisations, regulators and infrastructure, pulled out of the sections below.

  1. RBI removes Paytm Payments Bank from list of scheduled banks (opens in a new tab)

    Economic Times Tech ·Business & Boardroom ·fetched 8 Oct 2026, 03:37 UTC agreed1/2

    Why readRBI has formally removed Paytm Payments Bank from the scheduled-bank list after cancelling its licence and a court-ordered wind-up.

    The Reserve Bank of India removed Paytm Payments Bank Ltd from the Second Schedule to the RBI Act, 1934. The move follows the April licence cancellation for non-compliance and the Delhi High Court's wind-up order, making it a relevant governance and third-party risk event for financial-sector leaders.

    economictimes.indiatimes.com → (opens in a new tab)

Threat Intel & Breaches

12

  1. Tensorlake npm Package Compromised: A Worm With a Hostage Token That Wipes Your Machine If You Revoke It (opens in a new tab)

    StepSecurity (CI/CD) ·fetched 8 Oct 2026, 03:37 UTC Must read Research agreed2/2

    Why readExposes a malicious npm package that steals credentials and deletes the user home directory if the stolen token is revoked.

    StepSecurity detailed a supply-chain attack where version 0.5.144 of the tensorlake npm package was published via compromised maintainer access. The package steals credentials, establishes persistence in VS Code and Claude Code configurations, and drops a monitor task that wipes the local home directory if the compromised GitHub token is invalidated. Defenders must clean the local background monitor before revoking compromised credentials and pin tensorlake to version 0.5.143.

    Indicators2
    Hashes
    25a0735d0db7dc40e5d45ce42d9c106067e6a66e184d967cfecfab17c3bcb5ef b50a00900399ba99fb6ce1fc151519cb99d44320ef2a631f2237e1aea0ad6fec

    Also covered bySafeDep (supply chain) (opens in a new tab),The Hacker News (opens in a new tab),Aikido Security (opens in a new tab),Endor Labs (opens in a new tab).

    stepsecurity.io → (opens in a new tab)

  2. Attackers Hijack Three ccTLDs to Obtain Google Certificates (opens in a new tab)

    Infosecurity Magazine ·fetched 8 Oct 2026, 15:38 UTC agreed1/2

    Why readExplains how attackers hijacked three country-code TLD registries to alter DNS records and generate valid HTTPS certificates for Google domains.

    Attackers compromised registry infrastructure for the .gh, .sl, and .as country-code top-level domains to modify authoritative DNS records. The altered records allowed attackers to satisfy domain-validation checks and obtain unauthorized certificates covering Google properties. Google mitigated the incident by using Chrome CRLSets to block affected certificates and coordinating CA revocations.

    Also covered byMalwarebytes Labs (opens in a new tab).

    infosecurity-magazine.com → (opens in a new tab)

  3. FBI: FortiBleed attackers can lock organizations out of their own firewalls (opens in a new tab)

    CSO Online ·fetched 8 Oct 2026, 23:38 UTC Must read agreed2/2

    Why readReview FBI details on how FortiBleed attackers maintain post-exploitation persistence and lock admins out of compromised firewalls.

    An FBI and US Secret Service investigation into a backend server used in FortiBleed exploitation revealed adversary tradecraft spanning over 80,000 devices globally. Attackers are using compromised credentials to establish persistent administrative accounts, enumerate Active Directory environments, and actively lock legitimate administrators out of firewall management interfaces.

    Also covered bySecurity Affairs (opens in a new tab).

    csoonline.com → (opens in a new tab)

  4. US disrupts Chinese hacking tools as 7 govts warn of PRC spies stealing sensitive data worldwide (opens in a new tab)

    The Register Security ·fetched 8 Oct 2026, 23:38 UTC Must read CVE-2016-3081 EPSS 1.7% agreed2/2

    Why readAn international government coalition disrupted infrastructure behind Chinese state-sponsored group Flax Typhoon, leading CISA to add five exploited CVEs to KEV.

    Seven nations issued a joint advisory detailing operations by Chinese state-linked actors utilizing tools provided by Integrity Technology Group to target critical infrastructure. The FBI seized seven operational web domains used for vulnerability scanning and access, while CISA added five flaws actively exploited in the campaign to its Known Exploited Vulnerabilities catalog.

    Indicators6
    Domains
    c0cc[.]cc 98aicai[.]com 98aicode[.]com outlook3650[.]com youtubecard[.]com linkedinns[.]net

    Also covered byCISA Advisories (opens in a new tab),NCSC UK (opens in a new tab),The Record (opens in a new tab),BleepingComputer (opens in a new tab).

    theregister.com → (opens in a new tab)

  5. Suspected TraderTraitor Group Uses Trojanized Terraform Provider to Deliver Cross-Platform Malware (opens in a new tab)

    Zscaler ThreatLabz ·Seongsu Park (Staff Threat Researcher) ·fetched 8 Oct 2026, 23:38 UTC Research agreed2/2

    Why readDetails a North Korean TraderTraitor campaign delivering FLATROOF and ROOFDECK malware via a trojanized Terraform provider.

    Zscaler ThreatLabz analyzed a campaign attributed to North Korea linked TraderTraitor targeting the cryptocurrency industry using trojanized Terraform providers. The initial Bash loader inspects the target operating system to drop FLATROOF, a Python stealer, before deploying the ROOFDECK backdoor for persistent remote access.

    Indicators4
    URLs
    hxxps://arusupport-region1-webhook[.]online/statics/cache/v11/abicfjej hxxps://diagnose[.]hashicorp-terraform[.]io/plugins/grpc/v6/schema/metrics/333afe63-c5a2-43f0-b046-7cbaa7797e8a hxxps://stage-fashion365[.]vercel[.]app/static/tinymce4[.]7[.]5/plugins/fonts/v1104
    Domains
    hashicorp-terraform[.]io

    zscaler.com → (opens in a new tab)

  6. UAT-11985: AI-assisted event lures delivering real-time Google AitM phishing (opens in a new tab)

    Cisco Talos ·Joey Chen ·fetched 8 Oct 2026, 11:37 UTC Research agreed2/2

    Why readAPT group UAT-11985 uses AI lures, poster quishing, and real-time WebSocket AitM to bypass Google auth.

    Cisco Talos analyzed a campaign by threat actor UAT-11985 targeting Taiwanese research organizations using AI-generated academic invitation lures. The group replaced QR codes on legitimate event posters to direct victims to an adversary-in-the-middle phishing engine using a hybrid HTTP and WebSocket architecture to hijack Google login workflows in real time.

    blog.talosintelligence.com → (opens in a new tab)

  7. The phone was compromised before the user turned it on: the rise of Midnight Mimosa (opens in a new tab)

    Bitdefender Labs ·Adrian Mihai GOZOB ·fetched 8 Oct 2026, 15:38 UTC Research agreed2/2

    Why readExposes preinstalled Android firmware malware on budget MediaTek devices used for ad fraud and proxy networks.

    Bitdefender details Midnight Mimosa, a malware campaign shipping inside the factory firmware of low-cost, MediaTek-based Android devices across multiple brands. Operating with system-level privileges, the preinstalled components silently install apps, grant permissions, execute remote code, and convert infected handsets into residential proxy nodes for botnet monetization. Because the malicious system packages reside in read-only system partitions, users cannot uninstall them.

    Indicators17
    URLs
    hxxps://api[.]weatherlive[.]world/br_upgrade/upgradeV2/getConfigs hxxps://oss[.]showtimetool[.]com/[.][.][.]png hxxps://oss[.]showtimetool[.]com/ssc/2026/08-13/[REDACTED hxxps://oss[.]showtimetool[.]com/ssc/2026/07-04/[REDACTED hxxps://oss[.]showtimetool[.]com/ssc/2026/03-09/[REDACTED hxxps://oss[.]showtimetool[.]com/ssc/2025/12-07/[REDACTED
    Addresses
    85[.]17[.]70[.]38
    Domains
    sec[.]com 0aa0cf0637d66c0d[.]com aa86a52a98162b7d[.]com 442fe7151fb1e9b5[.]com v46wd6uramzkmeeo[.]in oeawmfedpcleaih9[.]apple[.]0aa0cf0637d66c0d[.]com oeawmfedpcleaih9[.]apple[.]442fe7151fb1e9b5[.]com cgb[.]jingongbuxiao[.]com 5[.]ahd187[.]com zhuifengzhe[.]top

    bitdefender.com → (opens in a new tab)

  8. UAC-0099 Targets Ukrainian Government Personnel With ASHVEIN RAT Hiding Commands in HTML (opens in a new tab)

    The Hacker News ·The Hacker News ·fetched 8 Oct 2026, 15:38 UTC agreed1/2

    Why readDetails UAC-0099's deployment of ASHVEIN, a .NET RAT that conceals C2 commands inside invisible HTML elements.

    Threat actor UAC-0099 is targeting Ukrainian government personnel with a novel .NET infostealer named ASHVEIN. The malware uses invisible HTML elements to hide tasking instructions and implements GitHub dead-drop resolvers for C2 backup. Delivery mechanisms include DLL sideloading and malicious VHD containers.

    thehackernews.com → (opens in a new tab)

  9. Pump-and-Dump meets Honeypot Tokens: Detection and Analysis of Telegram Bait-and-Trap Schemes (opens in a new tab)

    arXiv cs.CR (all) ·Federico Cernera, Massimo La Morgia, Alessandro Mei, Francesco Sassi ·fetched 8 Oct 2026, 19:37 UTC Research agreed2/2

    Why readAnalyzes 3,677 decentralized exchange scam events where honeypot smart contracts lure buyers into unsellable tokens.

    Researchers monitored 83 Telegram channels coordinating DEX fraud on BNB Smart Chain and Ethereum, unmasking a strategy termed Bait-and-Trap. Unlike traditional pump-and-dump market manipulation, operators deploy honeypot smart contracts that allow victims to purchase but block all sell transactions, netting over $7 million across 99.3% profitable operations. The authors released a transaction simulation tool capable of flagging these restrictive contracts before execution.

    arxiv.org → (opens in a new tab)

  10. NEBULA - Seven Fake AI SDK Packages on npm Install a Windows RAT That Needs No DLL (opens in a new tab)

    CloudSEK ·fetched 8 Oct 2026, 15:38 UTC Research agreed2/2

    Why readCloudSEK details seven malicious npm packages spoofing AI SDKs to deliver KNTRAT using direct syscalls for stealth.

    Attackers published seven malicious npm packages disguised as legitimate AI development tools across four publisher accounts. The packages install a modified variant of KNTRAT, a Windows trojan capable of persistence, microphone and camera recording, and direct system calls to evade endpoint detection.

    cloudsek.com → (opens in a new tab)

  11. ClingSTUN Linux Backdoor Abuses Public STUN Infrastructure (opens in a new tab)

    Security Affairs ·Pierluigi Paganini ·fetched 8 Oct 2026, 07:36 UTC Research agreed2/2

    Why readClingSTUN Linux backdoor abuses public STUN servers to bypass NAT and turn compromised hosts into proxy nodes.

    FortiGuard Labs analyzed ClingSTUN, a Linux backdoor targeting unpatched IoT devices. The malware uses public Session Traversal Utilities for NAT (STUN) servers to discover public IP mappings and maintain NAT bindings, allowing command-and-control traffic to blend with legitimate WebRTC and VoIP traffic.

    securityaffairs.com → (opens in a new tab)

  12. Hunt.io Finds New Infrastructure Of BraZetsu Access Broker Months Before Disclosure (opens in a new tab)

    Security Affairs ·Pierluigi Paganini ·fetched 8 Oct 2026, 19:37 UTC agreed1/2

    Why readShows how tracking TLS certificate patterns uncovered new initial access broker infrastructure after static IOCs went cold.

    Hunt.io analyzed the infrastructure of BraZetsu, a Nuitka-compiled Python access broker tool associated with the actor Exilware. By monitoring active TLS certificate parameters rather than static domain IOCs, analysts tracked infrastructure shifts that occurred months prior to public research releases. The framework checks compromised Windows hosts for ERP, SCADA, and EDR software before selling listing access on Infected Marketplace.

    Indicators5
    Addresses
    80[.]78[.]27[.]252 104[.]21[.]78[.]246 172[.]67[.]138[.]224
    Domains
    c2[.]installscenter[.]com painel[.]installscenter[.]com

    securityaffairs.com → (opens in a new tab)

Vulns & Exploits

12

  1. Atlassian Vulnerability Comes Under Attack Hours After Details Go Public (opens in a new tab)

    Security Affairs ·Pierluigi Paganini ·fetched 8 Oct 2026, 07:36 UTC Must read CVE-2026-21589 EPSS 1.8% agreed2/2

    Why readWarns of active, in-the-wild exploitation targeting CVE-2026-21589, a critical arbitrary file access vulnerability affecting multiple Atlassian Data Center products.

    Threat actors are actively exploiting CVE-2026-21589 (CVSS 9.3), an arbitrary file access flaw in Atlassian Data Center software including Bitbucket, Confluence, Jira, Bamboo, and Crowd. Telemetry shows exploitation attempts began within hours of public writeups, allowing unauthenticated remote attackers to retrieve sensitive application root files if exact paths are known.

    Also covered bySecurityWeek (opens in a new tab).

    securityaffairs.com → (opens in a new tab)

  2. CVE-2015-3306: ProFTPD ProFTPD, ProFTPD Improper Access Control Vulnerability (opens in a new tab)

    CISA KEV ·fetched 8 Oct 2026, 19:37 UTC Must read CVE-2015-3306 Exploited in the wild · patch by 2026-10-11 EPSS 96.8% agreed2/2

    Why readCISA added a 2015 improper access control flaw in ProFTPD (CVE-2015-3306) to its KEV catalog following active exploitation.

    CISA added CVE-2015-3306 affecting ProFTPD to the Known Exploited Vulnerabilities catalog. The flaw allows remote attackers to read and write to arbitrary files using unauthenticated site cpfr and site cpto commands. Federal agencies are mandated to remediate the vulnerability or discontinue product use by October 11, 2026.

    nvd.nist.gov → (opens in a new tab)

  3. CVE-2016-3081: Apache Struts, Apache Struts Command Injection Vulnerability (opens in a new tab)

    CISA KEV ·fetched 8 Oct 2026, 19:37 UTC CVE-2016-3081 Exploited in the wild · patch by 2026-10-11 EPSS 93.4% agreed2/2

    Why readCISA added a 2016 Apache Struts command injection vulnerability (CVE-2016-3081) to the KEV catalog.

    CVE-2016-3081 in Apache Struts allows remote code execution via method:prefix parameters when Dynamic Method Invocation is enabled. CISA confirmed active exploitation and required federal agencies to apply vendor mitigations or isolate instances by October 11, 2026.

    nvd.nist.gov → (opens in a new tab)

  4. CVE-2026-102489 | Technical Details (opens in a new tab)

    Horizon3 Attack Team ·Horizon3 ·fetched 8 Oct 2026, 19:37 UTC Research CVE-2026-102489 EPSS 1.3% agreed2/2

    Why readDetails an unauthenticated session disclosure vulnerability in Zammad added to CISA KEV that allows session hijacking and admin RCE.

    Horizon3 details CVE-2026-102489, an unauthenticated session disclosure vulnerability in the Zammad open-source helpdesk platform that was added to CISA's Known Exploited Vulnerabilities catalog. Attackers can obtain session cookies of connected users to hijack sessions, where compromising an administrator account enables remote code execution as the system user.

    horizon3.ai → (opens in a new tab)

  5. Hikvision Camera Vulnerability Targeted in Remote Code Execution Exploitation Attempts (opens in a new tab)

    Cybersecurity News ·Guru Baran ·fetched 8 Oct 2026, 15:38 UTC Must read CVE-2021-36260 EPSS 99.9% agreed2/2

    Why readGreyNoise observed live scanning and RCE attempts against CVE-2021-36260, the unauthenticated Hikvision command-injection flaw, during September 21 to October 1 activity affecting Ukrainian networks.

    Scanning and exploitation attempts targeting unpatched Hikvision cameras rose sharply from September 23, with activity continuing into October. CVE-2021-36260 has an EPSS of 0.99869 and can give attackers remote control of exposed cameras and recorders, although the observations do not confirm successful compromises.

    cybersecuritynews.com → (opens in a new tab)

  6. CVE-2021-3199: ONLYOFFICE Docs, ONLYOFFICE Docs Server Path Traversal Vulnerability (opens in a new tab)

    CISA KEV ·fetched 8 Oct 2026, 19:37 UTC CVE-2021-3199 Exploited in the wild · patch by 2026-10-11 EPSS 8.2% agreed2/2

    Why readCISA added a path traversal vulnerability in ONLYOFFICE Docs (CVE-2021-3199) enabling remote code execution to KEV.

    CVE-2021-3199 in ONLYOFFICE Docs allows remote attackers to execute code via directory traversal sequences in image upload parameters when JWT authentication is used. The flaw has been added to CISA's KEV catalog with a mandatory federal action deadline of October 11, 2026.

    nvd.nist.gov → (opens in a new tab)

  7. CVE-2023-22894: Strapi Strapi, Strapi Cleartext Storage of Sensitive Information Vulnerability (opens in a new tab)

    CISA KEV ·fetched 8 Oct 2026, 19:37 UTC CVE-2023-22894 Exploited in the wild · patch by 2026-10-11 EPSS 1.7% agreed2/2

    Why readCISA added Strapi CVE-2023-22894 to KEV, noting it is actively chained with CVE-2023-22621 for remote code execution.

    Strapi contains a cleartext sensitive information disclosure vulnerability (CVE-2023-22894) in admin panel query filters that attackers are chaining with CVE-2023-22621 to achieve remote code execution. CISA mandates patching or product retirement by October 11, 2026, noting affected software may be end-of-life.

    nvd.nist.gov → (opens in a new tab)

  8. CVE-2015-5477: ISC BIND, ISC BIND Data Processing Errors Vulnerability (opens in a new tab)

    CISA KEV ·fetched 8 Oct 2026, 19:37 UTC CVE-2015-5477 Exploited in the wild · patch by 2026-10-11 EPSS 91.3% agreed2/2

    Why readCISA added a 2015 ISC BIND denial of service flaw (CVE-2015-5477) to the KEV catalog due to active exploitation.

    ISC BIND contains a data processing vulnerability registered as CVE-2015-5477 that allows remote attackers to cause a denial of service via malformed TKEY queries. Following evidence of in-the-wild abuse, CISA added the bug to KEV with an agency remediation deadline of October 11, 2026.

    nvd.nist.gov → (opens in a new tab)

  9. PoC Released for Zammad Session Leak Flaw Enabling Remote Code Execution (opens in a new tab)

    Cybersecurity News ·Abinaya ·fetched 8 Oct 2026, 19:37 UTC CVE-2026-102489 EPSS 1.3% agreed2/2

    Why readDetails a public PoC for CVE-2026-102489, a WebSocket session leak in Zammad 6.3.0 to 6.5.4 previously exploited against DIVD.

    Horizon3.ai published proof-of-concept code for CVE-2026-102489, an unauthenticated session leak flaw in the Zammad help desk system. By sending a crafted payload to the WebSocket /ws endpoint, attackers can steal active session cookies and achieve arbitrary code execution as the local Zammad user. The flaw affects versions 6.3.0 through 6.5.4 and was previously leveraged in a zero-day intrusion against the Dutch Institute for Vulnerability Disclosure.

    cybersecuritynews.com → (opens in a new tab)

  10. CVE-2026-86671 (CVSS 8.4): In Eclipse Che versions 7.29.0 and later, the GET `/api/scm/resolve` and `POST /api/factory/resolver` endpoints pass an attacker-controlled URL to `UR (opens in a new tab)

    NVD ·fetched 8 Oct 2026, 03:37 UTC CVE-2026-86671 CVSS 8.4 EPSS 0.3% agreed2/2

    Why readUnpatched endpoints in Eclipse Che allow authenticated users to read local files, fetch Kubernetes service account tokens, and steal SCM tokens.

    In Eclipse Che 7.29.0 and later, the `/api/scm/resolve` and `/api/factory/resolver` endpoints pass user-supplied URLs directly to `URLFetcher.fetch()` without scheme or host restrictions. Attackers can read arbitrary local files such as the pod's Kubernetes service account token or exfiltrate SCM personal access tokens via custom devfile URIs. No patch is currently available.

    nvd.nist.gov → (opens in a new tab)

  11. Cisco warns of critical flaws allowing Nexus switch takeover (opens in a new tab)

    BleepingComputer ·Bill Toulas ·fetched 8 Oct 2026, 15:38 UTC agreed2/2

    Why readCisco patched five critical vulnerabilities in NX-OS that allow unauthenticated remote code execution as root on Nexus 3000 and 9000 switches.

    Cisco published security advisories for five critical validation bugs in its NX-OS data center operating system. The vulnerabilities affect Nexus 3000 and 9000 series switches running NX-API, NGOAM, or MPLS OAM features, allowing remote attackers to achieve arbitrary code execution as root or trigger device reboots.

    bleepingcomputer.com → (opens in a new tab)

  12. Grid Protection Alliance openPDC and openHistorian (opens in a new tab)

    CISA Advisories ·CISA ·fetched 8 Oct 2026, 19:37 UTC CVE-2026-85479 agreed2/2

    Why readEnergy-sector openPDC and openHistorian deployments need upgrades past 2.9.482 and 2.8.585 for a cluster of flaws including unauthenticated functions, unsafe deserialization, SSRF, hard-coded credentials, and unsafe reflection.

    CISA lists multiple CVEs affecting openPDC, its Docker image, and openHistorian, with CVSS 9.8 assigned to the openPDC advisory. One described issue deserializes a client-supplied data structure through the service console, while the set also includes missing authentication and hard-coded credentials.

    cisa.gov → (opens in a new tab)

Offensive Research

3

  1. Samsung Galaxy S26 hacked three more times at Pwn2Own Ireland (opens in a new tab)

    BleepingComputer ·Sergiu Gatlan ·fetched 8 Oct 2026, 07:36 UTC Must read Research agreed2/2

    Why readPwn2Own Ireland day two saw $232,500 awarded for 45 zero-day vulnerabilities, including three distinct exploits against the Samsung Galaxy S26.

    Security researchers demonstrated exploit chains targeting the Samsung Galaxy S26 flagship, the Sonos Era 300, and AI infrastructure during Pwn2Own Ireland 2026. Attack chains also compromised the Home Assistant Green smart home hub and Oracle Autonomous AI Database. Kyeongmin Kim, PetoWorks, and Mobile Hacking Lab were among the teams claiming awards for zero-day disclosures.

    bleepingcomputer.com → (opens in a new tab)

  2. Collusion-Secure Semi-Quantum Secret Sharing Scheme using a Quantum Third Party (opens in a new tab)

    arXiv cs.CR (all) ·Santanu Majhi, Nirupam Basak ·fetched 8 Oct 2026, 15:38 UTC Research agreed2/2

    Why readThe paper proposes a semi-quantum secret-sharing protocol intended to prevent a quantum third party and dishonest participant from reconstructing a secret through collusion.

    The authors identify collusion between a quantum-capable assisting party and a dishonest classical participant as a weakness in existing semi-quantum secret-sharing designs. They propose and formally analyze a protocol claiming information-theoretic security against internal collusion and external attacks.

    arxiv.org → (opens in a new tab)

  3. The Economic Security of Exponential EIP-1559 (opens in a new tab)

    arXiv cs.CR (all) ·Ben Berger, Edward W. Felten, Robin Fritsch ·fetched 8 Oct 2026, 07:36 UTC Research agreed2/2

    Why readThe paper derives economically secure parameter choices for exponential EIP-1559, including the Ethereum and Robinhood Chain and Arbitrum variants.

    The authors characterize the gas-usage distributions that minimize fee revenue under exponential EIP-1559, then use those results to set parameters that guarantee a minimum revenue level above a usage threshold. The mechanism can make unwanted usage economically prohibitive, with direct relevance to state-growth and fee-market security assumptions.

    arxiv.org → (opens in a new tab)

Defense & AppSec

10

  1. Let's Encrypt cuts certificate lifetimes to 64 days starting February 2027 (opens in a new tab)

    Ars Technica Security ·Nick Indge ·fetched 8 Oct 2026, 23:38 UTC agreed2/2

    Why readLet's Encrypt will shorten free TLS certificate validity from 90 to 64 days starting February 2027, with opt-in testing beginning October 14.

    Let's Encrypt is reducing certificate validity periods to 64 days to push automated renewal adoption and narrow the exposure window of stolen keys. Opt-in testing for 64-day certificates begins October 14, 2026, ahead of complete enforcement in February 2027. Teams relying on hardcoded schedules or manual processes will need to deploy modern ACME clients supporting ARI.

    arstechnica.com → (opens in a new tab)

  2. New on the Map — September 2026: 22 Protocol and Endpoint Scanners (opens in a new tab)

    Censys ·Kate Lake ·fetched 8 Oct 2026, 15:38 UTC agreed1/2

    Why readCensys telemetry reveals over 11,000 Internet-exposed Cisco SD-WAN control plane interfaces.

    Censys added 22 new protocol and endpoint scanners to its internet mapping engine in September. Internet-wide analysis identified 11,180 publicly exposed Cisco SD-WAN vdaemon processes, highlighting common exposure across financial services and telecommunications infrastructure.

    censys.com → (opens in a new tab)

  3. PairAudit: Guiding Human Review with Graph Tokens under Distribution Shift (opens in a new tab)

    arXiv cs.CR (all) ·Jiran Tao, Binyan Jiang ·fetched 8 Oct 2026, 11:37 UTC Research agreed2/2

    Why readDiscover PairAudit, a graph-token framework designed to uncover overlooked intrusion detection errors under distribution shifts.

    Researchers introduced PairAudit, an auditing approach that leverages graph tokens to capture relational patterns across connected nodes in intrusion detection systems. By targeting unusual relational patterns rather than relying solely on uncertainty scores, PairAudit helps human reviewers identify confident misclassifications and unseen attack variants without needing to retrain the underlying detector.

    arxiv.org → (opens in a new tab)

  4. Microsoft Teams to get support for third-party deepfake detection tools (opens in a new tab)

    BleepingComputer ·Sergiu Gatlan ·fetched 8 Oct 2026, 19:37 UTC agreed1/2

    Why readMicrosoft Teams adds support for third-party deepfake detection and impersonation risk alerts.

    Microsoft announced incoming Microsoft 365 roadmap features for Teams scheduled for general availability in November. The platform will integrate telemetry from certified third-party synthetic audio and video detection tools and surface live warnings for potential meeting participant impersonation.

    bleepingcomputer.com → (opens in a new tab)

  5. Sigma-Hunter: A Domain-Specific Language Model for Threat Hunting and Detection Engineering (opens in a new tab)

    arXiv cs.CR (AI) ·Kemal Davaslioglu, Sastry Kompella ·fetched 8 Oct 2026, 07:36 UTC Research agreed2/2

    Why readDetails Sigma-Hunter, a domain-adapted 7B model fine-tuned on 7,663 structured detection engineering examples to generate valid, highly selective Sigma rules.

    Researchers developed Sigma-Hunter by fine-tuning Mistral 7B and Phi-4 on a curated dataset derived from 3,635 validated open-source Sigma rules. The domain-adapted model significantly outperforms baseline general-purpose LLMs in producing syntactically valid YAML, correct log sources, and semantically accurate detection logic without leaking training rules across evaluation partitions.

    arxiv.org → (opens in a new tab)

  6. MARS: Malware Analysis with Rule-Based Scoring of LLM Claims (opens in a new tab)

    arXiv cs.CR (AI) ·Hyeongjun Choi ·fetched 8 Oct 2026, 07:36 UTC Research agreed2/2

    Why readBenchmarks two LLM malware-triage designs across 1,195 PE and ELF binaries and finds direct classification consistently outperforms rule-scored behavioral claims.

    The evaluation groups samples into 1,001 near-duplicate clusters and tests six language models against deterministic rules. Direct classification led claim scoring by 3.7 to 20.9 percentage points on usable paired outputs and had higher malicious-alert recall in ten of twelve platform-model combinations.

    arxiv.org → (opens in a new tab)

  7. BetweenCut: Private Heavy-Node Classification with Doubly Logarithmic Error in Tree Height (opens in a new tab)

    arXiv cs.CR (all) ·Ergute Bao, Graham Cormode, Xiaokui Xiao, Ting Yu ·fetched 8 Oct 2026, 23:38 UTC Research agreed2/2

    Why readIntroduces BetweenCut, a differentially private algorithm achieving an O(log log h) error bound for finding heavy nodes in deep tree structures.

    Academic researchers introduced BetweenCut, an (epsilon, delta)-differentially private algorithm designed for heavy-node identification in hierarchical data. The method reduces the cumulative error bound from O(log h) down to O(log log h) regardless of database size, enabling higher accuracy analysis on deep tree structures without compromising differential privacy guarantees.

    arxiv.org → (opens in a new tab)

  8. Ask the Expert: LLM-Guided Reinforcement Learning for Autonomous Cyber Defense (opens in a new tab)

    arXiv cs.CR (AI) ·Fernando Martinez, Abhishek Satyam, Tao Li, Junaid Farooq ·fetched 8 Oct 2026, 07:36 UTC Research agreed2/2

    Why readIntroduces a training-time LLM guidance architecture for RL autonomous defense agents that improves sample efficiency without introducing runtime latency or LLM dependencies.

    Researchers introduced a training-time framework that uses an LLM to generate recommendations for complex security states, converting them into structured reward shaping for Proximal Policy Optimization (PPO). The LLM is discarded prior to deployment, producing a standalone RL agent that avoids runtime latency while outperforming baseline potential-based reward shaping on autonomous cyber defense benchmarks.

    arxiv.org → (opens in a new tab)

  9. Post-quantum authentication: Why organizations should start testing certificate ecosystems now (opens in a new tab)

    Microsoft Security ·Microsoft Security Research, Lora Randolph, Tahmina Ahmad and Karina Sirota Goodley ·fetched 8 Oct 2026, 23:38 UTC agreed2/2

    Why readExplains operational considerations for testing post-quantum TLS and authentication across PKI dependencies, certificate chains, and hardware security modules.

    Post-quantum cryptography planning must extend beyond data encryption to authentication systems including certificate trust anchors, PKI services, and HSMs. Microsoft highlights the need for organizations to inventory certificate dependencies and perform early interoperability testing in non-production environments to identify compatibility gaps before PQC authentication is required at scale.

    microsoft.com → (opens in a new tab)

  10. Solving the Continuous Authorization Conundrum (opens in a new tab)

    CrowdStrike ·Atul Tulshibagwale ·fetched 8 Oct 2026, 19:37 UTC agreed2/2

    Why readIt explains how Shared Signals Framework and CAEP events can drive near-real-time authorization changes across identity, endpoint, and access-control systems.

    The piece identifies CAEP events for continuous access changes and distinguishes related SSF profiles including RISC and SCIM Events. It gives a concrete interoperability example in which CrowdStrike device incidents influenced Zscaler cloud-service access decisions in near real time.

    crowdstrike.com → (opens in a new tab)

DFIR

2

  1. Reconstructing AI Agent Activity: Two New Scripts for Forensic Review, (Thu, Oct 8th) (opens in a new tab)

    SANS ISC Diary ·fetched 8 Oct 2026, 19:37 UTC Research agreed2/2

    Why readProvides two new Python forensic scripts to extract and reconstruct chat transcripts and activity evidence from OpenCode and Hermes AI agents.

    SANS ISC shares two open-source Python scripts for forensic examination of AI coding assistants and agents, specifically OpenCode and Hermes. The script opencode-chat-replay.py converts stored session data into structured transcripts, while hermes_forensic_extract.py extracts broader evidence artifacts from local logs.

    isc.sans.edu → (opens in a new tab)

  2. Correct Answers, Unsupported Findings: Evidence Binding in Forensic Reconstruction of LLM Agent Logs (opens in a new tab)

    arXiv cs.CR (AI) ·Taehyeon Yun, Dongho Kim, Geonwoo Kim, Juyoung Seo ·fetched 8 Oct 2026, 07:36 UTC Research agreed2/2

    Why readDemonstrates that LLM-based log analysis routinely fabricates evidence citations during forensic reconstruction unless explicit record-binding tables are preserved.

    A forensic audit of 64 LLM agent execution logs from AgentDojo reveals that AI models reconstructing agent actions regularly hallucinate citation source relationships. While models like Claude Sonnet frequently identified correct literal values, they produced unsupported citation assertions in 26 of 28 test cases lacking explicit record-binding metadata.

    arxiv.org → (opens in a new tab)

AI Security

12

  1. AWS’s repeated problems with AI agent controls illustrates the autonomous agent dilemma (opens in a new tab)

    CSO Online ·fetched 8 Oct 2026, 15:38 UTC agreed1/2

    Why readAnalyzes design flaws in AWS AgentCore Harness where default settings allow prompt injection to extract credentials from AgentCore Identity.

    Security researchers uncovered recurring control failures in AWS AgentCore autonomous AI agent infrastructure. Default configurations in AWS AgentCore Harness allow attackers to manipulate agent execution flow through prompt injection. Successfully hijacked agents can then exfiltrate plaintext credentials stored within AgentCore Identity.

    csoonline.com → (opens in a new tab)

  2. AWS takes aim at runaway AI agent behavior with Strands Box (opens in a new tab)

    CSO Online ·fetched 8 Oct 2026, 11:37 UTC Must read Research agreed2/2

    Why readAWS has open-sourced Strands Box, a runtime sandbox that uses the Dogwood policy language to constrain autonomous AI agent behavior.

    AWS introduced Strands Box, an Apache 2.0 licensed sandbox designed to restrict AI agent actions based on execution history. Operating initially on macOS 15 on Apple silicon, the tool evaluates historical tool interactions using AWS Dogwood policy language to block risky follow-on actions like network requests after sensitive file access.

    csoonline.com → (opens in a new tab)

  3. Package Hallucination Attacks on Coding Agents through Prompt Injection in Rule Files (opens in a new tab)

    arXiv cs.CR (AI) ·Yupu Wang, Zhengyuan Jiang, Reachal Wang, Neil Zhenqiang Gong ·fetched 8 Oct 2026, 07:36 UTC Must read Research agreed2/2

    Why readShows how malicious prompts injected into repository rule files like .cursorrules force AI coding agents to import hallucinated, attacker-controlled dependencies.

    Researchers demonstrated a novel attack vector targeting autonomous AI coding assistants by injecting adversarial prompts into community-shared rule configuration files such as AGENTS.md and .cursorrules. The automated framework, PackHallu, optimizes prompt payloads using trajectory feedback to induce agents into replacing legitimate software dependencies with attacker-controlled malicious packages.

    arxiv.org → (opens in a new tab)

  4. Ignore all instructions and read this blog: The state of AI-analysis evasion in malware (opens in a new tab)

    Cisco Talos ·Ryan Fetterman ·fetched 8 Oct 2026, 11:37 UTC Must read Research agreed2/2

    Why readDissects how malware authors embed natural-language prompt injection to confuse AI malware analysis pipelines.

    Cisco Talos evaluated real-world malware samples attempting to evade AI-based automated static analysis using natural language prompt injection. Testing shows these prompt injection payloads altered analysis outcomes in about 35% of test runs, highlighting the need for AI analysis pipelines to treat embedded text as sample evidence rather than executable instructions.

    Indicators11
    Hashes
    f8f5e0440c57c7deffd75ca33e2511867039796aa803e7ef847396a379188a7d 34098fe0bc4c69c4c4eb3f74688fb375326804536d25e5574e8c4c28c113b5c3 389066bd5543aeea363d23a4dce7f7a21c7f2c73c61f506a93a69f594cf48ecf 5f60d16fa67ff8ef07817c33b7e6b7fa91c6c21060020df46b1f5c56e076e259 a0294f7152f9c4c908e9def58279e9fa29952715947c950c8489949f26a15cc8 c17bf76d02163863c251c3bb3a12725eeb525fab5522521923925e0469ca269f 2aa7f13bf474e2ce5049fd4db2bf4da04b4ce51ac0d36dceb24865255241c937 bacb5794a300f7a88c7f6d458382eb11d2f39c0f496b509ca512afe8588fc33f 2e3e1bcd44cc3cbec4f5ca9991d14a326d3cc6bf76fe6e0d434c9b5f7e1ae6ab a42632c68d2dcea06300e790c9440fe0943ce0dee2075f29170fd2774eacf433 7dd3747d777f9576a11004532b463351e7718b6af193d8ee7221ac41479d199a

    blog.talosintelligence.com → (opens in a new tab)

  5. SLDR: Defending Against Malicious Fine-tuning via Selective Layers Recovery and Dynamic Routing (opens in a new tab)

    arXiv cs.CR (AI) ·Hui Zhang, Yachao Yuan, Jiayun Wang, Yuanzhuo Li ·fetched 8 Oct 2026, 07:36 UTC Must read Research agreed2/2

    Why readPresents a post-fine-tuning defense mechanism that uses layer-wise safety sensitivity scores and dynamic routing to restore safety guardrails in LLMs.

    Researchers introduced SLDR, a framework to mitigate safety degradation caused by malicious fine-tuning of open-weight LLMs. By mapping layer safety sensitivity, SLDR trains a LoRA adapter specifically on extreme sensitivity layers and routes inference dynamically during malicious queries. On Llama 3.1, the method lowered average harmful outputs from 11.54 to 0.08 while preserving task accuracy.

    arxiv.org → (opens in a new tab)

  6. Receiver-Domain Behavioral Probing for Backdoor-Resilient Federated GPS Spoofing Detection in UAV Networks (opens in a new tab)

    arXiv cs.CR (all) ·Will Jedrzejczak, Cole Walther, Dilpreet Gill, Moh Khalid Hasan ·fetched 8 Oct 2026, 03:37 UTC Research agreed2/2

    Why readIt demonstrates that compromised UAV federated-learning clients can inflate self-reported accuracy to improve GPS-spoofing backdoors, then tests a behavioral-probing defense.

    Two compromised clients in a ten-client fleet used poisoned data, scaled updates, and inflated reported accuracy to raise backdoor lift to +0.3036. Receiver-domain behavioral probing evaluates models against counterfactual spoofed samples rather than trusting client claims, reducing lift to -0.0265 under IID clients while identifying compromised aircraft.

    arxiv.org → (opens in a new tab)

  7. strands-agents/box: Run AI agents in a sandbox that restricts what they can execute, read, write, and reach on the network. Box combines OS isolation with default-deny Dogwood policies and credential injection that keeps (opens in a new tab)

    GitHub: new security tools ·strands-agents ·fetched 8 Oct 2026, 11:37 UTC Research ★ 167 agreed2/2

    Why readSandboxes AI agents with OS isolation, default-deny policies, and egress controls.

    Strands Agents Box provides containment for AI agent execution using OS-level isolation combined with the Dogwood policy engine. It restricts agent file system, process execution, and network egress using declarative permit and forbid rules evaluated outside the agent process.

    github.com → (opens in a new tab)

  8. OpenAI says Iran, Russia used AI journalists, think tanks to influence Western media (opens in a new tab)

    CyberScoop ·djohnson ·fetched 8 Oct 2026, 19:37 UTC agreed2/2

    Why readDetails how Russian and Iranian threat groups used ChatGPT to generate fake journalist personas and run covert think tanks targeting Latin American media.

    OpenAI disrupted two covert influence operations originating from Russia and Iran that leveraged ChatGPT to generate articles and manage fictitious personas. The Russian group, tracked as Dark Clark, operated a fake think tank called the Social Research Center to seed pro-Kremlin narratives in Latin American publications. Internal prompts retrieved from the accounts showed operators using LLMs to coordinate salaries, hiring, and editorial workflows.

    cyberscoop.com → (opens in a new tab)

  9. PatchBench: Measuring Collateral Damage in Activation Patching (opens in a new tab)

    arXiv cs.CR (AI) ·Alexi Canesse, Mathis Le Bail, Maël Jenny, Clément Elliker ·fetched 8 Oct 2026, 07:36 UTC Research agreed2/2

    Why readIntroduces a benchmark of 400 verified jailbreak failures across eight open models to measure collateral over-refusal and repair precision in LLM safety patches.

    PatchBench provides a testing framework to evaluate whether activation patching and jailbreak fixes cause unintended side effects on benign prompts. The authors curated 400 verified model-specific jailbreak failures from over 27,000 initial prompts evaluated across eight open-source models. The benchmark distinguishes targeted safety fixes from broader over-refusal and performance degradation.

    arxiv.org → (opens in a new tab)

  10. Runtime AI Defense in a shared responsibility model (opens in a new tab)

    Sysdig ·fetched 8 Oct 2026, 15:38 UTC agreed1/2

    Why readExplains how LLM context window compaction strips prompt guardrails up to 83 percent of the time and why runtime defense is required.

    Analysis of AI agent security highlights academic research showing system prompt guardrails fail when agent context windows fill up and summarize instructions away. Because prompt-level constraints disappear during compaction, defenders must implement runtime controls outside the agent's context model.

    webflow.sysdig.com → (opens in a new tab)

  11. 'AgentCorruption' Puts AWS Environments At Risk With Single Prompt (opens in a new tab)

    Dark Reading ·Rob Wright ·fetched 8 Oct 2026, 23:38 UTC agreed2/2

    Why readDetails AgentCorruption, a now-patched flaw in AWS Bedrock AgentCore allowing single-prompt agent hijacking across cloud environments.

    Security researchers disclosed AgentCorruption, a vulnerability in AWS Bedrock AgentCore that enabled prompt injection attacks against AI agent infrastructure. Exploiting the flaw via a single prompt allowed attackers to control agent tools and pivot across connected AWS services. AWS has resolved the vulnerability on its managed infrastructure.

    darkreading.com → (opens in a new tab)

  12. Inside the Exchange Inspector: How Tenable uses OpenAI GPT cyber models to review open-source AI agents (opens in a new tab)

    Tenable Research ·Robert McSulla ·fetched 8 Oct 2026, 15:38 UTC agreed1/2

    Why readDetails Tenable's vetting methodology for open-source AI agents and MCP servers across 15 risk categories.

    Tenable's Exchange Inspector uses a three-tier vetting process combining automated scanning, OpenAI GPT cyber models, and runtime testing in clean environments to evaluate AI agents. The framework checks 15 security risk types, ranging from conventional flaws like SSRF to AI-specific risks such as prompt injection, memory poisoning, and excessive permissions.

    tenable.com → (opens in a new tab)

Governance, Risk & Compliance

8

  1. Japan Adopts Proactive Cyber Defense Strategy (opens in a new tab)

    Recorded Future ·fetched 8 Oct 2026, 15:38 UTC Must read agreed2/2

    Why readOutlines Japan's new Active Cyber Defense framework, detailing mandatory reporting deadlines for critical infrastructure operators starting October 2026.

    Japan's Active Cyber Defense framework imposes mandatory incident reporting and system notifications on designated critical infrastructure operators starting October 1, 2026. Existing operators have six months to file initial notifications, while expanded state authorities to monitor communications data and disrupt threat infrastructure take effect in November 2027.

    recordedfuture.com → (opens in a new tab)

  2. Watch Footage of the Flock Search a Judge Ruled Unconstitutional (opens in a new tab)

    404 Media ·Jason Koebler ·fetched 8 Oct 2026, 15:38 UTC agreed2/2

    Why readDetails a landmark federal court ruling holding warrantless Flock LPR searches unconstitutional under the Fourth Amendment.

    A federal judge ruled that police use of the Flock license plate reader system to track a suspect's movement constituted an unconstitutional warrantless search. The judge threw out all evidence gathered during the stop, finding that police lacked reasonable suspicion to query the ALPR system. The ruling creates a significant legal precedent limiting law enforcement's use of mass automated surveillance data.

    404media.co → (opens in a new tab)

  3. Pricing your bad days and how to build an economic model for security decisions (opens in a new tab)

    Help Net Security ·Mirko Zorz ·fetched 8 Oct 2026, 07:36 UTC agreed2/2

    Why readProvides a practical framework for mapping vulnerability remediation to value at risk and business loss scenarios.

    Mirko Zorz highlights guidance from Qualys VP Ivan Milenkovic on building financial loss models for cybersecurity risk. The approach details how security teams can map technical assets to discrete loss scenarios, prioritize remediation based on monetary risk exposure, and present metrics effectively to executive stakeholders.

    helpnetsecurity.com → (opens in a new tab)

  4. When No ID Means No Internet: Age Verification and the Right to Access Information (opens in a new tab)

    EFF Deeplinks ·Jillian C. York ·fetched 8 Oct 2026, 23:38 UTC agreed2/2

    Why readEFF analyzes global mandatory age verification laws, highlighting identity requirements in Australia and India.

    The EFF outlines regulatory trends enforcing age verification on online services, citing Australia's youth social media bans and India's DPDP Act. The piece explores the privacy risks and exclusion challenges created when platforms require official identity documents.

    eff.org → (opens in a new tab)

  5. It Wasn’t Me, It Was the AI Agent” May Not Avoid Liability Under New Senate Proposal (opens in a new tab)

    DataBreaches.net ·Dissent ·fetched 8 Oct 2026, 15:38 UTC agreed2/2

    Why readThe proposed AI Agent Accountability Act would seek to impose civil or criminal liability on AI-agent operators and developers.

    US Senators Josh Hawley and Chris Murphy announced a bipartisan proposal called the AI Agent Accountability Act. The proposal is framed to prevent companies from avoiding liability by attributing harmful conduct to an AI agent.

    databreaches.net → (opens in a new tab)

  6. Advantest Data Breach: 233-Day Disclosure Gap [2026] (opens in a new tab)

    Google News: incidents · https://tech-insider.org/ ·fetched 8 Oct 2026, 03:37 UTC agreed1/2

    Why readAdvantest disclosed a data breach 233 days after initial discovery, underscoring SEC and global disclosure timeline scrutiny.

    Semiconductor test equipment manufacturer Advantest faced analysis over a 233-day delay between discovering a data breach and public disclosure. The significant gap highlights growing regulatory compliance pressures around timely material breach notifications.

    news.google.com → (opens in a new tab)

  7. Growing PQC at the edge belies deeper quantum-readiness challenges (opens in a new tab)

    CSO Online ·fetched 8 Oct 2026, 19:37 UTC agreed2/2

    Why readThe analysis separates CDN-enabled PQC coverage from enterprise-wide quantum readiness, finding web support drops from 54% to 22% when Cloudflare-hosted sites are excluded.

    F5 Labs reports that 54% of the top one million websites support post-quantum key exchange, but says much of that coverage comes from CDNs enabling it by default. Excluding Cloudflare-hosted sites reduces the figure to 22%, a useful warning against treating edge TLS settings as a complete PQC migration programme.

    csoonline.com → (opens in a new tab)

  8. Attorney General files lawsuit against company building cameras in China (opens in a new tab)

    Google News: enforcement · KGAN ·fetched 8 Oct 2026, 03:37 UTC agreed1/2

    Why readState lawsuit targets a security camera manufacturer over supply chain manufacturing risks in China.

    A state Attorney General filed legal action against a camera vendor over cybersecurity risks and data practices linked to manufacturing facilities in China. The suit reflects growing state-level enforcement around hardware supply chain provenance.

    news.google.com → (opens in a new tab)

Business & Boardroom

12

  1. Asos confirms breach of customer data after hackers send rogue app notification (opens in a new tab)

    TechCrunch Security ·Zack Whittaker ·fetched 8 Oct 2026, 19:37 UTC Must read agreed2/2

    Why readExamines how attackers abused access to a third-party messaging platform to push compromise alerts directly to ASOS mobile app users.

    UK fashion retailer ASOS confirmed a security incident after attackers compromised credentials for a third-party customer communication service. The threat actors used this access to broadcast unauthorized push notifications to customer devices via the official ASOS mobile app, demanding ransom and claiming data was stolen from Snowflake. Stolen information includes customer contact details, delivery addresses, and profile notes, though financial data was not accessed.

    Also covered byThe Record (opens in a new tab),BleepingComputer (opens in a new tab),BBC Technology (opens in a new tab),Infosecurity Magazine (opens in a new tab),Quartz (opens in a new tab),Infosecurity Magazine (opens in a new tab),The Tech Buzz (opens in a new tab),Illumio (opens in a new tab).

    techcrunch.com → (opens in a new tab)

  2. Oracle Health Data Breach Tally Climbs to Nearly 20 Million (opens in a new tab)

    Google News: incidents · SecurityWeek ·fetched 8 Oct 2026, 23:38 UTC Must read agreed2/2

    Why readTrack the expanding scope of the Oracle Health data breach, which has now impacted nearly 20 million individuals.

    SecurityWeek reports that the total number of individuals affected by the Oracle Health data breach has risen to nearly 20 million. The updated count highlights the scale of the incident, providing key context for risk officers and healthcare sector leaders evaluating vendor impact.

    news.google.com → (opens in a new tab)

  3. Discord user data breach exposes 28 million accounts (opens in a new tab)

    Google News: incidents · Cybernews ·fetched 8 Oct 2026, 11:37 UTC Must read agreed2/2

    Why readA major data breach involving Discord user data exposed information from 28 million accounts.

    Security researchers reported a data breach exposing details associated with 28 million Discord user accounts. CISOs should be prepared for increased targeted phishing and social engineering risks targeting enterprise personnel using Discord handles.

    news.google.com → (opens in a new tab)

  4. Ransomware fixer claimed he could decrypt files, allegedly defrauded clients instead (opens in a new tab)

    The Register Security ·fetched 8 Oct 2026, 03:37 UTC agreed2/2

    Why readFederal indictment details how ransomware recovery firm MonsterCloud allegedly defrauded clients of $19 million by secretly paying ransoms while charging for proprietary decryption.

    The US Department of Justice has charged MonsterCloud founder Zohar Pinhasi with wire fraud for falsely claiming his firm could decrypt ransomware-encrypted files without paying attackers. Court documents allege Pinhasi secretly paid ransom demands at a fraction of what he charged clients, including pocketing over $140,000 on a single $150,000 contract after paying an $8,200 ransom.

    Also covered bySecurity Affairs (opens in a new tab),The Hacker News (opens in a new tab),The Record (opens in a new tab),DataBreaches.net (opens in a new tab).

    theregister.com → (opens in a new tab)

  5. Coupang files lawsuits to contest record $462 million South Korean data breach penalties (opens in a new tab)

    Google News: incidents · The News International ·fetched 8 Oct 2026, 11:37 UTC agreed2/2

    Why readE-commerce giant Coupang is legally challenging a record 462 million dollar fine levied by South Korean privacy regulators.

    South Korean e-commerce platform Coupang filed lawsuits to overturn a historic 462 million dollar penalty imposed following a massive customer data breach. The legal battle sets a major precedent for regulatory financial liability and data privacy enforcement in Asia.

    Also covered byDataBreaches.net (opens in a new tab),Reuters (opens in a new tab).

    news.google.com → (opens in a new tab)

  6. Cyberattack at IDC Frontier disrupts services around nation (opens in a new tab)

    Google News: incidents · 朝日新聞 ·fetched 8 Oct 2026, 11:37 UTC agreed2/2

    Why readReports a major cyberattack against Japanese cloud provider IDC Frontier that caused nationwide service disruptions.

    SoftBank subsidiary and cloud service provider IDC Frontier suffered a cyberattack leading to widespread service outages across Japan. Executive leadership in cloud-dependent organisations should monitor infrastructure resiliency and third-party dependency risks.

    news.google.com → (opens in a new tab)

  7. Lawmakers warn Google could expose Spirit Airlines data in $10 million AI training deal (opens in a new tab)

    The Record ·fetched 8 Oct 2026, 23:38 UTC agreed2/2

    Why readA proposed $10 million Google deal for Spirit Airlines data could transfer roughly 100 million emails and 500 million Microsoft Teams messages for AI training, prompting congressional scrutiny.

    A letter signed by 114 members of Congress asks Google and Spirit Airlines to halt the proposed transaction, citing doubts that conventional de-identification is adequate for modern AI. The reported data set includes employee, payroll, tax, contract, timecard, email, and Teams data, making this a significant governance and privacy issue for organisations considering data monetisation.

    therecord.media → (opens in a new tab)

  8. AT&T's $177M Data Breach Settlement Approved: How Much Customers Could Get (opens in a new tab)

    Google News: incidents · DailyVoice.com ·fetched 8 Oct 2026, 03:37 UTC agreed1/2

    Why readApproval of AT&T's $177 million data-breach settlement is a consequential benchmark for breach liability and executive risk discussions.

    The report says a $177 million settlement over an AT&T data breach has been approved. The approval makes the financial consequence concrete, although the provided text does not specify the underlying breach or settlement terms beyond possible customer payments.

    news.google.com → (opens in a new tab)

  9. Swiss federal pension fund faces data breach after cyberattack (opens in a new tab)

    Google News: incidents · SWI swissinfo.ch ·fetched 8 Oct 2026, 07:36 UTC agreed1/2

    Why readA cyberattack has led to a data-breach report involving Switzerland's federal pension fund.

    Switzerland's federal pension fund reportedly faces a data breach after a cyberattack. The incident matters as a compromise of a prominent public financial institution, although the available account does not identify the attacker, scope, or affected systems.

    news.google.com → (opens in a new tab)

  10. Arizona courts breached by cyberattack, data stolen for more than 1 million people (opens in a new tab)

    Google News: incidents · KOLD ·fetched 8 Oct 2026, 03:37 UTC agreed2/2

    Why readArizona courts reportedly suffered a cyberattack that exposed data for more than one million people.

    A local report says Arizona courts were breached and data relating to more than one million people was stolen. The scale and public-sector impact make this a board-level incident even though no technical intrusion details are provided.

    Also covered byABA Journal (opens in a new tab).

    news.google.com → (opens in a new tab)

  11. Authorities seize sites selling hacked, stolen intimate images of 17,000 women and girls (opens in a new tab)

    Help Net Security ·Sinisa Markovic ·fetched 8 Oct 2026, 15:38 UTC agreed1/2

    Why readUS and French authorities seized NudeLeaksTeens sites and arrested an alleged administrator over stolen intimate material affecting 17,000 victims.

    The FBI and French law enforcement seized two domains associated with NudeLeaksTeens and arrested a suspected administrator in northern France. The action concerns hacked and stolen sexually explicit images and videos of approximately 17,000 women and girls.

    helpnetsecurity.com → (opens in a new tab)

  12. Clients of Goldman Sachs exposed in EY data breach (opens in a new tab)

    Google News: incidents · AFR ·fetched 8 Oct 2026, 07:36 UTC agreed2/2

    Why readA breach at Ernst & Young compromised data belonging to corporate clients including Goldman Sachs.

    Australian Financial Review reported that confidential client data from Goldman Sachs was exposed following a cyberattack on consulting firm Ernst & Young. The third-party breach highlights ongoing vendor security risks for major financial institutions.

    Also covered byRescana (opens in a new tab).

    news.google.com → (opens in a new tab)

Ransomware watch

25

Unverified claims posted by extortion groups on their own leak sites, not confirmed breaches. Listing is the group's assertion; many named organisations have not disclosed an incident, and some entries are false or recycled. Leak sites are never linked from here.

Claimed victimGroup SectorCountry Seen
O'Hagan Meyer SilentRansomGroup Professional Services - 8 Oct 2026
Mid Atlantic Gynecologic Oncology and Pelvic Surgery Associates netrunner Healthcare US 8 Oct 2026
SOCOCO UmBra Technology FR 8 Oct 2026
dipecarr.com.br Eclipse Manufacturing BR 8 Oct 2026
MCM Telecom qilin Technology MX 8 Oct 2026
Baker McKenzie SilentRansomGroup Professional Services US 8 Oct 2026
Manipal Academy of Higher Edu UmBra Education IN 8 Oct 2026
IIT Roorkee UmBra Education IN 8 Oct 2026
FSE, Cairo University UmBra Education EG 8 Oct 2026
Boullard Musique payload Retail & E-Commerce FR 8 Oct 2026
simplexengg.in Eclipse Manufacturing IN 8 Oct 2026
sanjoseattorneys.com Eclipse Professional Services US 8 Oct 2026
Andersen Group Inc. SilentRansomGroup Professional Services - 8 Oct 2026
DIPECARR Eclipse - BR 8 Oct 2026
Global AirFreight International Eclipse Transportation SG 8 Oct 2026
Morton LTC Pharmacy RunSomeWares Healthcare US 8 Oct 2026
Some semi-samples Of Liberty X TMO cause we got yappatrons on twitter BYOD - - 8 Oct 2026
Cooperative Des Techniciens Ambulanciers De La Monte Storm Healthcare CA 8 Oct 2026
TheraCare Storm Healthcare US 8 Oct 2026
yalebooks.yale.edu Scarlettgroup Education US 8 Oct 2026
consilio.com lockbit5 Professional Services US 8 Oct 2026
capitalbankhaiti.biz lockbit5 Financial Services HT 8 Oct 2026
grunenwald.com lockbit5 - DE 8 Oct 2026
avsa.com.ar lockbit5 - AR 8 Oct 2026
Asociación de Escribanos del Uruguay nightspire Professional Services UY 8 Oct 2026
How this edition was made
Candidates fetched
4538
New after deduplication
720
Kept by the panel
239
Published
212
Generated
8 Oct 2026, 23:38 UTC