# Cybersecurity news this week: SSHole, AI attack agents and two supply-chain breaches Published: 2026-11-09 · By HackerGPT Team Canonical: https://hackergpt.org/blog/cybersecurity-news-week-november-9-2026/ > The security news of 2–9 November 2026 that still matters: an OpenSSH bypass attacked the day after its fix, seven additions to CISA's exploited list, AI agents attacking on their own, and two supply-chain breaches. What happened, and what to check on your own systems. **TL;DR:** The most urgent item of 2–9 November 2026 is SSHole (CVE-2026-123158), an OpenSSH login bypass on Alpine and other musl-based systems that attackers went after the day after its fix and that is now on CISA's exploited list; update `openssh` and restart sshd, or set `PasswordAuthentication no`. Six more actively exploited bugs joined that list this week, including SonicWall SSL-VPN and WordPress core. The wider story is AI on offense: Check Point found criminals running an AI pentest platform that attacks mostly on its own, and OpenAI reported 948 Linux root exploits produced by its own pipeline. This roundup is for people who run a website, a server or a container pipeline and want to know which of this week's news applies to them. For each story you get what happened, according to the people who found it, and the check to run on your own systems. Every check below is for systems you own or have written permission to test. ## What do I need to patch this week? Start with CISA's [Known Exploited Vulnerabilities (KEV) catalog](https://www.cisa.gov/known-exploited-vulnerabilities-catalog). A bug lands there only when CISA has evidence of exploitation in the wild, so it is the shortest honest list of what attackers are using right now. We pulled the catalog's JSON feed ourselves on 9 November at 21:34 UTC instead of relying on summaries: ```bash curl -s https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json \ | jq -r '"catalog \(.catalogVersion), \(.count) entries", (.vulnerabilities[] | select(.dateAdded >= "2026-11-02") | "\(.dateAdded) due \(.dueDate) \(.cveID) \(.vendorProject) \(.product)")' ``` ```text catalog 2026.11.09, 1773 entries 2026-11-09 due 2026-11-12 CVE-2026-102895 GPUStack GPUStack 2026-11-09 due 2026-11-12 CVE-2026-123158 OpenBSD OpenSSH 2026-11-09 due 2026-11-12 CVE-2026-116779 SignalWire FreeSwitch 2026-11-09 due 2026-11-12 CVE-2026-123218 IBM Aspera Faspex 2026-11-06 due 2026-11-20 CVE-2026-19868 SonicWall SonicOS SSLVPN 2026-11-04 due 2026-11-18 CVE-2024-27834 Apple Multiple Products 2026-11-03 due 2026-11-06 CVE-2026-119977 WordPress WordPress Core ``` The due date is the deadline for US federal agencies, but it is a useful clock for everyone: a three-day deadline means CISA considers the risk immediate. Two older entries fall due on 10 November: the Brotli decoder bug CVE-2026-111497, which we covered in our [HTTP2Shell brief](https://hackergpt.org/blog/http2shell-brotli-cve-2026-111497/), and an Adobe Commerce and Magento file-upload flaw, CVE-2026-48356. If you are new to the vocabulary, our glossary explains [what a CVE is](https://hackergpt.org/learn/cve/) and [how to read a CVSS score](https://hackergpt.org/learn/cvss/). ### SonicWall SSL-VPN: MFA that a password can skip [SonicWall's advisory SNWLID-2026-0014](https://psirt.global.sonicwall.com/vuln-detail/SNWLID-2026-0014) describes CVE-2026-19868: when SAML is enabled for SSL-VPN logins, valid credentials for another authentication method (LDAP, RADIUS, TACACS+) get a user in without going through SAML, and so without the MFA that SAML enforces. Stolen or sprayed passwords become VPN access. Gen7 is fixed in 7.3.4-7013 and later, Gen8 in 8.2.3-8015 and later; the 9 November revision adds a workaround, disabling the LDAP, RADIUS and TACACS+ servers. CISA listed it on 6 November. ### WordPress core: the upgrade window itself CVE-2026-119977 is a race in WordPress's full-package core upgrade: staging paths are predictable and reachable from the web, so an unauthenticated attacker can run code while an upgrade is in progress, as [Picus Security, who found it, explains](https://www.picussecurity.com/resource/blog/cve-2026-119977-when-updating-wordpress-becomes-the-vulnerability). It is fixed in [WordPress 7.1.4, released 28 October](https://wordpress.org/news/2026/10/wordpress-7-1-4/), with backports to every branch back to 4.7. Wordfence saw attempts the day of disclosure and [more than 240,000 blocked attempts by 5 November](https://www.wordfence.com/blog/2026/11/attackers-actively-probing-for-upgrade-window-rce-vulnerability-in-wordpress-core/), with no confirmed compromise reported. Check your version with WP-CLI: ```bash wp core version ``` Anything below 7.1.4, or below the patched release of your branch (7.0.8, 6.9.11, 6.8.12 and so on), needs the update. ## Is my server affected by SSHole, and how do I check? You are affected if you run OpenSSH's sshd built against musl libc without PAM, with password authentication on. In practice that means Alpine Linux's stock `openssh-server` package and container images built on it. glibc distributions such as Debian, Ubuntu and RHEL, and Alpine's `openssh-server-pam` with `UsePAM yes`, are not affected, according to [Red Hat](https://access.redhat.com/security/cve/cve-2026-123158) and the [Alpine advisory](https://alpinelinux.org/posts/2026-11-05-openssh-password-authentication-bypass/). ### How the bypass works Accounts that have no password, written as `*` in `/etc/shadow`, cannot be logged into with any password. To hide which accounts exist, sshd still hashes the password a client sends, borrowing the salt of another account. The [disclosure on oss-security](https://www.openwall.com/lists/oss-security/2026/11/05/3) shows what goes wrong on musl: with a password of roughly 257 to 1,024 bytes, musl's `crypt()` fails and returns `*`, which then equals the stored `*`, and sshd accepts the login. It needs another account on the machine with a SHA-256 or SHA-512 (`$5$` or `$6$`) hash. Alpine ships `root:*`, but its default `PermitRootLogin prohibit-password` blocks direct root logins; root is in reach only with `PermitRootLogin yes`, or through passwordless `doas` from an account that is exposed. The researcher's own write-up is at [sshhole.com](https://sshhole.com/). ![Timeline of SSHole from the OpenSSH fix on 5 November to the CISA KEV listing on 9 November 2026](https://hackergpt.org/blog/cybersecurity-news-week-november-9-2026/sshole-timeline.webp) The [OpenSSH 10.6p2 release](https://www.openssh.org/releasenotes.html) fixed it on 5 November. [Previdian](https://previdian.com/CVE-2026-123158) saw the first exploitation activity on 6 November, and [GreyNoise](https://sshole.events.greynoise.io/) recorded the first successful exploitation at 00:46 UTC on 7 November. That is a one-day gap between the public fix and attacks: the patch itself shows attackers where to look. ### The version number that matters is your package's The fix is "OpenSSH 10.6p2", but most Alpine systems will never run 10.6p2: Alpine backported the fix into every supported branch. We read [Alpine's security database](https://security.alpinelinux.org/vuln/CVE-2026-123158) directly to get the fixed package version per branch: ```bash for b in v3.21 v3.22 v3.23 v3.24 edge; do printf '%-6s ' "$b" curl -s "https://secdb.alpinelinux.org/$b/main.json" \ | jq -r '.packages[].pkg | select(.name=="openssh") | .secfixes | to_entries[] | select(.value | index("CVE-2026-123158")) | .key' done ``` ```text v3.21 9.9_p2-r1 v3.22 10.0_p1-r11 v3.23 10.2_p1-r1 v3.24 10.3_p1-r2 edge 10.6_p2-r0 ``` So an Alpine 3.24 host showing OpenSSH 10.3 is fixed if the package release is `-r2` or later, and an upstream-version check would wrongly flag it. That is the general lesson: compare against your distribution's fixed package, not the upstream release number. On a host or inside an image you own: ```bash apk info -v openssh-server # installed package and release sshd -T | grep -iE 'passwordauthentication|permitrootlogin|usepam' # effective config, as root awk -F: '$2 == "*" { print $1 }' /etc/shadow # accounts with no password set ``` Upgrade with `apk upgrade openssh-server` and restart sshd; Alpine notes that the running daemon keeps the old code until then. If you cannot upgrade today, `PasswordAuthentication no` closes the hole. Do not stop at hosts: base images in your CI pipeline and registry carry the same package, and an image built last week still ships the vulnerable release. ## What else was patched that I might run? These are not on CISA's list yet, but each has a vendor fix and a population worth checking. - **APT on Debian and Ubuntu (CVE-2026-109518).** When APT retries a download, an on-path attacker, malicious mirror or proxy can get it to install a package other than the verified one, and its scripts run as root. It needs `Acquire::Retries` above zero, the default since Debian 12 and Ubuntu 22.04. Fixed in [Debian DSA-6607-1](https://lists.debian.org/debian-security-announce/2026/msg00526.html) and [Ubuntu USN-8996-1](https://ubuntu.com/security/notices/USN-8996-1); `apt-cache policy apt` shows what you have. - **Zyxel GS1900 switches (CVE-2026-107441).** An out-of-bounds write that lets an attacker on the LAN run commands, fixed in firmware 2.90(xxxx.3)C0 per [Zyxel's advisory](https://www.zyxel.com/global/en/support/security-advisories/zyxel-security-advisory-for-out-of-bounds-write-vulnerability-in-gs1900-series-switches-11-03-2026). [Binarly](https://www.binarly.io/blog/zyxel-gs1900-rce-cve-2026-107441) found the attackers' exploit and showed it targets this bug. That is the exploit behind the 996 switches [GreyNoise reported compromised](https://www.greynoise.io/blog/open-season-on-kapibala-attacker-steals-government-records-wordpress-exploitation) in September, which GreyNoise had attributed to an older, June fix: switches patched in June were still exposed. - **Barracuda Email Security Gateway (CVE-2026-116606 and CVE-2026-111560).** An unauthenticated authorization bypass plus a command injection that, chained, give root on ESG below 9.4.0.030 when its admin interface is reachable, per [Barracuda](https://trust.barracuda.com/security/pre-authentication-authorization-bypass-command-injection). No exploitation has been reported. - **Android (CVE-2026-58868).** A critical remote code execution in the Framework component that needs no user interaction, on Android 14 to 17, fixed at the 2026-11-01 patch level in the [November bulletin](https://source.android.com/docs/security/bulletin/2026/2026-11-01). ## What did AI attack agents do this week? They moved from demos to operations. Three reports from different organisations, published within days of each other, describe AI agents doing offensive work with little human input. **ARTEX, an AI pentest platform used for crime.** [Check Point Research](https://research.checkpoint.com/2026/hacking-on-autopilot-inside-artex-cybercrime-campaigns/) found 2,557 systems that host or hosted panels of ARTEX, a Chinese open-source AI pentesting platform, 1,608 of them live and public at the time of research. Most of the attacks it analysed ran fully autonomously. The yield was real but smaller than the agents claimed: more than 190,000 sets of game-account credentials, of which Check Point estimates about 6,000 were valid, and more than 300 iOS signing certificates, of which three validated. The agents also reported cash-outs that never happened. [GitGuardian](https://blog.gitguardian.com/the-agents-have-now-found-your-secrets-on-github-artex-is-abusing-public-secrets-right-now/) reports ARTEX users harvesting secrets leaked on public GitHub and reaching Azure tenants in as little as five minutes. The lesson for site owners is old and now urgent: a key in a repository or a [JavaScript bundle](https://hackergpt.org/blog/hardcoded-credentials-javascript-bundle/) is found by machines, fast. **948 Linux root exploits from one pipeline.** [OpenAI reported](https://openai.com/index/a-thousand-fewer-paths-to-root/) on 6 November that its pipeline turned 2,308 candidate kernel findings into 1,169 working local privilege-escalation exploits covering 948 distinct bugs, for $1,213,430 in API credits. That works out to about $1,280 per distinct bug, end to end. The exploits need local access, the numbers are OpenAI's own, and the fixes are going to maintainers through its Patch the Planet programme, the same programme behind the audit that found [PHP2Shell](https://hackergpt.org/blog/php2shell-cve-2026-109618-109619/). What changes is the price of a working exploit, not the kind of bug. **One operator, a mesh of agents.** [Acronis's Threat Research Unit](https://www.acronis.com/tru/posts/bossai-inside-the-ai-attack-mesh-that-targeted-taiwans-emergency-services) describes "BOSSAI", a single operator running twelve configured AI agents, nine of them active, against hundreds of organisations in 11 countries over about 12 days, including emergency services, banks and universities; Acronis confirmed access at about ten of them. India's CERT-In issued a critical advisory, [CIAD-2026-0058](https://www.cert-in.org.in/s2cMainServlet?pageid=PUBVLNOTES02&VLCODE=CIAD-2026-0058), on "Crafty Falcon", an actor attacking banks and payment firms that enters through test environments, writes scripts on the fly with AI help and returns to victims whose remediation was incomplete. And [Bitdefender](https://bitdefender.com/en-us/blog/businessinsights/3am-ransomware-hybrid-ai-attack) traced a 3AM ransomware attack on a UK school estate that encrypted about 37 of 47 machines; Bitdefender's working explanation is a language model reading each command's result and retrying with different commands. The defensive read is the same across all three: attackers no longer need to be skilled or patient to work through a long list of exposed services. The test, staging and admin endpoints you forgot about are what an agent finds first, which is why [mapping your own attack surface](https://hackergpt.org/blog/map-web-app-attack-surface/) is worth doing before someone else does. ## Which supply-chain attacks should I know about? Two incidents this week put malicious code above the customer's own controls: in a security product's update channel and in device firmware. **cPGuard.** On 2 November at 02:06 UTC, a malicious update reached servers running the cPGuard security product, versions 5.88.12 to 5.88.19, and ran as root on about 2,950 of them, according to [OPSSHIELD's disclosure](https://opsshield.com/notice/security-incident-disclosure-servermanage.html). The attacker had registered a customer account under a false identity, gained command execution on the billing and licensing server, and stolen the key to the update storage. The script installed a hidden root shell and collected `/etc/shadow`, SSH keys, database and S3 credentials and shell history; about 7,900 billing-portal accounts were also exposed. Version 5.88.26 removes the shell and a follow-up update removes a leftover cron task, per the [vendor's notice](https://opsshield.com/notice/cpguard-update-channel-compromise.html). Updating is not enough: treat every credential on an affected server as stolen and rotate it. **XCPE.** [Google's Threat Intelligence Group](https://cloud.google.com/blog/topics/threat-intelligence/shutting-the-door-on-the-xcpe-botnet) dismantled a botnet of more than 1.3 million Android set-top boxes whose backdoor was preloaded in firmware, attributing it with high confidence to the manufacturer Videostrong, which sold access to criminal affiliates. The same firmware turned up in telehealth devices, karaoke machines and digital signage. Google banned the company and its future affiliates from certified Android, and says taking down the command servers cuts access to at least 10 million more compromised devices. That disrupts the botnet; it does not clean the devices. Two smaller items target people like this site's readers. [Doctor Web](https://news.drweb.com/show/?i=15311&lng=en) found more than 2,000 PCs infected through booby-trapped GitHub repositories posing as security tools, with malicious code in Visual Studio project files, Python and Go; the payload steals data and tries to remove ESET, McAfee and Malwarebytes. Read a tool's build files before you run it. And [XLab](https://blog.xlab.qianxin.com/cwpdoor-the-invisible-backdoor-living-inside-control-web-panel/) found CWPdoor, a trojanized `cwpsrv` for Control Web Panel that waits for signed commands in HTTP headers and keeps those requests out of its logs. ## Which breaches and incidents mattered? - **Binance.** The 29 October theft is the largest in crypto history: [Elliptic](https://www.elliptic.co/insights/binance-largest-hack-in-history/) puts it at $6.43 billion, more than four times the 2025 Bybit theft, and reports the FBI's attribution to North Korea's TraderTraitor group; [TRM Labs](https://www.trmlabs.com/resources/blog/north-korea-steals-usd-6-36-billion-from-binance-in-the-largest-hack-in-history) counts $6.36 billion across ten chains, most of it moved in about 20 minutes. Elliptic reports $2.85 billion frozen or seized. - **DigiCert.** In its [incident report to Mozilla](https://bugzilla.mozilla.org/show_bug.cgi?id=2084460), DigiCert says an internal system moved into an externally reachable zone in February 2025 stayed open to self-registered accounts for about 16 months. One read-only account pulled archived customer contact data, and identity documents for a small minority, between 27 June and 2 July 2026; DigiCert confirmed it on 19 October. - **Ukraine's Air Alert app.** On 1 November, false alerts went out nationwide, including radiation and chemical warnings. The [Ministry of Digital Transformation](https://t.me/s/mintsyfra) said the app had likely been subjected to enemy influence and that there was no radiation threat; the volunteer tracker [alerts.in.ua](https://blog.alerts.in.ua/p/false-alerts-november-1) counted 122 false alerts in seven minutes. Alerting systems are now a target in their own right. - **Queens Public Library.** A ransomware attack took down the [library's](https://connect.queenslibrary.org/) website, app and public computers; its locations stayed open. ## What else happened that teaches something? - **APT29 on phones.** [Sekoia](https://www.sekoia.com/blog/poisoned-source-apt29-watering-holes-targets-mobile-devices-with-n-day-exploits) attributes, with moderate to high confidence, a watering-hole campaign that injected scripts into legitimate sites, including through a week-long compromise of the Plerdy analytics tracker, and served n-day exploits to iOS and Android visitors without any click. Third-party scripts on your site are part of your attack surface. - **15,579 infected websites.** [Scott Helme](https://scotthelme.co.uk/we-know-15-579-websites-are-infected-how-do-we-tell-them/) found 15,579 infected sites with more than 135 million visits a month between them, serving card skimmers, ClickFix lures, iOS exploit kits and redirects; a 9 November update raised the count to 16,068. A Content Security Policy is one of the [security headers](https://hackergpt.org/learn/security-headers/) that limits what an injected script can load. - **Flowise and "Greeter".** [Era Security](https://era.security/blog/greeter-flowise-ai-infrastructure-attack) found 4,372 internet-exposed Flowise AI servers, 703 of them without authentication, and an attack framework exploiting CVE-2024-58351 that sends its code inside each request, so the saved chatflows look clean when you inspect them. - **AI infrastructure bugs.** [RunSybil](https://www.runsybil.com/blog/cpython-0-day-vulnerabilities-threaten-the-ai-supply-chain) showed a CPython bug that lets a crafted chat template in a model download escape the Jinja sandbox and run code in vLLM; the [fix](https://github.com/python/cpython/pull/155761) is in 3.13, 3.14 and 3.15 as an ordinary bug fix, with no CVE, and not in 3.12 or older. [JFrog](https://jfrog.com/blog/reachy-http-request-hijacks-ai-robot/) chained bugs in the Reachy Mini robot to root and motor control, including from a malicious web page. - **Default passwords still work.** [Sam Curry and Shubham Shah](https://samcurry.net/forgerock) reached a ForgeRock identity console at two carmakers through a path-confusion bypass, logged in with the default `admin:admin`, and could have unlocked and started customers' cars; they published it on 7 November. Both carmakers fixed it within about two hours of the report. ## Key takeaways - Update Alpine's `openssh-server` to your branch's fixed release and restart sshd, or set `PasswordAuthentication no`; check container base images, not only hosts. - Compare installed packages against your distribution's fixed release, not the upstream version number: backports make the upstream number misleading. - Treat CISA's KEV catalog as your weekly patch list; a three-day due date means exploitation is under way. - Assume leaked keys and forgotten test or admin endpoints are found by automated agents within minutes, and remove them first. - After a supply-chain compromise, updating removes the implant but not the theft: rotate every credential the compromised software could read. ## FAQ ### What is SSHole, and am I affected? SSHole (CVE-2026-123158) is a password-authentication bypass in OpenSSH's sshd when it is built against musl libc without PAM, as in Alpine Linux's stock `openssh-server`. With password authentication on, an attacker can log into an account that has no password set by sending an overlong password. glibc distributions such as Debian, Ubuntu and RHEL are not affected. Update to your distribution's fixed package and restart sshd. ### What is the CISA KEV catalog, and why should a small team care? It is CISA's list of vulnerabilities with evidence of exploitation in the wild, with a remediation deadline for US federal agencies. A small team cannot patch everything at once, and KEV is a short, evidence-based list of what attackers are using right now. Checking it weekly against what you run is a cheap way to set priorities. ### Are AI agents really attacking sites on their own now? Yes, within limits. Check Point found that most ARTEX attacks it analysed ran without a human steering them, and Acronis described one operator running several AI agents against hundreds of organisations. The agents still make mistakes and overstate their success, but they work through exposed services, leaked secrets and known bugs faster than a person would. ### I ran cPGuard 5.88.12 to 5.88.19. What should I do? Update to 5.88.26 or later and apply the follow-up update that removes the leftover cron task, as OPSSHIELD describes. Then rotate everything the script could read: root and user passwords, SSH keys, database credentials and S3 keys. Check the server for other changes made after 2 November. ### Is my WordPress site safe after an automatic update? It is fixed once it runs 7.1.4 or the patched release of its branch, such as 7.0.8 or 6.9.11; `wp core version` or the Dashboard's Updates page shows which you have. The bug is exploitable while an upgrade runs and attacks began on 28 October, so if your site upgraded in the days after that, check its files and admin accounts for changes you did not make. ## Sources - [CISA Known Exploited Vulnerabilities catalog](https://www.cisa.gov/known-exploited-vulnerabilities-catalog) (feed read 9 November 2026, catalog 2026.11.09) - [OpenSSH release notes](https://www.openssh.org/releasenotes.html) and [oss-security disclosure of CVE-2026-123158](https://www.openwall.com/lists/oss-security/2026/11/05/3) - [Alpine Linux advisory](https://alpinelinux.org/posts/2026-11-05-openssh-password-authentication-bypass/) and [Alpine security tracker](https://security.alpinelinux.org/vuln/CVE-2026-123158) - [Red Hat: CVE-2026-123158](https://access.redhat.com/security/cve/cve-2026-123158) - [sshhole.com](https://sshhole.com/), [Previdian](https://previdian.com/CVE-2026-123158), [GreyNoise SSHole tracker](https://sshole.events.greynoise.io/) - [SonicWall SNWLID-2026-0014](https://psirt.global.sonicwall.com/vuln-detail/SNWLID-2026-0014) - [WordPress 7.1.4](https://wordpress.org/news/2026/10/wordpress-7-1-4/), [Picus Security](https://www.picussecurity.com/resource/blog/cve-2026-119977-when-updating-wordpress-becomes-the-vulnerability), [Wordfence](https://www.wordfence.com/blog/2026/11/attackers-actively-probing-for-upgrade-window-rce-vulnerability-in-wordpress-core/) - [Debian DSA-6607-1](https://lists.debian.org/debian-security-announce/2026/msg00526.html), [Ubuntu USN-8996-1](https://ubuntu.com/security/notices/USN-8996-1) - [Zyxel advisory](https://www.zyxel.com/global/en/support/security-advisories/zyxel-security-advisory-for-out-of-bounds-write-vulnerability-in-gs1900-series-switches-11-03-2026), [Binarly](https://www.binarly.io/blog/zyxel-gs1900-rce-cve-2026-107441), [GreyNoise](https://www.greynoise.io/blog/open-season-on-kapibala-attacker-steals-government-records-wordpress-exploitation) - [Barracuda advisory](https://trust.barracuda.com/security/pre-authentication-authorization-bypass-command-injection) - [Android Security Bulletin, November 2026](https://source.android.com/docs/security/bulletin/2026/2026-11-01) - [Check Point Research: ARTEX](https://research.checkpoint.com/2026/hacking-on-autopilot-inside-artex-cybercrime-campaigns/), [GitGuardian](https://blog.gitguardian.com/the-agents-have-now-found-your-secrets-on-github-artex-is-abusing-public-secrets-right-now/) - [OpenAI: A thousand fewer paths to root](https://openai.com/index/a-thousand-fewer-paths-to-root/) - [Acronis TRU: BOSSAI](https://www.acronis.com/tru/posts/bossai-inside-the-ai-attack-mesh-that-targeted-taiwans-emergency-services), [CERT-In CIAD-2026-0058](https://www.cert-in.org.in/s2cMainServlet?pageid=PUBVLNOTES02&VLCODE=CIAD-2026-0058), [Bitdefender: 3AM](https://bitdefender.com/en-us/blog/businessinsights/3am-ransomware-hybrid-ai-attack) - [OPSSHIELD disclosure](https://opsshield.com/notice/security-incident-disclosure-servermanage.html) and [notice](https://opsshield.com/notice/cpguard-update-channel-compromise.html) - [Google Threat Intelligence: XCPE](https://cloud.google.com/blog/topics/threat-intelligence/shutting-the-door-on-the-xcpe-botnet), [Doctor Web](https://news.drweb.com/show/?i=15311&lng=en), [XLab: CWPdoor](https://blog.xlab.qianxin.com/cwpdoor-the-invisible-backdoor-living-inside-control-web-panel/) - [Elliptic](https://www.elliptic.co/insights/binance-largest-hack-in-history/), [TRM Labs](https://www.trmlabs.com/resources/blog/north-korea-steals-usd-6-36-billion-from-binance-in-the-largest-hack-in-history) - [DigiCert incident report](https://bugzilla.mozilla.org/show_bug.cgi?id=2084460), [Ministry of Digital Transformation of Ukraine](https://t.me/s/mintsyfra), [alerts.in.ua](https://blog.alerts.in.ua/p/false-alerts-november-1), [Queens Public Library](https://connect.queenslibrary.org/) - [Sekoia: Poisoned Source](https://www.sekoia.com/blog/poisoned-source-apt29-watering-holes-targets-mobile-devices-with-n-day-exploits), [Scott Helme](https://scotthelme.co.uk/we-know-15-579-websites-are-infected-how-do-we-tell-them/), [Era Security](https://era.security/blog/greeter-flowise-ai-infrastructure-attack) - [RunSybil](https://www.runsybil.com/blog/cpython-0-day-vulnerabilities-threaten-the-ai-supply-chain), [CPython fix](https://github.com/python/cpython/pull/155761), [JFrog](https://jfrog.com/blog/reachy-http-request-hijacks-ai-robot/), [Sam Curry: ForgeRock](https://samcurry.net/forgerock) Want to practise finding exposed admin consoles and leaked keys legally? Try it on [CyberGym](https://cybergym.hackergpt.app), our deliberately vulnerable practice target.