A now-patched vulnerability in AWS Bedrock AgentCore could have allowed attackers to take over every agent in an AWS account and region with a single chatbot prompt. Tamir Ishay Sharbat of Zenity Labs, an AI security vendor, revealed the flaw — dubbed “AgentCorruption” — at SecTor 2026 in Toronto.

The issue stemmed from agents running inside Firecracker MicroVMs without proper network isolation, letting attackers query AWS’s Instance Metadata Service (IMDS) for temporary credentials. From there, broad default permissions meant attackers could invoke other agents, read sessions, and raid AWS Secrets Manager.

AWS patched the flaw in February, enforcing IMDSv2 on all new agents and stripping permissions from default roles — but disputes Zenity’s framing, telling Dark Reading the research recasts documented, intended behavior as a vulnerability. Zenity is now investigating whether similar weaknesses exist on other cloud platforms, and Sharbat says the IMDS problem isn’t specific to AWS. Keep each agent’s role narrow, he says, and anyone who reaches IMDS finds a smaller blast radius.

Source: Dark Reading

ASOS has admitted that hackers stole far more customer data than it initially disclosed — and only after BBC News told the retailer the criminals had sent it a sample of the stolen data. Names, addresses, phone numbers, emails, customer numbers, dates of birth, and even search histories are now in criminal hands. When the hack surfaced on October 6, 2026, ASOS had said only basic contact details may have been accessed.

ASOS says the breach began when hackers impersonated a trusted contact to obtain an employee’s login credentials. The group, calling themselves Xuanyewen (also reported as Xuanye Group), told the BBC they reached the data through Simon AI, a platform built on Snowflake — a claim ASOS has not confirmed, and Snowflake says its own platform was not breached.

No passwords or bank details were taken, and ASOS says customers don’t need to take any action. But experts warn them to watch for convincing phishing attempts and impersonation scams using their stolen personal details.

Source: BBC News

A cyberattack on Arizona's court system compromised personal information for 1.3 million people, believed to have started when a court employee clicked a malicious link in an email. Those 1.3 million were referred to the court's FARE debt collection program over unpaid fees, fines, and restitution for traffic and criminal violations going back 30 years. The copied records include names and Social Security numbers.

The attackers also took nearly 30,000 active and inactive orders of protection, along with 150,000 reports dating back to 2010 from the Foster Care Review Board, which advises in cases where parents are alleged to be unfit or unable to care for a child. They did not take information about jurors, witnesses, or court employees.

Court technology staff shut the attack down on a backup server about two hours after spotting it on September 24. No records were altered or deleted, and the court says the data came off a highly compressed backup, so how much of it can actually be read is unclear. The FBI is investigating.

The court has notified those affected by text and email and is urging them to freeze their credit. Anyone holding a protective order does not need to refile — the orders remain valid. Court spokesperson Alberto Rodriguez said there's no evidence the data has been used or shared, and that no court cases have been affected or delayed.

Source: SecurityWeek

ClickFix, the social engineering trick that fools users into pasting malicious commands into PowerShell, Windows Run, or the macOS Terminal, is evolving to dodge detection. Two new campaigns show attackers rebuilding the moment where a pasted command turns into a payload.

Flare researcher Assaf Morag detailed a campaign spreading CrocoRat — a remote access Trojan and crypto stealer. The victim still pastes a command from the clipboard, but that command now queries a DNS TXT record on an attacker-controlled server, and the record hands back the next PowerShell instruction. The next stage never touches the clipboard, so defenders have fewer forensics to work with.

Flare also found an unexecuted Python launcher in the package, built to choose its payload by environment: a quiet, persistent foothold on systems that look corporate, and the full RAT plus credential and crypto stealers on systems that look personal. The operator appears to be limiting theft on corporate targets to avoid tripping detection.

Microsoft Threat Intelligence flagged a separate cluster of compromised websites that pre-fetch a script into the browser cache, disguised as a PNG, before the fake prompt ever appears. That hides the payload and gets around the Run dialog's character limit. Morag's advice goes after the step itself: alert on clipboard-to-Run patterns, harden PowerShell script-block logging, stop interpreters running from writable directories, and train users on the tells — fake CAPTCHAs, press-Windows+R prompts, and verification-failed screens

Source: Dark Reading

ASOS customers woke up on Tuesday to a startling push notification — apparently sent by hackers who have compromised the fashion retailer's systems. The message, addressed to ASOS's data protection officer and IT team, claimed the attackers had "fully compromised the Snowflake instance" and threatened to leak the data unless the company engaged. App users in Australia, France, Sweden and Ireland received it too.

A group calling itself Xuanye Group took credit, linking to a freshly created Telegram channel. Dan Bird of security firm Horizon3 pointed out that sending a push notification needs access to ASOS's notification system, which is separate from Snowflake — so if both claims hold up, the attackers had credentials opening more than one door.

Snowflake says its own investigation has so far found no compromise of its platform. ASOS confirmed the unauthorised notification on Tuesday afternoon, said it immediately restricted access to the notification platforms, and told customers some basic personal information may have been accessed — but not payment cards or account passwords.

ASOS shares fell around a tenth on Tuesday. The company has not yet reported a breach to the UK's Information Commissioner's Office, and the National Cyber Security Centre has offered assistance. How many customers received the notification is still unclear.

Source: BBC News

A fresh zero-day vulnerability is hitting Citrix NetScaler appliances — even ones administrators had just patched days earlier. Tracked as CVE-2026-88779, the high-severity memory overflow flaw affects NetScaler ADC and Gateway instances configured as SAML SP or IdP, causing Denial of Service through repeated crashes. Those earlier patches covered CVE-2026-88771 and CVE-2026-88772, fixed the week before.

Security researcher Kevin Beaumont, who named this one PitScaler 2, confirmed seeing exploitation attempts against patched honeypots, and reported one honeypot running a downloaded malware binary. Admins reported reboots alongside logs showing shell commands buried in username fields. One admin who retrieved the script said it tries to plant web shells, survive reboots and upload configuration and backups — while cautioning there was no proof it ever ran.

Security firm watchTowr has since reproduced the flaw and found it can only crash systems, with no route to code execution. The firm suspects attackers crashed machines deliberately to make exploitation of CVE-2026-88771 faster. Citrix says the same thing from its side: service availability is affected, customer data integrity is not.

CISA added the flaw to its KEV catalog on October 4, giving federal agencies until October 7 to remediate. It's the sixth exploited NetScaler vulnerability CISA has cataloged in 2026.

Source: SecurityWeek

A newly established ransomware group called BYOD claims to have stolen personal data from Trump Mobile customers, posting a file to its dark web leak site last week containing 3,615 records — names, addresses, phone numbers, emails, and order details. Straight Arrow News reported it on October 5, 2026, after reviewing the data and calling people listed in it.

Several confirmed their details were accurate. Several others confirmed the same, then denied being Trump Mobile customers and hung up — so the file holds up for some entries without standing up all 3,615. Among those listed is Eric Brunnett, vice president and chief information officer at the Trump Organization. Straight Arrow found no Trump family members in the data.

BYOD says it got in by infecting an employee at Florida-based Liberty Mobile with malware. The malware family, the infection method and the access path remain unestablished, and nothing in the reporting shows ransomware encrypted any systems. Trump Mobile is owned by T1 Mobile and licenses its branding from the Trump Organization.

BYOD also claims it still has access to Trump Mobile's backend dashboard, backing that up with a screenshot of customer data. Whether that access is live is the open question. Trump Mobile had not commented at publication time. No passwords or card numbers have been shown as exposed, but customers should watch for phishing attempts and verify any unexpected outreach through official channels.

Source: Cybersecurity News

Saif al-Din Khader, known as "Rey" and believed to lead the ShinyHunters extortion group, has been arrested in Jordan and is reportedly cooperating with the FBI to identify other members. Jordan confirmed the arrest without naming him — the identification comes from Reuters sources, who place the detention on September 29. The FBI has declined to comment on any arrest in Jordan.

Khader, a teenager from Amman, is also linked to the Scattered Lapsus$ Hunters group, where Brian Krebs placed him as a key member when he first named him in November 2025. Some reporting calls him ShinyHunters' leader; most call him a suspected member. Charges and extradition haven't been disclosed.

The arrest follows ShinyHunters hacking and defacing the FBI's jobs site, FBIJobs.gov, claiming to have stolen 2-3 terabytes of data including personal and health information of all current and former FBI employees. Last week, Dutch police arrested a 24-year-old Amsterdam man in the same investigation.

Dutch police and the FBI both declined to name the Amsterdam suspect, though independent reports point to Pepijn van der Stap, a convicted hacker on supervised release. The FBI says the man it arrested took part in hacking more than 140 organizations and collected at least $70 million in extortion payments. Krebs reported that Khader apparently tried to pin the FBI hack on van der Stap, planting a defacement image of Umbreon, van der Stap's old nickname. Patel says more arrests are "on the table."

Source: SecurityWeek

Citrix has pushed emergency patches for a zero-day vulnerability in NetScaler ADC and Gateway appliances that attackers are already exploiting. Tracked as CVE-2026-88779 with a CVSS 4.0 score of 8.7, the memory overflow flaw hits systems configured as SAML service providers or identity providers, triggering denial of service through repeated crashes. No credentials or user interaction are needed to exploit it — just network access, and attack complexity is low.

Citrix confirmed targeted attacks and says the impact is availability only, with no sign of data theft. Its CVSS vector agrees, rating confidentiality and integrity impact at none. But the question isn't closed. Administrators logged authentication requests carrying shell commands, Kevin Beaumont saw a malware binary running on a patched honeypot, and watchTowr reproduced the flaw within hours. Nobody has confirmed code execution, and nobody has ruled it out.

Either way, the clock is running. CISA added the flaw to its Known Exploited Vulnerabilities catalog on October 4 and gave federal agencies until October 7 to fix it, with forensic triage required under BOD 26-04 to establish whether an appliance was hit before the patch went on. To find out if you're in scope, check the configuration for add authentication samlAction or add authentication samlIdPProfile — either entry meets the precondition.

Administrators who patched for the two earlier NetScaler zero-days, CVE-2026-88771 and CVE-2026-88772, must upgrade again. Go to 14.1-73.41 or 13.1-64.28 on the standard branches, 14.1-73.41 FIPS for FIPS appliances, or 13.1-37.282 for 13.1 FIPS and NDcPP. Citrix-managed cloud services are covered already. Global Deny Lists buy time; patching ends it.

Source: Cyber Security News

GitLab has pushed urgent security updates to fix a critical vulnerability in its AI Gateway component, tracked as CVE-2026-90970 with a near-perfect CVSS 3.1 score of 9.9. The flaw lets authenticated users with Duo Agent Platform access submit crafted flow configurations that break out of the prompt template sandbox and execute arbitrary commands on the gateway server. Security researcher invisiblemeerkat reported it.

Affected builds run from 18.1.6 up to 19.2.4, plus 19.3 before 19.3.2 and 19.4 before 19.4.1 — and those three releases are the fixes, available now. Note that's the gateway's own version, not your GitLab instance's, so check the deployment itself before deciding you're clear.

Most customers don't have to. GitLab.com, Dedicated, and self-managed instances pointed at a GitLab-hosted gateway are already covered. Only operators running their own AI Gateway must upgrade manually — immediately. GitLab hasn't reported any exploitation, but the attack needs low complexity and no user interaction at all, so the bar is a valid Duo-enabled account and nothing more.

Source: Cybersecurity News