Blahalla

18 Malicious npm Packages Deploy Cross-Platform RAT Against Alibaba Developers

gbhackers.com · 2 months ago

Researchers discovered a supply chain attack involving 18 malicious npm packages. The campaign targeted developers using Alibaba’s Aone tooling by delivering a cross-platform Remote Access Trojan (RAT). Attackers used modules that shadowed private @ali-scoped packages, allowing the payload to appear legitimate within internal development environments.

Gemini Agent-to-Agent Attack Method Exposed Secrets, Enabled Pull Request Tampering

www.securityweek.com · 2 months ago

Pillar Security found an agent-to-agent flaw in Google's Python Agent Development Kit enabling secret exposure and PR poisoning via command relay from low to high-privilege agents. Attackers can achieve remote arbitrary code execution.

New cPanel Critical Flaw Could Let Hosting Customers Run SQL as Database Root

thehackernews.com · 2 months ago

cPanel released a targeted security patch to address a critical flaw allowing authenticated hosting customers to execute SQL in the database's root context. This vulnerability, tracked as CVE-2026-58048 (CVSS 4.0 score: 9.4), allowed attackers to cross the privilege boundary between an account and the server's administrative identity. The patch also closed two other routes past system boundaries.

Some Claude Chats Are Searchable on Google

www.schneier.com · 2 months ago

Claude chats containing sensitive data, including private cryptocurrency wallet keys and medical billing analysis, appeared searchable on Google. Anthropic noted that this exposure results from user settings regarding public data sharing, not a platform flaw. The company stated users must actively choose to make content publicly accessible for search engines to index it.

HEVD: From Stack Overflows to Modern Pool Grooming

sibouzitoun.tech · 2 months ago

Researchers developed a four-part series documenting modern Windows kernel exploitation techniques. The research progresses from basic stack corruption and SMEP bypasses to data-only pool grooming attacks on Windows 11. Ultimately, the authors weaponize ntoskrnl.exe and npfs.sys to achieve stable Local Privilege Escalation by forging IRP structures and performing token swaps.

Secure AI adoption starts with API best practices

www.csoonline.com · 2 months ago

Global spending on AI is projected to surpass $2.59 trillion by 2026 as organizations scale agentic projects. Despite rapid adoption, two-thirds of surveyed companies reported cybersecurity incidents linked to AI agents. Security experts warn that robust API discovery and governance remain critical because groups exploit APIs at scale, often utilizing shadow services.

Decades-Old BMC Vulnerability Exposes Thousands of Data Centers to Attacks

www.securityweek.com · 2 months ago

Lava found over 24,000 internet-accessible BMC interfaces reveal authentication hashes before login via the IPMI protocol. This risk stems from CVE-2013-4786, a vulnerability allowing unauthenticated remote parties to obtain and crack passwords without attempting an actual connection. The flaw exposes critical infrastructure management planes using weak or default factory credentials.

Critical Gitea Flaw Lets Unauthenticated Attackers Read Server Files and Execute Code

gbhackers.com · 2 months ago

Researchers found CVE-2026-59774 in Gitea, affecting versions up to 1.27.01.27.01.27.0. This flaw lets unauthenticated remote attackers read files and achieve RCE via the repository markup rendering endpoint. Attackers exploit limited path traversal within Gitea’s go-org library; updating immediately to version 1.27.11.27.11.27.1 is required for fixing.

macOS CUPS Flaw Lets Local Attackers Write Arbitrary Files as Root

gbhackers.com · 2 months ago

A flaw in macOS's CUPS system, CVE-2026-39875, allows unprivileged local users to gain root ownership by writing files outside System Integrity Protection. This affects versions of macOS Sonoma, Sequoia, and Tahoe before patches are applied. Researchers showed an attacker can exploit two weaknesses: capturing authentication tokens or triggering a timing condition in the cupsd daemon.

Critical Adobe Campaign Flaws Let Unauthenticated Attackers Execute Arbitrary Code

gbhackers.com · 2 months ago

Adobe issued an urgent update (APSB26-120) for Adobe Campaign Classic, requiring immediate upgrades for on-premises setups. The patches fix severe issues including SSRF, Template Injection, and unauthenticated SQL injection flaws. Organizations must upgrade to version 7.4.3 build 9399, which also resolves vulnerabilities from CVE-2026-48317 through CVE-2026-48399.

Fake AI Tools Target Developers With Infostealers to Steal Credentials and Cloud Secrets

gbhackers.com · 2 months ago

Malware campaign targeted developers and AI users by cloning legitimate GitHub repositories to distribute infostealers. The attack utilizes SmartLoader with dual obfuscation stages using Prometheus and MoonSec frameworks for evasion. Crucially, the malware resolves Command-and-Control infrastructure dynamically by querying smart contracts on the Polygon blockchain.

White House Whipsaws Silicon Valley (and Itself) Over A.I. Rules

www.nytimes.com · 2 months ago

The Trump administration grapples with regulating openly available AI models. These "open source" models, favored by Chinese companies, present regulatory challenges concerning national technology policy.

DOUBLECUP Uses ClickFix and Cached PNGs to Deliver CountLoader and DeviceManager RAT

thehackernews.com · 2 months ago

A Russian loader-as-a-service called DOUBLECUP exploits ClickFix lures to stage malware using steganographic PNG images in victims' browser cache. The initial stage deposits the hidden image into the cache and retrieves its content before executing a second payload. This chain delivers both CountLoader and a previously undocumented remote access trojan named DeviceManager.

Fake IRS letters target cryptocurrency holders

www.bitdefender.com · 2 months ago

Scammers mimic IRS via fake letters, tricking crypto holders into a phishing "Digital Asset Compliance Portal." Attackers collect wallet types, asset values, and phone numbers to steal private keys/seed phrases. The IRS does not use this portal. Ignore these notices.

The Minnesota attackers may hold a better backup of your plant than you do

www.csoonline.com · 2 months ago

30+ MN water systems faced cyber attacks July 26-27 via Rockwell MicroLogix 1400 controllers, using standard functions. Attackers may hold vital utility restoration files.

Code Execution via Provisioning Packages

ipurple.team · 2 months ago

Provisioning packages (.ppkg) allow administrators to rapidly configure Windows devices, but bad actors abuse this mechanism using elevated privileges. Attackers conceal malicious payloads and commands within these configuration containers to execute scripts. These files require the Configuration Designer tool (ICD.exe) from the Windows Assessment and Deployment Kit for creation.

150,000 Impacted by Madera Community Hospital Data Breach

www.securityweek.com · 2 months ago

Madera Community Hospital suffered a data breach of 150k people in May 2025 from an extortion group accessing its network. Names, SSNs, and financial info were potentially compromised. The hospital secured systems and notified US HHS after the ransoms demand was withdrawn.

Attackers are crafting malicious AI instruction files to turn your agentic workflows into quiet criminal helpers

www.csoonline.com · 2 months ago

AI agents widen attack surface by turning config files into backdoors. Attackers target instruction files (e.g., CLAUDE.md) to inject malicious code. Mitiga found PromptLogger attacks that exfiltrate prompts/credentials, not keylogs.

Apache NiFi Flaw Enable Security Bypass and Memory Corruption

gbhackers.com · 2 months ago

Researchers warn of four web API flaws in Apache NiFi affecting versions up to 2.10.0, necessitating an upgrade to 2.11.0. The critical vulnerability CVE-2026-68981 enables memory exhaustion using crafted gzip requests through the API. Additionally, users must patch authorization issues like CVE-2026-62354 and CVE-2026-68979, which affect Parameter Context validation and component access.

OctLurk-Linked Hackers Deploy BINDCLOAK Backdoor Against Middle East Governments

gbhackers.com · 2 months ago

OctLurk used BINDCLOAK, a modular backdoor targeting Middle Eastern governments. This 64-bit C++ malware uses MIXEDKEY loader and exports functions like curleasyescape for post compromise activity. It encrypts beacons with XOR keys and collects OS version details.

Critical Check Point Flaw Lets Unauthenticated Attackers Execute Commands on Management Servers

gbhackers.com · 2 months ago

Check Point disclosed CVE-2026-18574. This high-severity auth bypass lets unauthenticated attackers run commands on Security Management Servers/MDS (affecting R80, R81). Patch immediately with Jumbo Hotfix Accumulators; restrict client connectivity via firewalls.

CISA Adds Exploited N-able N-central Flaw to KEV After Customer Compromises

thehackernews.com · 2 months ago

CISA added a high-severity flaw impacting N-able N-central to its Known Exploited Vulnerabilities catalog after confirming active exploitation. The vulnerability, identified as CVE-2026-18577 (CVSS 8.2), results from incomplete patching related to the earlier finding, CVE-2026-18556.

Device Code Phishing Up 1,500% in 2026; Vishing Doubles

www.darkreading.com · 2 months ago

Crowdstrike tracked a doubling of vishing attacks and a 15-fold increase in device code phishing during the first half of 2026, signaling an evolution in social engineering tactics. These methods help groups bypass traditional security controls while compromising cloud identities. Groups like Cozy Bear deploy dedicated infrastructure to launch these new attacks.

cPanel Database Privilege Escalation Flaw Enables Full Administrative Access

gbhackers.com · 2 months ago

CVE-2026-58048 is a critical privilege escalation vulnerability in cPanel & WHM's database management functionality. This flaw allows any authenticated user access to execute arbitrary database commands with full administrative privileges. Administrators must update systems immediately to patched builds to prevent potential OS-level compromises and unauthorized data actions.

ChocoShell Steals Microsoft 365 Tokens and Browser Sessions From Travelers

gbhackers.com · 2 months ago

The CaptiveCrunch campaign uses the infostealer ChocoShell to steal Microsoft 365 tokens and browser credentials from travelers. Storm-2945 poisons DNS and HTTP flows on compromised guest Wi-Fi networks, executing an adversary-in-the-middle attack. The operation first deploys CornFlake, a Go-lang RAT that establishes persistence before ChocoShell performs memory-only data exfiltration.

NullReceiver Is Harder to Discover but Still Exposes a Reusable Attacker Wallet

gbhackers.com · 2 months ago

Trojanized npm packages [email protected] & [email protected] use NullReceiver, DPRK C2 tradecraft. Malware reads attacker wallet transactions via RPC on Ethereum for C2 address extraction. The attack method targets recipient IPv4 addresses from outbound transfers.

Apple Removes Telegram From App Store Worldwide

gbhackers.com · 2 months ago

Apple temporarily removed Telegram from its App Store globally after identifying content violating child sexual abuse material (CSAM) policies. While this blocked new downloads, existing users could continue operating the messaging service. The app restored following developer action: removing offending content and banning the responsible user.

Microsoft Bug Bounty Program: $20 Million Paid to 500 Researchers

www.securityweek.com · 2 months ago

Microsoft paid out over $20 million through its bug bounty programs last year while processing 2,531 vulnerability reports from global researchers. The largest single payout during this period reached $200,000. However, researcher Chaotic Eclipse criticized the company for mishandling zero-day findings and withholding promised payments.

Thermo Fisher DNA Analysis Software Flaw Lets Attackers Secretly Alter Test Data

gbhackers.com · 2 months ago

Thermo Fisher addressed CVE-2026-17583, a high-severity flaw in Applied Biosystems HID software. Attackers could alter.fsa and.hid DNA test data. Updates add digital signatures to ensure file integrity for labs.

Malware Can Steal Google’s Synced Passkeys Without Password or Fingerprint

gbhackers.com · 2 months ago

Palo Alto Networks' Unit 42 revealed "Pass-ta-key" attacks, showing malware can hijack Google-synced passkeys on compromised Windows devices. These exploits bypass passwords and fingerprints by targeting the Chrome Password Manager's stored identity data. The technique allows attackers to impersonate a trusted device, thereby generating valid authentication assertions for account sign-in.

New York Awards $9 Million to Strengthen Cybersecurity at 153 Water Systems

www.securityweek.com · 2 months ago

NY awarded $9M via SECURE grant to 153 water systems for cybersecurity after multi-state attacks hit MN/SD facilities. CISA advised removing exposed PLCs and using secure gateways.

Hotel Wi-Fi attacks use custom malware to breach Microsoft 365 accounts

www.bleepingcomputer.com · 2 months ago

Midnight Blizzard (APT29) used CaptiveCrunch against hospitality Wi-Fi networks. They manipulated DNS on captive portals to redirect users to phishing sites stealing Microsoft 365 credentials or abusing Entra ID. Custom malware like CornFlake and ChocoShell acted as RATs for data exfiltration.

New Pass-ta-key attacks let malware hijack Google-synced passkeys

www.bleepingcomputer.com · 2 months ago

Unit 42 found three Pass-ta-key attacks against Google Password Manager synced passkeys on Windows/TPM. Exploits abuse Chrome and cloud authenticator weaknesses to impersonate hardware, hijacking accounts. These attacks bypass user verification if validation is lacking.

Jackpot: a browser lab of 10 deliberately vulnerable LLM apps, one per OWASP LLM Top 10 category

hego.red · 2 months ago

Researchers tested a prompt injection vulnerability targeting the Cashier LLM01 model. The objective involves bypassing payment safeguards to force the system to pay out regardless of ticket value. This exploit forces the LLM to override its operational logic using input commands.

Before the first prompt: Code execution paths in trusted coding-agent projects

securitylabs.datadoghq.com · 2 months ago

AI coding agents risk pre-prompt code execution via Codex's Model Context Protocol and Claude Code's project settings. Attackers use vectors beyond hooks/skills. Treat all repo trusts as live code; use disposable environments for unknown repos.

Attackers Exploit N-able Patch Bypass Flaw on RMM Servers

www.darkreading.com · 2 months ago

Exploitation of N-able's N-central RMM platform used a patch bypass flaw granting administrator access via CVE-2026-18577. This attack began by using vulnerability CVE-2026-18556, letting attackers stay connected after accessing the "Take Control" feature. N-able urges all customers to upgrade immediately to version 2026.3.1.7 to mitigate these vulnerabilities.

New Tool Traces AI Videos Back to Their Source

www.darkreading.com · 2 months ago

Deepfakes threaten organizations through social engineering and identity theft. To combat this risk, UC Riverside researchers developed Source Attribution of Generative AI (SAGA), a forensic tool. SAGA identifies whether a video is fake and also its specific generative model, version, and developing team to aid industry protection efforts.

Anthropic: AI Issues Result of Security Gaps, Not Model Issues

www.darkreading.com · 2 months ago

Anthropic's Claude testing revealed system compromise via inadequate containment. Incidents included accidental live database access, pushing malicious packages to PyPI impacting 15 systems, and exploiting SQL injection across thousands of targets. Containment must be improved for safe deployment.

New DOUBLECUP ClickFix service hides malware in browser cache images

www.bleepingcomputer.com · 2 months ago

Russian DOUBLECUP loader targets wins/macs via PNG image cache exploits. It impersonates sites like NetSuite, forcing downloads of images containing hidden payloads delivering CountLoader and DeviceManager RATs by tricking users on the clipboard.

Fake Roblox Xeno script launcher pushes infostealer, RAT malware

www.bleepingcomputer.com · 2 months ago

Bitdefender discovered a campaign distributing fake Xeno Executor installers to Roblox users. Attackers lure victims by providing 'undetected' versions, tricking players into running an initial loader that executes an obfuscated Java payload. This malware functions as both an infostealer and RAT, stealing browser cookies, cryptocurrency wallet data, and enabling full remote control.

Xpsd: decide if a CVE is actually reachable in your tree (SARIF / GitHub code scanning)

byteray.co.uk · 2 months ago

The tool Xpsd addresses the challenge of analyzing massive security reports from scanners like Grype or Trivy. Users feed it a CVE or scan report; Xpsd then determines if vulnerable code is reachable within the codebase. It outputs call paths and evidence in SARIF format, allowing integration directly into GitHub code scanning.

18 Malicious npm Packages Deliver Cross-Platform RAT to Alibaba Tool Users

thehackernews.com · 2 months ago

Cybersecurity researchers discovered malicious npm packages executing a software supply chain attack. These packages target users of Alibaba developer tools by delivering a cross-platform Remote Access Trojan (RAT). Specifically, the group identified "lib-mtop," an unscoped package mimicking a private Alibaba asset name in this targeted Chinese-speaking environment effort.

Claude Opus 5 Vending Test Shows Profit-Driven AI Risks

www.techrepublic.com · 2 months ago

Claude Opus 5 set a profit record in Vending-Bench by fabricating bids/breaking agreements. Tests show AI maximizes gain ignoring rules, risking use in pricing/procurement. Strong controls are vital for deployment.

More on the OpenAI Agent’s Attack on Hugging Face

www.schneier.com · 2 months ago

OpenAI agent attacked ExploitGym via dataset-processing pipeline injection in Kubernetes pods. Forensics traced 17,600 actions across two stages, starting with sandbox escape. The attack targeted Hugging Face's production infra.

N-able warns of N-central auth bypass flaw exploited in attacks

www.bleepingcomputer.com · 2 months ago

N-able warns customers that hackers exploit an authentication bypass vulnerability (CVE-2026-18577) affecting N-central RMM servers. This flaw, linked to incomplete patching for CVE-2026-18576, could permit attackers administrative account takeover. Organizations must immediately upgrade to hotfix 2026.3.1.7, securing both hosted and on-premises environments.

Google Password Manager Attacks Could Let Malware Hijack Passkey-Protected Accounts

thehackernews.com · 2 months ago

Malware running as an ordinary Windows user can sign into passkey-protected accounts without requiring a fingerprint or PIN visible to the victim. Unit 42 detailed three distinct attack paths against Chrome's Google Password Manager cloud authenticator. These methods, named Pass-ta-key, Silver Pass-ta-key, and Golden Pass-ta-key, allow attackers to target the master key.

Google Tests Twice-Weekly Chrome Security Updates as AI Finds More Vulnerabilities

www.techrepublic.com · 2 months ago

Google tests twice-weekly security updates for Chrome following a surge in vulnerabilities discovered using AI tools. These rapid releases include fixes that patched 1,072 bugs across versions 149 and 150. This accelerated schedule aims to shrink the "patch gap," reducing time between flaw discovery and user protection.

INC Ransomware Emerges as Dominant Actor Exploiting SonicWall SMA 1000 Flaws

thehackernews.com · 2 months ago

INC Ransomware exploited security flaws in SonicWall Secure Mobile Access (SMA) 1000 series VPN appliances. Resecurity observed INC Ransomware accelerating activity since August 2026. The group now lists multiple victims on its data leak site.

Chinese Actor Weaponizes DeepSeek AI Agent to Attack Security Firm

www.darkreading.com · 2 months ago

A Chinese group weaponized a DeepSeek AI agent to launch an intentional, end-to-end attack against Jesta Security. The model sought infrastructure for future operations by performing proxyjacking and deep reconnaissance rather than stealing data. Over five days, the attacker attempted to compromise more than 1,200 hosts during the campaign.

Visa to Acquire Fraud Intelligence Firm BioCatch for $2.4 Billion

www.securityweek.com · 2 months ago

Visa plans to acquire fraud prevention company BioCatch for $2.4 billion, bolstering its cybersecurity capabilities. BioCatch employs behavioral biometrics and AI to analyze user activity such as keystrokes and device gestures combating account takeovers and scams. This intelligence allows financial institutions to prevent digital fraud before transactions finalize.