In-depth analysis of malware, threat actors, attack campaigns, techniques, and emerging cyber threats.
2492 articles
Datadog researchers trace compromised LiteLLM PyPI releases 1.82.7 and 1.82.8 to TeamPCP’s wider supply-chain campaign. The payload harvested credentials, established persistence and could spread into Kubernetes; responders are urged to rotate exposed credentials and be
ANY.RUN’s analysis describes Kamasers as a multi-vector DDoS botnet that retrieves C2 addresses through public-service Dead Drop Resolver channels and can download and execute additional payloads; it details the malware’s commands, infrastructure and detection.
Sysdig researchers observed a stealer matching the Trivy compromise in Checkmarx’s ast-github-action about four days later. The activity suggests stolen credentials enabled further action poisoning; runtime detection identified the behavior.
Huntress reports a campaign active since January 2026 using tax-themed Google Ads, stacked cloaking, and rogue ScreenConnect installers to deploy HwAudKiller, which uses a signed Huawei driver to terminate security processes.
Researchers found credential-stealing code in LiteLLM PyPI versions 1.82.7 and 1.82.8. Version 1.82.8 used a .pth file to run at Python startup; PyPI quarantined the project.
Insikt Group analyzed five ClickFix clusters that trick users into running obfuscated commands in Windows or macOS tools. The campaigns delivered RATs and information stealers; the report details their execution chains and recommends hardening native utilities.
The article says compromised LiteLLM versions 1.82.7 and 1.82.8 execute malicious code that steals credentials and Kubernetes secrets, and recommends removing the package, investigating indicators, blocking listed domains, and rotating credentials.
A scan of 127,906 IPv6 prefixes found two active ip6.arpa zones using wildcard A records, a technique that can evade reputation-based defenses. Another 382 zones had Cloudflare nameservers but no active A records; phishing delivery from the second campaign was not seen.
Netskope traces TroyDen’s Lure Factory to 300+ trojanized GitHub packages using paired LuaJIT components, sandbox-evasion delays and screenshot exfiltration. Analysis links three lure campaigns; credential theft is suspected, not confirmed.
Malicious LiteLLM versions 1.82.7 and 1.82.8 used Python startup execution and persistence to collect credentials and other sensitive data, then exfiltrate it. PyPI quarantined the packages; LiteLLM published a security update.
CIS reports a phishing campaign targeting U.S. SLTT government entities with IRS- and SSA-themed TryCloudflare links that automatically download legitimate RemotePC software, potentially giving attackers access to victims’ computers.
LevelBlue analysts traced a reusable malware delivery framework using obfuscated VBS, fileless PowerShell, PNG-embedded .NET loaders, and a separate fake-PDF shortcut chain. SentinelOne prevented execution of the initially detected VBS file.
SEKOIA’s TDR tracks Silver Fox campaigns across South Asia from 2025–2026, detailing tax-themed phishing, ValleyRAT delivery, abuse of a misconfigured Chinese RMM tool, and a WhatsApp-themed Python stealer. It assesses mixed espionage and profit motives.
ReversingLabs found malicious npm packages that mimic install logs to phish for sudo passwords, then retrieve and run a RAT designed to steal crypto wallets and sensitive data. Payload details came from Telegram or teletype[.]in.
Malwarebytes describes FriendlyDealer, a campaign using fake app-store pages across 1,500+ domains to install PWAs that redirect users to gambling offers. The kit collects telemetry but, according to the report, does not install traditional malware or steal passwords.
StepSecurity’s analysis found LiteLLM 1.82.7 and 1.82.8 packages contain credential-stealing payloads that encrypt and attempt to exfiltrate secrets, install persistent backdoors, and could spread across Kubernetes nodes; it recommends rotating exposed credentials.
Hunt.io researchers found an exposed directory on Proton66 containing a TheGentlemen-attributed ransomware toolkit. Mimikatz logs with victim credentials indicate prior use, and scripts show extensive preparation to disable defenses and deploy ransomware.
Aikido says TeamPCP’s Kubernetes campaign installs CanisterWorm on non-Iranian nodes and uses a destructive payload on systems matching Iran’s timezone and locale. Variants also spread via SSH or unauthenticated Docker API access.
Wiz reports TeamPCP compromised KICS GitHub Action tags and two Checkmarx OpenVSX extensions, deploying credential-stealing malware with GitHub, cloud and Kubernetes access capabilities. PyPI packages were also affected; maintainers and PyPI took action.
XM Cyber describes eight AWS Bedrock attack paths involving permissions, integrations, logs, knowledge bases, agents, flows, guardrails and prompts, and recommends tighter permissions and security posture controls.
AhnLab reports that Larva-26002 targets exposed MS-SQL servers with weak credentials and deploys the Go-based ICE Cloud Client, which receives scan targets and credentials from C2 and reports successful logins.
StepSecurity says TeamPCP used npm tokens stolen through the Trivy v0.69.4 compromise to spread CanisterWorm, which installs a persistent Python backdoor, steals npm tokens, and can deliver destructive Kubernetes payloads.
Huntress says an EvilTokens campaign used Railway-hosted infrastructure for Microsoft 365 device-code phishing, affecting 344 organizations. The investigation details token replay and phishing methods; Huntress reports blocking further attempts.
Google’s GTIG details DarkSword, a JavaScript iOS exploit chain abusing six vulnerabilities to gain kernel privileges, and its use by commercial surveillance vendors and suspected state actors to deploy three payload families. Apple patched the flaws; GTIG advises users
