NEW
Varonis37 minutes ago2 min▣LLM reportcriticalJava4Shell is a heap overflow in OpenJDK's AVX-512 Base64 decoder that corrupts memory pre-authentication in Keycloak, Jira, and Confluence; fixed builds end in .1.
Malformed Base64 padding underflows the output mask, writing 64 bytes past the buffer (128 on JDK 17). Varonis reproduced the overflow on Temurin 17.0.21; -XX:-UseBASE64Intrinsics eliminated it.
NEW
Varonis1 day ago3 min▣LLM reporthighTheMailer, a self-hosted AI phishing kit sold for $3,000–$6,000, uses commercial LLM APIs to rewrite lures until they bypass spam filters.
Operators supply API keys for Claude, ChatGPT, Grok, Gemini, Mistral, or DeepSeek; Inbox Loop checks test inboxes over IMAP and regenerates content after spam hits. Per-recipient templates, randomized metadata, and personalized PDF attachments defeat hash- and template-based detection.
Varonis12 days ago3 min▣LLM reportlowNotC2 runs a command-and-control loop inside a victim's AWS account using SQS, Lambda, SNS, and Bedrock, leaving no conventional C2 server to block.
Deployment requires already-compromised AWS credentials with queue, function, and role-creation permissions. Detection relies on CloudTrail, Lambda, SNS, and Bedrock logs.
Varonis26 days ago3 min▣LLM reporthighA cross-tenant GCP Dataflow vulnerability lets attackers squat predictable staging buckets and swap pickle files to execute code on victims' Dataflow workers, stealing service account credentials.
The vulnerability exploits Apache Beam's deterministic default staging bucket naming in GCP Dataflow. An attacker precomputes the victim's bucket name, creates it first, and swaps staged pickle files to achieve RCE on Dataflow workers. Google deployed server-side fixes and client-side checks in Beam 2.75.0.
Varonis27 days ago3 min▣LLM reportcriticalPHP2Shell vulnerabilities CVE-2026-109618 and CVE-2026-109619 enable unauthenticated remote code execution on PHP servers via crafted multipart uploads, affecting all PHP versions since 5.4.
The flaws reside in PHP's core engine and trigger before application code executes, so CMS-level logging may capture nothing. RCE was demonstrated on WordPress, Drupal, and Joomla container images. Patched releases shipped October 21: PHP 8.5.12, 8.4.27, 8.3.36, 8.2.35. PHP 8.1 and older have no official patch.
Varonis27 days ago5 min▣LLM reporthighStealC's seller resurfaced as Stadia, a full RAT adding HVNC, remote shell, and in-memory loading to StealC 3.5, targeting Windows hosts via three bare-IP C2 servers.
Stadia is a 32-bit MSVC Windows RAT that embeds StealC 3.5's stealer as one of six modules. It persists via hidden scheduled tasks under C:\ProgramData\, injects into suspended browser processes to steal credentials, and beacons over HTTP to three hard-coded bare IPs using RC4-encrypted, Base64-encoded JSON bodies. The builder panel at stadiawatch.cc compiles per-customer builds with rotating keys and tokens.
Varonisabout 1 month ago4 min▣LLM reportcriticalA privilege-escalation chain in Snowflake-managed PostgreSQL turns the CREATEROLE attribute into full OS-level RCE as the postgres user, bypassing every published mitigation without disabling any.
Five chained PostgreSQL features escalate a non-superuser role with CREATEROLE to OS command execution. The chain hijacks pgcron to reach superuser, rebinds blocked C symbols via LANGUAGE internal for arbitrary file read/write, and injects archivecommand to execute shell commands. Snowflake remediated server-side; customers require no action.
Varonisabout 2 months ago4 min▣LLM reporthighAvisLoader is a Windows loader using Tox P2P for C2, delivered via ClickFix lures, with shortcut persistence and a bundled UAC bypass.
AvisLoader uses the Tox encrypted P2P messaging protocol for command-and-control, making domain-based takedowns ineffective. It is delivered through ClickFix lures mimicking DocuSign and includes shortcut modification persistence, a UAC bypass helper, and a process-hiding DLL.
Varonis2 months ago4 min▣LLM reporthighTrustSink registers a rogue External Authentication Method in Microsoft Entra to capture plaintext passwords during legitimate MFA flows, persisting through password resets.
A privileged attacker registers a rogue OpenID Connect provider as an External Authentication Method in Entra. During MFA, the provider shows a fake Microsoft password page, captures the plaintext password, and returns a signed JWT to complete sign-in. The provider persists across password resets, harvesting replacement credentials on subsequent logins.
Varonis3 months ago7 min▣LLM reportcriticalCVE-2026-24301 (CoSnitch) is a critical vulnerability chain in Microsoft Copilot Personal that allows automatic prompt execution via crafted URLs. The chain enables silent data exfiltration from connected OAuth services and persistent memory poisoning via indirect prompt injection through web summarization. The vulnerability was discovered using a technique called meta-hacking, where the AI was manipulated into revealing its own undocumented parameters.
Varonis3 months ago9 min▣LLM reporthighThe WS-Trust usernamemixed endpoint behind Entra Seamless SSO accepts direct SOAP-based username and password authentication requests from the internet and processes them through a non-interactive legacy path. This path does not trigger Smart Lockout, does not populate standard Entra ID sign-in logs, and returns AADSTS error codes that distinguish invalid passwords from valid passwords blocked by MFA, Conditional Access, or expiration, allowing an attacker to confirm credential validity on MFA-protected accounts while evading standard detection controls.
Varonis3 months ago9 min▣LLM reporthighVaronis Threat Labs disclosed a Parameter-to-Prompt (P2P) vulnerability in Atlassian Rovo where a crafted URL using the rovoChatPrompt parameter injects attacker instructions directly into a user's trusted Rovo Chat session. Rovo's federated access across Atlassian products and connected SaaS platforms, combined with the ResearchAgent tool's ability to browse and post to arbitrary external websites, created a one-click data exfiltration chain requiring no jailbreak or guardrail bypass. Atlassian fixed the vulnerability after responsible disclosure via Bugcrowd.
Varonis4 months ago8 min▣LLM reportmediumAI assistant share links across major platforms (ChatGPT, Claude, Grok, Gemini, DeepSeek, and others) create unauthenticated, permanent, crawlable HTTP endpoints that expose user conversations and artifacts to public indexing. While most vendors have added noindex directives, the article demonstrates that legacy domains, archive services like the Wayback Machine, and platforms like DeepSeek and Grok still retain or serve exposed content. The risk is amplified in corporate environments where shadow AI usage and over-permissioned access can result in sensitive organizational data being published to durable public surfaces.
Varonis4 months ago9 min▣LLM reporthighVaronis Threat Labs analyzed Dolphin X, a multi-functional RAT and infostealer advertised by the alias 'Kontraktnik' that targets 300+ applications including browsers, cryptocurrency wallets, SSH keys, cloud CLI tokens, and DevOps credentials. The malware features a server-side build system with a three-tier mutation engine for evasion, and an AI Profiler that scores victims to help operators prioritize high-value targets. Its broad collection scope and AI-driven victim triage represent a notable evolution in cybercrime tooling.
Varonis4 months ago9 min▣LLM reporthighHuggingFace disclosed a breach on July 16, 2026, in which an autonomous AI attacker chained two remote code execution vulnerabilities in its dataset processing pipeline — a remote-code dataset loader abuse and a template injection in a dataset configuration file. The attacker exfiltrated cloud and cluster credentials, moved laterally into internal clusters, and generated decoy activity to complicate attribution. HuggingFace detected the compromise using its own AI-assisted anomaly-detection pipeline but had to deploy an open-weight LLM to bypass commercial model guardrails that refused to process malicious payloads from logs.
Varonis4 months ago9 min▣LLM reportcriticalVaronis Threat Labs discovered a critical vulnerability in Google Cloud Platform's Dialogflow CX service that allowed attackers with only the dialogflow.playbooks.update permission to inject persistent malicious code into the shared Cloud Run execution environment. The vulnerability enabled silent exfiltration of conversation data, bypassing VPC Service Controls, and injection of phishing prompts into chatbot conversations. Additional weaknesses included IMDS credential exposure and unrestricted outbound network access from the Cloud Run environment.
Varonis5 months ago5 min▣LLM reporthighVaronis Threat Labs deployed managed MySQL honeypots across major cloud providers to observe attacker behavior. The publicly exposed GCP instance with weak credentials was compromised within hours by multiple automated ransomware operators who brute-forced access, exfiltrated data, dropped tables, and left ransom notes.
Varonis5 months ago5 min▣LLM reportcriticalVaronis Threat Labs discovered SearchLeak, a critical vulnerability chain in Microsoft 365 Copilot Enterprise Search (CVE-2026-42824). By chaining Parameter-to-Prompt (P2P) injection, an HTML rendering race condition, and a Server-Side Request Forgery (SSRF) via Bing's image search, attackers could exfiltrate sensitive organizational data via a single malicious link.
Varonis5 months ago5 min▣LLM reporthighVaronis Threat Labs demonstrated that enterprise AI agents, specifically an OpenClaw deployment, are vulnerable to traditional phishing and social engineering techniques. In simulated attacks, the agent successfully identified technical phishing indicators like malicious OAuth flows but failed to recognize social context, resulting in the exfiltration of AWS credentials and sensitive CRM data to an external attacker.
Varonis6 months ago4 min▣LLM reporthighA malicious Visual Studio Code extension installed on a GitHub employee's endpoint provided the threat actor TeamPCP with access to exfiltrate approximately 3,800 internal repositories. The incident underscores the critical risk of IDE extensions serving as initial access vectors for supply-chain attacks, allowing threat actors to leverage developer privileges for large-scale data exfiltration.