Search/hubspot
● Vendor

hubspot

Known CVEs
0
Highest CVSS
—
In KEV
0
Vendor
jinjava
Connections
21 relationships
DataGrout helps enterprises control AI usage, governance and LLM costs
DataGrout helps enterprises control AI usage, governance and LLM costs SelectHub has announced the launch of DataGrout, its specialized AI research lab introducing an LLM inference optimization platform and AI governance solution for enterprises. DataGrout’s mission is to drive token reduction for agentic workflows, chatbots and AI tools, while equipping IT and FinOps leadership with a policy-driven, auditable LLM payload and cost monitoring system to track company-wide AI utilization. As enterprise users increase LLM usage through context-intensive chat sessions and coding tools such as Claude Code, Cursor and Replit, uncontrolled token waste and hidden API expenses are putting significant strain on corporate budgets. This is especially impactful as organizations also begin expanding agentic workflow deployments across diverse use cases such as customer service and sales outreach. DataGrout addresses this challenge through a unified connection layer that enables agents to securely access all enterprise applications with built-in dynamic context pruning, alongside a symbolic inference layer that reduces calls to the underlying LLM. Early tests have shown promising results with token reduction averaging 60% for data-intensive tasks involving ERP and CRM integration, without compromising workload accuracy. “Companies are flying blind into the AI era,” said Venkat Devraj, CEO at SelectHub. “Through DataGrout, IT and Finance teams gain precise token tracking and visibility into LLM consumption patterns and payloads, enabling leadership to clearly measure the value delivered by AI investments while systematically eliminating waste on a per LLM-call basis. Deterministic workloads are kept off the LLM entirely without users and developers having to manually segregate and route workloads. The net positive impact on the P&L is instantaneous.” The critical market need for tools that help IT and FinOps / AI Tokenomics teams audit and eliminate token waste is being echoed across enterprise leadership. Deepak Dodani, Chief Information Officer at global transportation management and freight audit firm AFS Logistics, noted: “The twin challenge for companies deploying AI is token economics and governance. Right now, most companies cannot accurately assess the extent to which their LLM resources are driving business value versus being drained by unnecessary or unauthorized side-projects. DataGrout’s premise in terms of reducing token waste while enabling visibility into LLM usage, is very timely.” Bharath Prabhakaran, Chief Digital Officer at University of Cincinnati, also emphasized the need for enterprise LLM usage tracking and agent governance: “Rampant LLM token usage has many CIOs and CFOs worried. AI line items can quickly become one of their fastest-growing operational expenses. Without visibility into LLM usage and token consumption metrics that show which teams and use cases consume the most tokens, forecasting ROI is nearly impossible. DataGrout endeavors to reduce token usage while increasing the transparency needed to reduce waste and prevent abuse.” DataGrout acts as a MCP gateway and control plane with agents connecting via its Conduit SDK or via MCP or JSON-RPC. DataGrout allows any third-party MCP server to be accessed through its platform, while also offering its own MCP servers for popular SaaS applications such as Salesforce, HubSpot, Oracle ERP, SAP, ServiceNow, Snowflake and QuickBooks. DataGrout can be connected to LLMs directly with companies bringing their own key, or via popular AI gateways such as Amazon Bedrock and Kong.
helpnetsecurity.comAug 13, 2026extracted
Cybersecurity Firms Impacted by Klue Supply Chain Attack
Cybersecurity firms Huntress and Recorded Future have disclosed the impact of a supply chain attack that hit market intelligence platform Klue. The attack started on June 11 and affected systems associated with software platform integrations. The hackers connected to Klue’s backend servers and executed unauthorized commands, pushing a code update to harvest OAuth tokens for customers’ Klue integrations. Klue notified customers of the incident on June 12, warning that it had deactivated OAuth tokens for all customers and disabled integrations with Salesforce, HubSpot, SharePoint, Zoom, Gong, Chorus, Clari, Google Drive, and Slack. According to ReliaQuest, the hackers abused the Salesforce REST API to exfiltrate large volumes of customer relationship management (CRM) data over a 24-hour window, “including a concentrated burst of nearly a thousand queries in 15 minutes and sustained extraction windows lasting over 6 hours”. On June 17, Salesforce disabled the Klue Battlecards app integration, warning that it “detected unusual activity involving the app that may have resulted in unauthorized access to a subset of customer data via the app’s connection to Salesforce”. On Thursday, both Huntress and Recorded Future confirmed that they were among the companies affected by the supply chain attack. “The data that was copied from our Salesforce account includes business contacts, price quotes, and other sales-related data and messaging. No threat data, passwords, payment card information, or engineering data relating to the Huntress agent or telemetry we collect was affected,” Huntress said. Recorded Future noted, “While our investigation is ongoing, we believe the impact was limited to business data fields stored in our Salesforce database, such as client contact names and email addresses. Certain business contract information may also have been potentially included in the impacted data.” The incident was limited to the Klue-Salesforce integration, and the attackers did not access any systems belonging to or maintained by the two cybersecurity firms. Huntress noted that several other cybersecurity companies use Klue, but no other firm appears to have publicly disclosed impact from the attack. The attack follows the same pattern observed in previous Salesforce, Salesloft Drift, and Gainsight incidents, which have been attributed to ShinyHunters and UNC6395, but appears to have been mounted by a new threat actor. Huntress said it received attempted extortion communication from a threat actor calling itself “mr bean”, who pointed to a Session Messenger ID associated with Icarus, an extortion group that emerged in April 2026. Icarus’ leak site has one entry from early May, with the data allegedly stolen from the victim already published (albeit no longer available), and another from June 16, which points to data stolen from Salesforce. “With those matching data points, we have high confidence that the Icarus actor is responsible for the Klue compromise and this supply chain attack,” Huntress says. While it has shared details of the attack with its customers, Klue has not made a public announcement on the matter. SecurityWeek has emailed the company for a statement and will update this article if it responds. Update, June 22: Klue publicly acknowledged the attack, other cybersecurity firms disclosed the impact of the incident, and Icarus listed Klue on its leak site. Related: Atomic Arch Supply Chain Attack Hits 1,500 AUR Packages Related: Over 100 NPM, PyPI Packages Hit in New Shai-Hulud Supply Chain Attacks Related: Maine Disables Data Breach Portal Due to Fake Submissions Related: White House Issues Memo to Bolster NSS Cybersecurity
securityweek.comJun 19, 2026extracted
Klue OAuth breach linked to 'Icarus' Salesforce data theft attacks
Market intelligence platform Klue suffered a OAuth breach that enabled the "Icarus" threat actors to steal Salesforce CRM data from multiple organizations in an ongoing extortion campaign. Sources told BleepingComputer of the attack yesterday, telling us that numerous organizations had their Salesforce data stolen and were now being extorted by the relatively new extortion group. Cybersecurity firms ReliaQuest and Huntress have both published reports confirming the security incident, with Huntress stating that their Salesforce data was stolen in the attack. Salesforce has since disabled the Klue Battlecards integration on its platform while the breach is investigated. "To protect our customers, Salesforce has disabled the connection between the Klue Battlecards app, installed by individual customers, and Salesforce as part of our response to a recent security incident," Salesforce warned yesterday. "As a result, organizations will not be able to connect to Salesforce via this app until further notice." If you have any information regarding this incident or other undisclosed attacks, you can contact us confidentially via Signal at 646-961-3731 or at [email protected]. Stolen OAuth credentials used to steal Salesforce data ReliaQuest stated that attackers gained access to Klue Battlecards integration service accounts and used OAuth tokens associated with customer Salesforce instances to carry out data theft. The researchers observed the threat actors generating OAuth tokens and then using automated Python scripts to query Salesforce's REST API for nearly 24 hours. The activity began with reconnaissance of an organization's Salesforce instances through the '/services/data/v59.0/sobjects' endpoint before exfiltrating data using the '/services/data/v59.0/query'. ReliaQuest said that for one of the organizations, the attackers slowly mapped out their Salesforce objects to identify valuable objects and then rapidly stole data once they knew what they wanted. "The attacker then hit the same endpoint, sending almost a thousand queries in a 15-minute window in at least one environment," explained ReliaQuest. "Where the first stage was a slow, steady pull designed to blend in, this burst traded stealth for speed, suggesting either time pressure or a shift to targeted records. In another case, the exfiltration was observed over 6 hours." The researchers said the activity closely resembled previous Salesforce third-party integration data theft attacks by the ShinyHunters extortion group, but were unable to attribute the attacks to the threat actor. However, BleepingComputer learned yesterday that ShinyHunters was not behind this attack, but rather a relatively new threat actor known as "Icarus" who had already begun emailing extortion demands to Klue customers impacted by the breach. A ransom note shared with BleepingComputer showed that the emails were sent using the alias "mr bean" and included a Session Messenger ID to contact them. The threat actors' data leak site also contains a message hinting at the extortion campaign in a simple post titled "Get Ready," stating, "big corps getting listed. be ready." Icarus is believed to have launched in April 2026, and initially listed two victims on its leak site, with BleepingComputer learning that at least one of these victims is connected to the Klue campaign. That company has now been removed from the data leak site, which may indicate that negotiations are underway. Today, Huntress disclosed that it was among the organizations impacted by the Klue breach, confirming that they had received a similar extortion email as seen by BleepingComputer. However, the Session ID used in later emails was different and was instead the one listed on the Icarus data leak site, providing additional evident that they were behind the attack. "In the initial email, the adversary suggests, 'we advice you to write to us on Session' (sic)," reported Huntress. "The Session Messenger ID that they provided matched the same values included on the dark web leak site of a new extortion group dubbed 'Icarus.'" According to Huntress, Klue told customers that attackers first compromised the company's backend systems and then pushed a malicious code update that stole OAuth tokens customers use to integrate the Battlecards product with third-party platforms. The attackers reportedly used a dormant but still active credential created by Klue for a prototype integration. After gaining access to Klue's environment, they stole customer OAuth tokens and used them to query connected Salesforce environments directly. Klue later disabled integrations with Salesforce, HubSpot, SharePoint, Zoom, Gong, Chorus, Clari, Google Drive, and Slack while responding to the incident. Huntress said the stolen data includes CRM-related information, including business contacts, sales communications, price quotes, competitive intelligence reports, and account data. The cybersecurity company said there was no evidence that threat intelligence, customer telemetry, passwords, payment card information, or engineering systems were compromised. Both ReliaQuest and Huntress shared IP addresses linked to the attacks, which are listed below: 138.226.246.94 212.86.125.24 213.111.148.90 94.154.32.160 Organizations using Klue integrations are advised to review Salesforce and related SaaS logs for activity originating from these addresses, revoke and rotate OAuth tokens, terminate active sessions, and review Salesforce logs for unusual API activity. Overall prevention scores can hide what happens after initial access. Once attackers are using valid credentials, prevention drops sharply. The Blue Report 2026 measures defenses technique by technique across 338 million simulations run in customer production environments. Get the report
bleepingcomputer.comJun 18, 2026extracted
Bonfy ACS 2.0 helps organizations control data use in AI environments
Bonfy ACS 2.0 helps organizations control data use in AI environments Bonfy.AI announced Bonfy Adaptive Content Security (Bonfy ACS) 2.0, a platform built to secure enterprise content across all systems, applications, and AI agents – anywhere data moves, resides, or is processed. As organizations race to deploy copilots, custom AI apps, and increasingly autonomous AI agents, security leaders are struggling with blind spots around how these systems access, transform, and share sensitive data, gaps that legacy DLP and DSPM tools were never designed to handle. By 2028, Gartner projects that 22% of cyberattacks and data leaks will involve generative AI, and through 2029 over 50% of successful cybersecurity attacks against AI agents will exploit access‑control issues. Bonfy delivers real-time, contextual protection across email, SaaS apps, collaboration tools, browsers, cloud and on‑prem file stores, AI systems, and agent frameworks, so enterprises can safely accelerate AI adoption without flying blind. With native coverage for Microsoft 365 (Exchange Online, SharePoint, Entra, Copilot and Purview), Google Workspace (Gmail, Google Drive, Google Directory), Salesforce, HubSpot, Slack, on‑premises file stores, AWS S3, and more, Bonfy becomes the unifying data security layer that follows content regardless of channel or AI workflow. “AI agents are about to become the biggest catalyst in our market, and the biggest new attack surface enterprises have seen in a decade,” said Gidi Cohen, co-founder and CEO of Bonfy. “They read, write, click, send, and connect to everything. Our mission with Bonfy 2.0 is simple: protect the data across that entire lifecycle, no matter which agent, platform, or workflow is in play.” Built for agentic and autonomous AI Bonfy ACS 2.0 is engineered specifically for system‑level and browser‑based AI agents that plan, reason, call tools, and execute actions across enterprise systems. By treating agents as first‑class entities, not just extensions of users, Bonfy allows security teams to see which agents accessed which data, how they used it, and where the outputs ultimately landed. Bonfy ACS 2.0 fits cleanly into a customer’s existing security and productivity stack: it complements Microsoft Purview and M365 DLP, integrates with Microsoft Entra and Google Directory, and plugs into SIEM/SOAR tools such as Splunk, Sentinel, and Rapid7 for workflow automation, while also integrating via its MCP Server interface and APIs with modern AI platforms including Microsoft Copilot Studio, OpenAI, Anthropic Claude, Google Gemini, and other enterprise agent frameworks. Bonfy is designed for executive level visibility and governance for CISOs, CIOs, and to be operated by security teams, security architects, and AI platform teams responsible for GenAI and agent deployments in financial services, insurance, technology, biotech/pharmaceutical, healthcare companies, and more. It provides one policy and automation engine that spans traditional data security, AI data governance, and AI agent guardrails, eliminating the need to stitch together separate point products for systems, and agents. The Bonfy platform can be used both for projects where organizations consume AI and build AI. “The AI agent factory is already here,” said Tal Hornstein, CISO of Cast & Crew. “Microsoft, Google, Salesforce, ServiceNow, and the hyperscalers are all shipping frameworks for building agents. Bonfy gives enterprises a way to say: Yes, we want the productivity, but we’re not willing to lose control of our data to get it.” Headline capabilities in Bonfy ACS 2.0 Bonfy ACS 2.0 introduces six major capabilities that together form a second-generation, high‑performance data security platform for the AI era. AI agent data guardrails (MCP and agent framework support) Bonfy adds “data in use” security solution by adding an MCP server interface, API, and agent‑aware controls so enterprises can inspect and govern the content AI agents read, share, and generate during planning, reasoning, and execution, not just in the final output. Agents can call Bonfy inline to label and risk‑score content before it reaches external services or users, stopping AI‑driven leakage and trust‑boundary violations. Browser extension for Shadow AI and agentic activity A lightweight browser extension delivers real‑time, content‑aware inspection of web traffic, including unsanctioned AI tools and browser‑based assistants. Bonfy separates safe AI use from risky disclosure, detects shadow AI automations, and shows security teams exactly where sensitive data is going. Full Google workspace support Bonfy 2.0 adds native support for Gmail, Google Drive, and Google Directory, achieving parity with Microsoft 365 integrations and extending multi‑channel protection across both ecosystems. Organizations running on Google get unified, entity‑aware controls including contextual, automated classification labeling. Data surface visibility for AI-era risk A new “data surface visibility” view gives CISOs a live map of where sensitive content lives across data stores such as Microsoft SharePoint, Google Drive, AWS S3 buckets, On-prem file stores and AI systems, and how employees and agents use it. Teams can drill from high‑level exposure down to specific actors and flows to understand real business risk, not just isolated events. On-premises and cloud file store coverage Bonfy covers on‑premises file stores and cloud object storage such as AWS S3, alongside existing SharePoint, Google Drive and other SaaS applications. This creates a unified control plane for unstructured data at rest, in motion, and in use. Data minimization, encryption enhancements, and SOC 2 Bonfy 2.0 tightens data minimization, encryption, and configurable retention so the platform itself has a smaller, better‑protected footprint. Completing SOC 2 Type 2 certification as part of the release reinforces Bonfy’s readiness for highly regulated industries. “System-level agents won’t run on endpoints you control. They’ll run in compute infrastructure provided by Microsoft, Google, OpenAI, Anthropic, Salesforce, and everywhere else your data already lives, thereby making it extremely simple and straightforward for these agents to access that data (and more) through available tools to enhance their learning with enterprise-specific data” said Vishnu Varma, Head of Product, Bonfy. “Endpoint-, browser-only DLP can’t see that world. Bonfy was built as a multi-channel platform precisely so we can protect content wherever those agents operate.” Bonfy ACS 2.0 is available immediately. RSAC 2026 attendees can schedule a live demo by contacting Vishnu Varma.
helpnetsecurity.comMar 20, 2026extracted
Data breach di Odido: quando il social engineering batte qualsiasi firewall
Immaginate un attaccante che non usa vulnerabilità zero-day, non forza nessun sistema di autenticazione e non acquista tool sofisticati sul dark web. Immaginate, invece, qualcuno che scrive un’e-mail, fa una telefonata e poi, con la sola forza della manipolazione psicologica, apre le porte di uno dei più grandi operatori telefonici d’Europa. È esattamente quello che è successo a Odido, la principale telco dei Paesi Bassi con oltre 6 milioni di clienti: una violazione che non solo espone dati sensibili di massa, ma riaccende in modo prepotente il dibattito su quanto il fattore umano rimanga il tallone d’Achille irrisolto della cyber security aziendale. Come sottolinea Sandro Sana, Ethical Hacker e membro del Comitato Scientifico Cyber 4.0, “il caso Odido dimostra, ancora una volta, che la MFA (autenticazione a più fattori) non viene “bucata”: viene gentilmente consegnata. Quando il social engineering entra in gioco, firewall e controlli tecnici diventano comparse”. Indice degli argomenti Il 14 febbraio 2026 Odido – nato dalla fusione tra T-Mobile Netherlands e Tele2 Netherlands nel 2022 – ha comunicato ufficialmente alla propria base clienti un grave incidente di sicurezza informatica. L’azienda ha notificato l’Autorità olandese per la protezione dei dati (AP) e avviato un processo di notifica che, secondo fonti interne, richiederebbe fino a due giorni per raggiungere tutti i destinatari. Il vettore iniziale dell’attacco, ricostruito da fonti vicine all’indagine e riportato dalla testata olandese NOS, segue uno schema di multi-stage social engineering ormai consolidato ma ancora drammaticamente efficace: Fase 1, phishing: nella prima fase, gli attaccanti hanno inviato email di phishing mirate a dipendenti del call center — in particolare a operatori di call center esternalizzati — riuscendo a sottrarre le relative credenziali di accesso. Fase 2, vishing ed MFA Bypass: con le credenziali in mano, ma bloccati dall’autenticazione multi-fattore (MFA), gli attaccanti hanno chiamato telefonicamente le stesse vittime spacciandosi per tecnici del reparto IT interno di Odido, convincendole ad approvare il tentativo di accesso. Fase 3, esfiltrazione dati: una volta dentro, gli attaccanti hanno avuto accesso al sistema CRM Salesforce, da cui hanno iniziato un’estrazione sistematica dei dati dei clienti tramite scraping programmatico. La tecnica descritta corrisponde perfettamente a quello che in gergo tecnico si chiama MFA fatigue attack o real-time phishing proxy: l’attaccante non rompe la MFA, la scavalca semplicemente convincendo l’utente legittimo ad approvare la richiesta di autenticazione. Una distinzione sottile ma fondamentale. Odido ha ufficialmente comunicato che il data breach ha interessato un “sistema di contatto clienti” che, secondo alcune fonti, potrebbe essere identificato con la piattaforma CRM Salesforce utilizzata dai call center. La buona notizia, se di buona notizia si può parlare, è che password, dettagli delle chiamate e dati di fatturazione non sono stati compromessi. La cattiva notizia è che per una parte significativa degli utenti notificati (stimati in circa 6,2 milioni tra clienti attuali ed ex-clienti) i dati potenzialmente esposti includono elementi ad alto rischio di sfruttamento secondario: nomi e cognomi; indirizzi di residenza e indirizzi e-mail; IBAN bancari; date di nascita; numeri di documento di identità: passaporto e/o patente di guida. Nel data breach di Odido sarebbero interessati anche i clienti del brand sussidiario Ben, operatore low-cost che eroga servizi su base contrattuale senza piani a lungo termine. La combinazione di questi dati (identità anagrafica, coordinate bancarie, documento d’identità) rappresenta un kit pressoché completo per attività di identity fraud e spear phishing. Le incognite sul furto dell’intero database clienti Secondo fonti vicine all’indagine sul data breach di Odido, non è ancora possibile escludere che gli attaccanti abbiano scaricato l’intero database clienti. Un dump completo dei dati avrebbe richiesto diversi giorni di attività non rilevata, uno scenario tecnicamente plausibile ma non ancora confermato. L’esperto di cyber security Seymen Rouvhof ha confermato: “Con questo tipo di attacco, raccogliere tutti i record richiede, in genere, giorni. Le azioni devono rimanere inosservate durante quel periodo”. È inoltre importante sottolineare che La combinazione IBAN + nome + indirizzo + numero del documento d’identità è sufficiente per: aprire conti bancari o linee di credito fraudolente; effettuare SIM swap per aggirare l’autenticazione bancaria; costruire spear phishing altamente personalizzati verso i 6,2M di vittime; impersonare Odido in fatture false o comunicazioni urgenti. Per chi lavora ogni giorno nel campo della cyber security difensiva, il data breach di Odido non è una sorpresa: è la conferma di un pattern osservabile che si ripete con cadenza sempre più frequente. Il social engineering di nuova generazione non è il classico principe nigeriano delle e-mail spam degli anni 90. È un’operazione chirurgica, scalabile e spesso automatizzata, costruita su tre pilastri psicologici fondamentali: Autorità (Authority). L’attaccante si è presentato come membro dell’IT interno, sfruttando il principio di compliance all’autorità descritto dallo psicologo statunitense Robert Cialdini nel suo libro “Le armi della persuasione”, insieme alle altre sei tendenze umane (autorevolezza, reciprocità, scarsità, coerenza, riprova sociale e simpatia). I dipendenti dei call center – spesso terzisti, poco integrati nella cultura di sicurezza dell’azienda madre – tendono a obbedire alle richieste percepite come “tecniche” provenienti dall’IT, anche se non le comprendono pienamente. Urgenza (Scarcity/Urgency). La narrativa usata durante la chiamata vishing era quasi certamente costruita intorno all’urgenza: “C’è un problema di sicurezza sul tuo account, devi approvare immediatamente questa verifica”. L’urgenza inibisce il pensiero critico e accelera la decisione impulsiva. Scarcity of Information. I dipendenti target non avevano modo rapido di verificare l’identità del chiamante. I sistemi di autenticazione interna per le telefonate aziendali sono ancora oggi raramente implementati, creando un gap di verifica che gli attaccanti sfruttano sistematicamente. È utile sottolineare, inoltre, che il vettore d’attacco descritto è riconducibile alla famiglia degli attacchi di tipo MFA bypass via social engineering. In particolare, la tecnica del “vishing post-phishing” è documentata nel framework MITRE ATT&CK alla voce T1566.004 (Spearphishing Voice) combinata con T1621 (Multi-Factor Authentication Request Generation). La particolarità di questo attacco è la fase di handoff: il phishing cattura le credenziali, mentre il vishing completa il bypass MFA in real-time. Il tutto senza alcun exploit tecnico. Una tecnica, questa, che è stata osservata anche negli attacchi contro MGM Resorts (2023), Uber (2022) e Twilio (2022): tre casi in cui il social engineering ha preceduto o sostituito l’hacking tecnico. Uno degli elementi più rilevanti per i professionisti IT emerge dal vettore di esfiltrazione nel data breach di Odido: il sistema compromesso pare sia stato Salesforce, la piattaforma CRM dominante nel mercato Enterprise. Questo dettaglio merita un’analisi specifica. I sistemi CRM come Salesforce aggregano per definizione i dati più sensibili di un’organizzazione lato customer. Ironia della sorte, più il CRM è ben implementato, con profili cliente completi, storico interazioni, dati anagrafici completi, più diventa appetibile come target di esfiltrazione massiva. Nel data breach di Odido, una volta ottenute le credenziali di un account operatore legittimo, gli attaccanti hanno potuto sfruttare le API native del CRM o semplicemente la navigazione programmatica del portale per estrarre dati in modo automatizzato: una tecnica che gli analisti definiscono una “programmatic scraping session”. “Abbiamo implementato la MFA, siamo al sicuro”. Questa frase, ripetuta in quasi ogni assessment aziendale, è diventata pericolosamente fuorviante: il caso Odido è l’ennesima dimostrazione che la MFA è necessaria, ma non sufficiente. La MFA basata su notifiche push (come quelle di Microsoft Authenticator o Google Authenticator) presuppone che l’utente sappia distinguere una notifica legittima da una fraudolenta. Presuppone, cioè, un livello di consapevolezza che, soprattutto in ambienti di call center ad alto turnover e bassa formazione specifica, non può essere dato per scontato. Gli attaccanti, nel caso Odido, hanno semplicemente chiamato il dipendente mentre la notifica push era attiva, creando un contesto narrativo convincente (“Sono dell’IT, sto configurando il tuo account, approva la notifica che ti è arrivata”) che ha portato la vittima ad approvare consapevolmente, pur non sapendo cosa stesse approvando, l’accesso non autorizzato. Per configurare un sistema di autenticazione multifattoriale resistente ai tentativi di attacchi di tipo social engineering è utile adottare le seguenti raccomandazioni: Utilizzare passkey e sistemi FIDO2: eliminano il problema alla radice. L’autenticazione è device-bound ed è crittograficamente legata al sito o all’app di destinazione, per cui non c’è alcuna notifica push da approvare. Number matching: se si usa la push MFA, è consigliabile abilitare il “number matching” per mostrare all’utente un codice da inserire nel telefono. Ciò riduce drasticamente il rischio di false approvazioni di accesso. Security awareness contestuale: è molto utile anche addestrare i dipendenti specificamente sullo scenario “qualcuno vi chiama e vi chiede di approvare una notifica”, spiegando loro come comportarsi in situazioni del genere. MFA resistente al phishing: dovrebbe essere obbligatoria per i sistemi in cui vengono trattati dati sensibili. Zero-trust per i call center terzisti: violazioni di sicurezza come quella di Odido dimostrano che il modello “trust but verify” non funziona più. I call center esternalizzati devono operare in ambienti containerizzati con visibilità ridotta e monitoraggio continuo delle sessioni. Come evidenzia ancora Sandro Sana, “il vero problema non è Salesforce né la tecnologia utilizzata, ma l’illusione che l’autenticazione multi-fattore sia una misura risolutiva senza formazione, processi e controlli sull’operato dei terzisti”. “Questo data breach”, continua l’analista, “è l’ennesima prova che il perimetro più fragile resta quello umano e che la supply chain, soprattutto nei call center esternalizzati, continua a essere il punto cieco di molte organizzazioni. La sicurezza non si compra a licenza: si governa”. Il data breach di Odido apre scenari interessanti e preoccupanti sul piano regolatorio europeo. La violazione coinvolge dati di cittadini olandesi ed europei, gestiti attraverso una piattaforma (Salesforce) e operatori di call center esternalizzati. Una catena di responsabilità complessa che il GDPR e la direttiva NIS2 devono intercettare. Odido ha già notificato l’Autorità olandese per la protezione dei dati (AP) nei tempi previsti dall’art. 33 del GDPR, entro 72 ore dall’individuazione della violazione, e la notifica ai soggetti interessati è in corso. Sul fronte NIS2, Odido come operatore di telecomunicazioni rientra tra i soggetti essenziali, con obblighi rafforzati in materia di gestione del rischio e notifica agli enti competenti. Il nodo più critico, però, riguarda i call center terzisti. Se la violazione è originata da account di dipendenti di fornitori esterni, si pone con forza la questione dell’adeguatezza dei controlli di sicurezza imposti contrattualmente ai responsabili del trattamento ai sensi dell’art. 28 GDPR. La domanda è semplice: i contratti con i call center prevedevano obblighi specifici di formazione anti-phishing e di gestione sicura dell’MFA? In caso contrario, Odido ne risponde in solido. Odido ha comunicato ai clienti di aspettarsi possibili abusi derivati dall’esposizione dei loro dati, con una specifica attenzione alle fatture false che potrebbero sembrare provenire dall’operatore stesso. Ecco i consigliamo da seguire concretamente per proteggersi da questi tentativi di truffa, sottolineando che questi stessi consigli sono validi sempre indipendentemente dal caso specifico: Verificare ogni comunicazione finanziaria ricevuta da Odido accedendo direttamente al portale Odido, senza cliccare su link nelle email o SMS. Attivare alert sulle movimentazioni bancarie collegate all’IBAN potenzialmente esposto. Monitorare eventuali richieste di credito o aperture di conti effettuate a proprio nome, richiedendo una verifica alla propria banca. Non condividere ulteriori dati sensibili a seguito di comunicazioni non verificate che si presentano come Odido. In caso di sospetto utilizzo fraudolento dei documenti esposti (passaporto, patente), contattare le autorità competenti per richiedere un alert sul proprio codice documento. Da questo incidente emergono almeno cinque lezioni strutturali che ogni responsabile della sicurezza dovrebbe portare al prossimo board meeting: Il perimetro umano è il perimetro più vulnerabile.Nessun firewall, nessun EDR, nessun SIEM protegge dall’ingegneria sociale se i dipendenti non sono adeguatamente formati. La security awareness non è un costo overhead: è un controllo di sicurezza critico con ROI misurabile (riduzione del tasso di click su phishing simulati). I terzisti sono estensioni del perimetro aziendale. Un call center esterno che gestisce dati dei clienti è, a tutti gli effetti, parte del perimetro di sicurezza aziendale. I loro dipendenti devono ricevere la stessa formazione che ricevono i vostri. I loro sistemi devono rispettare i vostri standard. Questo va negoziato contrattualmente, non dato per scontato. Il CRM è infrastruttura critica. Salesforce, HubSpot, Microsoft Dynamics: questi sistemi contengono i gioielli della corona dei dati cliente. Trattarli come applicazioni ordinarie, senza monitoraggio comportamentale dedicato e controlli di accesso granulari, è un rischio sistemico inaccettabile. La MFA si bypassa: usare MFA resistente al phishing. Come evidenziato, la push MFA tradizionale non è più sufficiente contro attaccanti motivati. La transizione verso Passkeys e FIDO2 non è più un nice-to-have: è una necessità operativa, specialmente per i sistemi che contengono dati personali di milioni di utenti. Detection veloce, prevention perfetta. Nella realtà operativa, la prevenzione assoluta è un’utopia. Il vero differenziatore è la velocità di rilevamento. Nel caso Odido, l’esfiltrazione ha richiesto potenzialmente giorni. Un sistema di rilevamento comportamentale sulle sessioni Salesforce – anomalie nel volume di record consultati, orari di accesso insoliti, geoIP anomali – avrebbe potuto bloccare l’operazione molto prima. Il data breach di Odido non sarà l’ultimo del suo genere. Anzi: con la crescente sofisticazione degli strumenti di AI generativa – che permettono di creare e-mail di phishing grammaticalmente perfette, deepfake vocali convincenti per il vishing e profili social costruiti ad hoc per aumentare la credibilità degli attaccanti – il social engineering di nuova generazione diventerà il vettore dominante degli anni a venire. La risposta non può essere solo tecnologica, ma richiede un cambio di paradigma culturale nelle organizzazioni: la sicurezza non è responsabilità esclusiva del team IT, è responsabilità di ogni singola persona che ha accesso a sistemi aziendali, dai manager ai call center, dai developer agli stagisti. Dunque, il messaggio da portare sempre ai propri clienti è questo: “Il vostro attaccante più pericoloso non è nel dark web. Probabilmente ha già il vostro numero di telefono interno“. E Odido lo ha scoperto nel modo più costoso possibile.
cybersecurity360.itFeb 18, 2026extracted
Over 100 Organizations Targeted in ShinyHunters Phishing Campaign
Many major organizations appear to have been targeted in a recent cybercrime campaign linked to the ShinyHunters group, according to security firm Silent Push. Over the past 30 days, Silent Push has identified domains suggesting that the threat actors have been preparing or conducting attacks against at least 100 organizations in sectors such as software and technology, financial, biotech and pharma, financial services, real estate, energy and utilities, healthcare, logistics and transportation, manufacturing, retail, and insurance. Silent Push has named major companies such as Atlassian, Adyen, Canva, Epic Games, HubSpot, Moderna, ZoomInfo, GameStop, WeWork, Halliburton, Sonos, and Telstra. The hackers have set up fake domains targeting these companies, but it’s unclear whether any attacks were conducted or whether their attempts to gain access to systems were successful. In the campaign, the cybercriminals used voice phishing (vishing) to target single sign-on (SSO) accounts associated with Okta and other identity platforms. In attacks observed by Okta and others, threat actors used specialized phishing kits that enable them to intercept credentials and trick victims into helping them bypass multi-factor authentication. “The most critical of these features are client-side scripts that allow threat actors to control the authentication flow in the browser of a targeted user in real-time while they deliver verbal instructions or respond to verbal feedback from the targeted user,” Okta explained. It added, “It’s this real-time session orchestration that delivers the plausibility required to convince the threat actor’s target to approve push notifications, submit one time passcodes (OTP) or take other actions the threat actor needs to bypass MFA controls.” ShinyHunters is the public-facing entity that has taken credit for the attacks, but Silent Push has attributed the campaign — based on TTPs — to Scattered LAPSUS$ Hunters, the group formed last year by Lapsus$, Scattered Spider, and ShinyHunters members. On the ShinyHunters leak website, the cybercriminals recently listed companies such as Betterment, Crunchbase, and SoundCloud, all of which have confirmed suffering a data breach. Alon Gal, CTO of threat intelligence firm Hudson Rock, learned from ShinyHunters that these are victims of the Okta SSO vishing campaign. The hackers have released millions of records allegedly stolen from these companies. Google’s Mandiant has also been tracking this campaign, which it has described as active and ongoing. “After gaining initial access, these actors pivot into SaaS environments to exfiltrate sensitive data. An actor that identifies as ShinyHunters has approached some of the victim organizations with an extortion demand,” said Charles Carmakal, CTO of Mandiant Consulting. “While this is not the result of a security vulnerability in vendors’ products or infrastructure, we strongly recommend moving toward phishing-resistant MFA, such as FIDO2 security keys or passkeys where possible, as these protections are resistant to social engineering attacks in ways that push-based or SMS authentication are not. Administrators should also implement strict app authorization policies and monitor logs for anomalous API activity or unauthorized device enrollments,” Carmakal added. *updated with statement from Charles Carmakal Related: Google Warns of Vishing, Extortion Campaign Targeting Salesforce Customers Related: Organizations Warned of Rise in Okta Support Phishing Attacks Related: Security Industry Skeptical of Scattered Spider-ShinyHunters Retirement Claims
securityweek.comJan 27, 2026extracted
The Golden Scale: 'Tis the Season for Unwanted Gifts
In October 2025, we published two Insights blogs on threat activity affiliated with the cybercriminal alliance known as Scattered LAPSUS$ Hunters (SLSH). After a few weeks of apparent inactivity, the threat actors have returned with a vengeance based on open-source reporting and conversations obtained from a new Telegram channel (scattered LAPSUS$ hunters part 7). This latest Insights threat blog will detail several notable observations made by Unit 42 since mid-November, and prepares organizations as we head into the holiday season. On Nov. 20, 2025, Salesforce released a security advisory acknowledging that they had detected “unusual activity involving Gainsight-published applications.” This led the company to revoke “all active access and refresh tokens associated with Gainsight-published applications” while also temporarily removing such applications from their AppExchange while they conduct an investigation. At the time of this writing time, Salesforce assesses that the activity was not a result of any vulnerability in their platform and that “this activity may have enabled unauthorized access to certain customers’ Salesforce data through the app’s connection.” The company has notified all impacted customers and issued an additional advisory on Nov. 22, 2025 with a number of indicators of compromise (IoCs) related to this activity. Based on BleepingComputer’s reporting, Bling Libra (aka ShinyHunters) claimed to have gained access to an additional 285 Salesforce instances by breaching Gainsight. The threat group asserted they accomplished this using secrets obtained via their supply chain attack targeting Salesloft Drift in August 2025, which Unit 42 previously reported on Sep. 10, 2025. Gainsight acknowledged on Sept. 3, 2025 that they were breached via stolen OAuth tokens linked to the Salesloft Drift attack. In this security alert the company confirmed the following types of information were likely accessed by the threat actors: Names Business email addresses Phone numbers Regional/location details Gainsight product licensing information Plain text content from certain support cases (not including attachments) On Nov. 20, 2025, SLSH representatives posted a message within their newly created Telegram channel. It included an image that appears to represent a new dedicated leak site (DLS) with text reading “24 November 2025, stay tuned” as shown in Figure 1. This seemingly implies a deadline set for any companies affected by this latest data theft campaign to pay a ransom. On Nov. 21, 2025, SLSH posted another message shown in Figure 2, which functions as a warning to companies that have not yet been affected by their Salesforce data theft campaigns. On Nov. 19, 2025, BleepingComputer reported on a new ransomware-as-a-service (RaaS) program dubbed “ShinySp1d3r” which is allegedly still under active development by SLSH. The ransomware currently only works on Windows systems but representatives for the criminal syndicate told reporters that they are close to producing versions for Linux and ESXi systems. Unit 42 previously alluded to the development of ShinySp1d3r ransomware in our last Insights blog on SLSH. Additionally, last week, we also published timely threat intelligence on our research into IoCs likely associated with this form of ransomware. Figures 3 and 4 provide further information on the encryptor portion of ShinySp1d3r upon successful execution. On Nov. 21, 2025, SLSH posted another Telegram message shown in Figure 5 where they threaten to deploy ShinySp1d3r ransomware for all of New York City and the State of New York. On Nov. 21, 2025, CrowdStrike confirmed to BleepingComputer that an employee had shared screenshots of internal systems with SLSH which were then posted to the group’s Telegram channel. CrowdStrike asserted that the individual was terminated last month and that none of its systems were breached as a result of this activity. Bling Libra confirmed to reporters that they agreed to pay the insider $25,000 for access to CrowdStrike’s network. On the same day, SLSH posted several more Telegram messages further illustrated in Figures 6 and 7. The first image shown below highlights the industries that the threat actors were looking to solicit insiders from, which includes retail and hospitality organizations. The second image shown below illustrates how the threat actors are attempting to calm any unease that potential insiders may be feeling in the aftermath of CrowdStrike’s insider detection. On Nov. 24, 2025, Gainsight announced that connections to other SaaS platforms such as HubSpot and Zendesk were being temporarily suspended due to the supply chain attack. The company also encouraged customers to rotate their S3 keys as a precautionary measure. At time of publication, Unit 42 had yet to identify any communications by the threat actors claiming to have leaked information related to their alleged Gainsight data theft campaign. However, they did post the following message to their Telegram channel on Nov. 24, 2025: “pretty sure the 2025 victim count by us in total is ~1.5k (1000 already publicly reported) and still increasing” My overall prediction when it comes to these financially-motivated threat actors in 2026 and beyond is more of the same: unwavering chaos. We previously expected SLSH to take a break and reemerge at the beginning of the new calendar year with the aforementioned activities, but they have seemingly decided to expedite that timeline based on these latest observations. The emergence of a RaaS program, in conjunction with an EaaS offering, makes SLSH a formidable adversary in terms of the wide net they can cast against organizations using multiple methods to monetize their intrusion operations. Additionally, the insider recruitment element adds yet another layer for organizations to defend against. The timing of these developments could not be worse for most organizations, especially retailers, as they ramp up for the biggest shopping weeks of the calendar year. Figure 8 provides more insight on how the threat actors plan to operate in the coming weeks, which seemingly alludes to more customer data potentially being leaked to their DLS. Palo Alto Networks recently predicted that 2026 will be the “Year of the Defender” with regards to applying AI-driven defenses to combat AI-powered attacks. I strongly believe that this sentiment of 2026 being the year of the defender also needs to hold true if we are to collectively defeat the many fronts that SLSH is targeting organizations from. One of the best gifts you can give your organization this time of year is joining and actively participating in an industry-specific Information Sharing and Analysis Center — this enables your network defenders to learn from other peer institutions and collectively shift the outcome to “left of bang.” Unit 42 is ready to help support your organization with an active compromise or to provide a proactive assessment to lower your organization's risk related to this evolving threat activity.
unit42.paloaltonetworks.comNov 26, 2025extracted
Salesforce Gainsight compromise: Early findings and customer guidance
Salesforce Gainsight compromise: Early findings and customer guidance In the wake of Salesforce’s announcement about “unusual activity involving Gainsight-published applications” and the company’s revocation of access and refresh tokens associated with them, Gainsight has been doing a good job keeping customers updated on current investigation findings. On the status page following the incident, the company confirmed that, as a precautionary measure, the Gainsight app has been temporarily pulled from the Hubspot Marketplace and Zendesk connector access has been revoked. They have also shared that: The suspicious activity Salesforce detected were API calls using the Gainsight Connected App coming from non-whitelisted IPs. Salesforce has yet to confirm data exfiltration from customer instances. They haven’t seen attackers sending out phishing emails or using bulk email features. “At the moment only three orgs are known to be impacted. The Gainsight Salesforce connection should be the only impacted product,” an employee with the company noted earlier today. Customers should open a support ticket to request the IP ranges/subnets that Salesforce login events from the Gainsight connector should originate from. If they want to review API calls, they should request Salesforce logs from that company. “Gainsight, Salesforce, and a third-party forensics firm are jointly reviewing all security layers. They will not restore API access until fully cleared,” the employee added. “Our third-party will issue a formal report and any remediation guidance. Gainsight will likely move to a packaged version of the Connected App to ensure a clean and secure reset. While no one can guarantee absolute protection, we will only turn services back on once fully vetted.” Affected customers are being notified by Salesforce and Mandiant (the “third-party forensics firm”). While several things point to the attackers being the ShinyHunters cyber extortion collective – and they confirmed to DataBreaches.net their involvement – an official attribution of the attack has yet to be made. They group claims to have exfiltrated data from affected Salesforce instances of many companies, including “Verizon, Gitlab, F5, Sonicwall, and others”. Subscribe to our breaking news e-mail alert to never miss out on the latest breaches, vulnerabilities and cybersecurity threats. Subscribe here!
helpnetsecurity.comNov 21, 2025extracted
OpenAI confirms GPT-5 is now better at handling mental and emotional distress
OpenAI confirmed that it shipped an update on October 5, which allows GPT-5 to better handle sensitive conversations. After the update, GPT will automatically recognize when the user is not doing well, including situations like emotional or mental distress. However, OpenAI says only GPT-5 Instant has been updated to handle such use cases. GPT-5 Instant is one of the fastest and low-end models in the lineup. It's also the default model most of the time, especially if you have a free account. But what if you're conversing with a GPT-5-thinking model and it detects emotional distress or a related topic? It'll automatically route your request to GPT-5 Instant. "We've been using our real-time router to direct sensitive parts of conversations—such as those showing signs of acute distress—to reasoning models. GPT-5 Instant now performs just as well as GPT-5 Thinking on these types of questions," OpenAI noted in a new support document. In our tests, BleepingComputer previously found that GPT-5 automatically routes requests through different models depending on the question. In fact, if you choose GPT-5-instant specifically, GPT will still route your request through a thinking model if it thinks a different model can answer better. OpenAI says it worked with health experts on its new GPT-5 update OpenAI insists that these updates were guided by mental health experts and help ChatGPT de-escalate conversations. OpenAI will point people to real-world crisis resources when appropriate, while still using language that feels supportive and grounding. GPT-5 was updated with emotional support earlier this month, and the updated model is now available for everyone. In addition, OpenAI confirmed ChatGPT can now access your organization’s context from your connected apps. GPT's Company Knowledge support works through Slack, SharePoint, Google Drive, GitHub, HubSpot, Asana, and other connectors. This is similar to how Microsoft 365 Copilot works. Like Microsoft's AI solution, GPT can now provide answers tailored to your company and projects, as it has access to all relevant information. Overall prevention scores can hide what happens after initial access. Once attackers are using valid credentials, prevention drops sharply. The Blue Report 2026 measures defenses technique by technique across 338 million simulations run in customer production environments. Get the report
bleepingcomputer.comOct 30, 2025extracted
Firezone: Open-source platform to securely manage remote access
Firezone: Open-source platform to securely manage remote access Firezone is an open-source platform that helps organizations of any size manage secure remote access. Unlike most VPNs, it uses a least-privileged model, giving users only the access they need. Firezone was built to scale from the start, so you can add more gateways as traffic grows. It uses WireGuard, a fast and secure VPN protocol, and adds extra protections like short-lived encryption keys and firewall hole-punching to reduce the attack surface. It is also simple to manage, with a Policy Engine that removes the need for complex firewall rules or ACLs and makes it easier to control and review access. Firezone can help solve many common secure access problems. It can block DNS queries to known malicious domains to protect your team from harmful sites. You can scale access to a VPC by adding multiple gateways, with an optional Terraform example to help with setup. It also allows routing certain traffic through a single static IP, which is useful when a service requires requests to come from approved IP addresses. It can secure access to databases like Postgres, manage access to public SaaS apps such as HubSpot or GitHub, and connect to hosts by their private IP addresses. With Firezone, you can reach private networks like homelabs or VPCs without opening firewall ports and securely access privately hosted web apps such as GitLab or Metabase. Firezone is available for free on GitHub. Must read: 35 open-source security tools to power your red team, SOC, and cloud security GitHub CISO on security strategy and collaborating with the open-source community Subscribe to the Help Net Security ad-free monthly newsletter to stay informed on the essential open-source cybersecurity tools. Subscribe here!
helpnetsecurity.comSep 29, 2025extracted
ShadowLeak Zero-Click Flaw Leaks Gmail Data via OpenAI ChatGPT Deep Research Agent
Cybersecurity researchers have disclosed a zero-click flaw in OpenAI ChatGPT's Deep Research agent that could allow an attacker to leak sensitive Gmail inbox data with a single crafted email without any user action. The new class of attack has been codenamed ShadowLeak by Radware. Following responsible disclosure on June 18, 2025, the issue was addressed by OpenAI in early August. "The attack utilizes an indirect prompt injection that can be hidden in email HTML (tiny fonts, white-on-white text, layout tricks) so the user never notices the commands, but the agent still reads and obeys them," security researchers Zvika Babo, Gabi Nakibly, and Maor Uziel said. "Unlike prior research that relied on client-side image rendering to trigger the leak, this attack leaks data directly from OpenAI's cloud infrastructure, making it invisible to local or enterprise defenses." Launched by OpenAI in February 2025, Deep Research is an agentic capability built into ChatGPT that conducts multi-step research on the internet to produce detailed reports. Similar analysis features have been added to other popular artificial intelligence (AI) chatbots like Google Gemini and Perplexity over the past year. In the attack detailed by Radware, the threat actor sends a seemingly harmless-looking email to the victim, which contains invisible instructions using white-on-white text or CSS trickery that tell the agent to gather their personal information from other messages present in the inbox and exfiltrate it to an external server. Thus, when the victim prompts ChatGPT Deep Research to analyze their Gmail emails, the agent proceeds to parse the indirect prompt injection in the malicious email and transmit the details in Base64-encoded format to the attacker using the tool browser.open(). "We crafted a new prompt that explicitly instructed the agent to use the browser.open() tool with the malicious URL," Radware said. "Our final and successful strategy was to instruct the agent to encode the extracted PII into Base64 before appending it to the URL. We framed this action as a necessary security measure to protect the data during transmission." The proof-of-concept (PoC) hinges on users enabling the Gmail integration, but the attack can be extended to any connector that ChatGPT supports, including Box, Dropbox, GitHub, Google Drive, HubSpot, Microsoft Outlook, Notion, or SharePoint, effectively broadening the attack surface. Unlike attacks like AgentFlayer and EchoLeak, which occur on the client-side, the exfiltration observed in the case of ShadowLeak transpires directly within OpenAI's cloud environment, while also bypassing traditional security controls. This lack of visibility is the main aspect that distinguishes it from other indirect prompt injection vulnerabilities similar to it. ChatGPT Coaxed Into Solving CAPTCHAs The disclosure comes as AI security platform SPLX demonstrated that cleverly worded prompts, coupled with context poisoning, can be used to subvert ChatGPT agent's built-in guardrails and solve image-based CAPTCHAs designed to prove a user is human. The attack essentially involves opening a regular ChatGPT-4o chat and convincing the large language model (LLM) to come up with a plan to solve what's described to it as a list of fake CAPTCHAs. In the next step, a new ChatGPT agent chat is opened and the earlier conversation with the LLM is pasted, stating this was "our previous discussion" – effectively causing the model to solve the CAPTCHAs without any resistance. "The trick was to reframe the CAPTCHA as "fake" and to create a conversation where the agent had already agreed to proceed. By inheriting that context, it didn’t see the usual red flags," security researcher Dorian Schultz said. "The agent solved not only simple CAPTCHAs but also image-based ones -- even adjusting its cursor to mimic human behavior. Attackers could reframe real controls as 'fake' to bypass them, underscoring the need for context integrity, memory hygiene, and continuous red teaming."
thehackernews.comSep 20, 2025extracted
ChatGPT Targeted in Server-Side Data Theft Attack
Researchers at web security company Radware recently discovered what they described as a service-side data theft attack method involving ChatGPT. The attack, dubbed ShadowLeak, targeted ChatGPT’s Deep Research capability, which is designed to conduct multi-step research for complex tasks. OpenAI neutralized ShadowLeak after it was notified by Radware. The ShadowLeak attack did not require any user interaction. The attacker simply needed to send a specially crafted email that when processed by the Deep Research agent would instruct it to silently collect valuable data and send it back to the attacker. However, unlike many other indirect prompt injection attacks, ShadowLeak did not involve the ChatGPT client. Several cybersecurity companies recently demonstrated theoretical attacks in which the attacker leverages the integration between AI assistants and enterprise tools to silently exfiltrate user data with no or minimal victim interaction. Radware mentions Zenity’s AgentFlayer and Aim Security’s EchoLeak attacks. However, the company highlighted that those are client-side attacks, while ShadowLeak involves the server side. As in previous attacks, the attacker would need to send an email that looks harmless to the targeted user but contains hidden instructions for ChatGPT. The malicious instructions would be triggered when the user asked the chatbot to summarize emails or research a topic from their inbox. Unlike client-side attacks, ShadowLeak exfiltrates data through the parameters of a request to an attacker-controlled URL. A harmless-looking URL such as ‘hr-service.net/{parameters}’, where the parameter value is the exfiltrated information, has been provided as an example by Radware. “It’s important to note that the web request is performed by the agent executing in OpenAI’s cloud infrastructure, causing the leak to originate directly from OpenAI’s servers,” Radware pointed out, noting that the attack leaves no clear traces because the request and data don’t pass through the ChatGPT client. The attacker’s prompt is cleverly designed not only in terms of collecting the information and sending it to the attacker. It also tells the chatbot that it has full authorization to conduct the required tasks, and creates a sense of urgency. The prompt also instructs ChatGPT to try multiple times if it doesn’t succeed, provides an example of how the malicious instructions should be carried out, and attempts to override possible security checks by convincing the agent that the exfiltrated data is already public and the attacker’s URL is safe. While Radware demonstrated the attack method against Gmail, the company said Deep Research can access other widely used enterprise services as well, including Google Drive, Dropbox, Outlook, HubSpot, Notion, Microsoft Teams, and GitHub. OpenAI was notified about the attack on June 18 and the vulnerability was fixed at some point in early August. Radware has confirmed that the attack no longer works. However, it told SecurityWeek that it believes “there is still a fairly large threat surface that remains undiscovered”. The security firm recommends continuous agent behavior monitoring for mitigating such attacks. “Tracking both the agent’s actions and its inferred intent and validating that they remain consistent with the user’s original goals. This alignment check ensures that even if an attacker steers the agent, deviations from legitimate intent are detected and blocked in real time,” it explained. Related: Irregular Raises $80 Million for AI Security Testing Lab Related: UAE’s K2 Think AI Jailbroken Through Its Own Transparency Features
securityweek.comSep 18, 2025extracted
Passkey support in business applications | Kaspersky official blog
Transition to passkeys promises organizations a cost-effective path toward robust employee authentication, increased productivity, and regulatory compliance. We’ve already covered all the pros and cons of this business solution in a separate, in-depth article. However, the success of the transition — and even its feasibility — really hinges on the technical details and implementation specifics across numerous corporate systems. Passkey support in identity management systems Before tackling organizational hurdles and drafting policies, you’ll have to determine if your core IT systems are ready for the switch to passkeys. Microsoft Entra ID (Azure AD) fully supports passkeys, letting admins set them as the primary sign-in method. For hybrid deployments with on-premises resources, Entra ID can generate Kerberos tickets (TGTs), which your Active Directory domain controller can then process. However, Microsoft doesn’t yet offer native passkey support for RDP, VDI, or on-premises-only AD sign-ins. That said, with a few workarounds, organizations can store passkeys on a hardware token like a YubiKey. This kind of token can simultaneously support both the traditional PIV (smart cards) technology and FIDO2 (passkeys). There are also third-party solutions for these scenarios, but you’ll need to evaluate how using them impacts your overall security posture and regulatory compliance. Good news for Google Workspace and Google Cloud users: they offer full passkey support. Popular identity management systems like Okta, Ping, Cisco Duo, and RSA IDplus also support FIDO2 and all major forms of passkeys. Passkey support on client devices We have a detailed post on the subject. All modern operating systems from Google, Apple, and Microsoft support passkeys. However, if your company uses Linux, you’ll likely need extra tools, and overall support is still limited. Also, while for all major operating systems it might look like full support on the surface, there’s a lot of variety in how passkeys are stored, and that can lead to compatibility headaches. Combinations of several systems like Windows computers and Android smartphones are the most problematic. You might create a passkey on one device and then find you can’t access it on another. For companies with a strictly managed device fleet, there are a couple of ways to tackle this. For example, you could have employees generate a separate passkey for each company device they use. This means a bit more initial setup: employees will need to go through the same process of creating a passkey on every device. However, once that’s done, signing in takes minimal time. Plus, if they lose one device, they won’t be completely locked out of their work data. Another option is to use a company-approved password manager to store and sync passkeys across all employees’ devices. This is also a must for companies using Linux computers, as its operating system can’t natively store passkeys. Just a heads-up: this approach might add some complexity when it comes to regulatory compliance audits. If you’re looking for a solution with almost no issues with sync and multiple platforms, hardware passkeys like the YubiKey are the way to go. The catch is that they can be significantly more expensive to deploy and manage. Passkey support in business applications The ideal scenario for bringing passkeys into your business apps is to have all your applications sign in through single sign-on (SSO). That way, you only need to implement passkey support in your corporate SSO solution, such as Entra ID or Okta. However, if some of your critical business applications don’t support SSO, or if that support isn’t part of your contract (which, unfortunately, happens), you’ll have to issue individual passkeys for users to sign in to each separate system. Hardware tokens can store anywhere from 25 to 100 passkeys, so your main extra cost here would be on the administrative side. Popular business systems that fully support passkeys include Adobe Creative Cloud, AWS, GitHub, Google Workspace, HubSpot, Office 365, Salesforce, and Zoho. Some SAP systems also support passkeys. Employee readiness Rolling out passkeys means getting your team up to speed regardless of the scenario. You don’t want them scratching their heads trying to figure out new interfaces. The goal is for everyone to feel confident using passkeys on every single device. Here are the key things your employees will need to understand. Why passkeys beat passwords (they’re much more secure, faster to sign in with, and don’t need to be rotated) How biometrics work with passkeys (the biometric data never leaves the device, and isn’t stored or processed by the employer) How to get their very first passkey (for example, Microsoft has a Temporary Access Pass feature, and third-party IAM systems often send an onboarding link; the process needs to be thoroughly documented, though) What to do if their device doesn’t recognize their passkey What to do if they lose a device (sign in from another device that has its own passkey, or use an OTP, perhaps given to them in a sealed envelope for just such an emergency) How to sign in to work systems from other computers (if the company’s policies permit it) What a passkey-related phishing attempt might look like Passkeys are no silver bullet Moving to passkeys doesn’t mean your cybersecurity team can just cross identity threats off their risk list. Sure, it makes things tougher for attackers, but they can still do the following: Target systems that haven’t switched to passkeys Go after systems that still have fallback login methods like passwords and OTPs Steal authentication tokens from devices infected with infostealers Use special techniques to bypass passkey protections While it’s impossible to phish the passkey itself, attackers can set up fake web infrastructure to trick a victim into authenticating and validating a malicious session on a corporate service. A recent example of this kind of AiTM attack was documented in the U.S. In that incident, the victim was lured to a fake authentication page for a corporate service, where attackers first phished their username and password, and then the session confirmation by having them scan a QR code. In this incident, the security policies were configured correctly, so scanning this QR code did not lead to successful authentication. But since such a mechanism with passkeys was implemented, the attackers hope that somewhere it is configured incorrectly, and the physical proximity of the device on which authentication is carried out and the device where the key is stored is not checked. Ultimately, switching to passkeys requires detailed policy configuration. This includes both authentication policies (such as disabling passwords when a passkey is available, or banning physical tokens from unknown vendors) and monitoring policies (such as logging passkey registrations or cross-device scenarios from suspicious locations).
kaspersky.comJul 30, 2025extracted