Over the past year, Phishguard observed an increase in phishing campaigns leveraging Scalable Vector Graphics (SVG) files as initial delivery vectors, with attackers favoring this format due to its flexibility and the challenges it presents for static detection.
SVGs are an XML-based format designed for rendering two-dimensional vector graphics. Unlike raster formats like JPEGs or PNGs, which rely on pixel data, SVGs define graphics using vector paths and mathematical equations, making them infinitely scalable without loss of quality. Their markup-based structure also means they can be easily searched, indexed, and compressed, making them a popular choice in modern web applications.
However, the same features that make SVGs attractive to developers also make them a highly flexible - and dangerous - attack vector when abused. Since SVGs are essentially code, they can embed JavaScript and interact with the Document Object Model (DOM). When rendered in a browser, they aren’t just images - they become active content, capable of executing scripts and other manipulative behavior. In other words, SVGs are more than just static images; they are also programmable documents.
The security risk is underestimated, with SVGs frequently misclassified as innocuous image files, similar to PNGs or JPEGs - a misconception that downplays the fact that they can contain scripts and active content. Many security solutions and email filters fail to deeply inspect SVG content beyond basic MIME-type checks (a tool that identifies the type of a file based on its contents), allowing malicious SVG attachments to bypass detection.
We’ve seen a rise in the use of crafted SVG files in phishing campaigns. These attacks typically fall into three categories:
Redirectors - SVGs that embed JavaScript to automatically redirect users to credential harvesting sites when viewed
Self-contained phishing pages - SVGs that contain full phishing pages encoded in Base64, rendering fake login portals entirely client-side
DOM injection & script abuse - SVGs embedded into trusted apps or portals that exploit poor sanitisation and weak Content Security Policies (CSPs), enabling them to run malicious code, hijack inputs, or exfiltrate sensitive data
Given the capabilities highlighted above, attackers can now use SVGs to:
Gain unauthorized access to accounts
Create hidden mail rules
Phish internal contacts
Steal sensitive data
Initiate fraudulent transactions
Maintain long-term access
Our telemetry shows that manufacturing and industrial sectors are taking the brunt of these SVG-based phishing attempts, contributing to over half of all targeting observed. Financial services follow closely behind, likely due to SVG’s ability to easily facilitate the theft of banking credentials and other sensitive data. The pattern is clear: attackers are concentrating on business sectors that handle high volumes of documents or frequently interact with third parties.
AhnLab SEcurity intelligence Center (ASEC) recently identified a phishing malware being distributed in Scalable Vector Graphics (SVG) format. SVG is an XML-based vector image file format commonly used for icons, logos, charts, and graphs, and it allows the use of CSS and JS scripts within the code. In November 2024, the ASEC Blog introduced SVG […]
In this post I’ll show you how I found a zeroday vulnerability in the Linux kernel using OpenAI’s o3 model. I found the vulnerability with nothing more complicated than the o3 API – no scaffolding, no agentic frameworks, no tool use.
Recently I’ve been auditing ksmbd for vulnerabilities. ksmbd is “a linux kernel server which implements SMB3 protocol in kernel space for sharing files over network.“. I started this project specifically to take a break from LLM-related tool development but after the release of o3 I couldn’t resist using the bugs I had found in ksmbd as a quick benchmark of o3’s capabilities. In a future post I’ll discuss o3’s performance across all of those bugs, but here we’ll focus on how o3 found a zeroday vulnerability during my benchmarking. The vulnerability it found is CVE-2025-37899 (fix here), a use-after-free in the handler for the SMB ‘logoff’ command. Understanding the vulnerability requires reasoning about concurrent connections to the server, and how they may share various objects in specific circumstances. o3 was able to comprehend this and spot a location where a particular object that is not referenced counted is freed while still being accessible by another thread. As far as I’m aware, this is the first public discussion of a vulnerability of that nature being found by a LLM.
Before I get into the technical details, the main takeaway from this post is this: with o3 LLMs have made a leap forward in their ability to reason about code, and if you work in vulnerability research you should start paying close attention. If you’re an expert-level vulnerability researcher or exploit developer the machines aren’t about to replace you. In fact, it is quite the opposite: they are now at a stage where they can make you significantly more efficient and effective. If you have a problem that can be represented in fewer than 10k lines of code there is a reasonable chance o3 can either solve it, or help you solve it.
Benchmarking o3 using CVE-2025-37778
Lets first discuss CVE-2025-37778, a vulnerability that I found manually and which I was using as a benchmark for o3’s capabilities when it found the zeroday, CVE-2025-37899.
CVE-2025-37778 is a use-after-free vulnerability. The issue occurs during the Kerberos authentication path when handling a “session setup” request from a remote client. To save us referring to CVE numbers, I will refer to this vulnerability as the “kerberos authentication vulnerability“.
The Likely Exploited Vulnerabilities (LEV) equations can help augment KEV- and EPSS-based remediation prioritization.
Researchers from CISA and NIST have proposed a new cybersecurity metric designed to calculate the likelihood that a vulnerability has been exploited in the wild.
Peter Mell of NIST and Jonathan Spring of CISA have published a paper describing equations for what they call Likely Exploited Vulnerabilities, or LEV.
Thousands of vulnerabilities are discovered every year in software and hardware, but only a small percentage are ever exploited in the wild.
Knowing which vulnerabilities have been exploited or predicting which flaws are likely to be exploited is important for organizations when trying to prioritize patching.
Known Exploited Vulnerabilities (KEV) lists such as the one maintained by CISA and the Exploit Prediction Scoring System (EPSS), which relies on data to estimate the probability that a vulnerability will be exploited, can be very useful. However, KEV lists may be incomplete and EPSS may be inaccurate.
LEV aims to enhance — not replace — KEV lists and EPSS. This is done through equations that take into account variables such as the first date when an EPSS score is available for a specified vulnerability, the date of the most recent KEV list update, inclusion in KEV, and the EPSS score for a given day (measured across multiple days).
LEV probabilities can be useful for measuring the expected number and proportion of vulnerabilities that threat actors have exploited.
It can also be useful for estimating the comprehensiveness of KEV lists. “Previously, KEV maintainers had no metric to demonstrate how close their list was to including all relevant vulnerabilities,” the researchers explained.
In addition, LEV probabilities can help augment KEV- and EPSS-based vulnerability remediation prioritization — in the case of KEV by identifying higher-probability vulnerabilities that may be missing, and in the case of EPSS by finding vulnerabilities that may be underscored.
While in theory LEV could turn out to be a very useful tool for vulnerability prioritization, the researchers pointed out that collaboration is necessary, and NIST is looking for industry partners “with relevant datasets to empirically measure the performance of LEV probabilities”.
Malicious npm packages targeting React, Vue, Vite, Node.js, and Quill remained undetected for two years while deploying destructive payloads.
Socket's Threat Research Team discovered a collection of malicious npm packages that deploy attacks against widely-used JavaScript frameworks including React, Vue.js, Vite, Node.js, and the open source Quill Editor. These malicious packages have remained undetected in the npm ecosystem for more than two years, accumulating over 6,200 downloads. Masquerading as legitimate plugins and utilities while secretly containing destructive payloads designed to corrupt data, delete critical files, and crash systems, these packages remained undetected.
The threat actor behind this campaign, using the npm alias xuxingfeng with a registration email 1634389031@qq[.]com, has published eight packages designed to cause widespread damage across the JavaScript ecosystem. As of this writing, these packages remain live on the npm registry. We have formally petitioned for their removal.
Notably, the same account has also published several legitimate, non-malicious packages that function as advertised. This dual approach of releasing both harmful and helpful packages creates a facade of legitimacy that makes malicious packages more likely to be trusted and installed.
Dramatic revelations shed fresh light on investigation into whether Chinese tech firm tried to buy influence in EU politics.
Belgian security agents bugged a corporate box at the RSC Anderlecht football stadium that was being used by Chinese tech giant Huawei to schmooze members of the European Parliament.
They also listened into other conversations involving one of Huawei’s leading lobbyists, including in his car. The surveillance operations, confirmed by three people with close knowledge of the investigation, formed part of a wide-ranging probe into allegations of corruption that was first revealed in March. They contributed to the Belgian prosecutor’s decision, reported by POLITICO on Monday, to request that a group of MEPs have their immunities lifted so they can be investigated.
The extraordinary revelations are the latest chapter in a saga that combines concerns about the reach of China in European politics and how susceptible EU lawmakers are to bribery and shady lobbying practices, even after a string of similar scandals.
Enterprise management solutions provider Serviceaide has informed the Department of Health and Human Services (HHS) that a data leak impacts the personal and medical information of nearly half a million Catholic Health patients.
California-based Serviceaide, whose solutions are used by organizations worldwide, discovered in November 2024 that an Elasticsearch database maintained for one of its customers, Buffalo, New York-based non-profit healthcare system Catholic Health, had been inadvertently made publicly available.
An investigation showed that the database had been exposed between September 19 and November 5, 2024.
While Serviceaide did not find any evidence that the information was exfiltrated, the company said it cannot definitively rule it out.
According to a data breach notice posted on the Serviceaide website, the exposed information varies for each individual, but it can include name, SSN, date of birth, medical record number, patient account number, medical information, health insurance information, prescription and treatment information, clinical information, healthcare provider details, email or username, and password.
Impacted individuals are being notified and offered 12 months of free credit monitoring and identity theft protection services.
Serviceaide informed the HHS, according to the government organization’s incident tracker, that just over 483,000 individuals are impacted by the data breach.
It’s not uncommon for healthcare data breaches to impact hundreds of thousands of individuals, and some incidents affect millions and even tens of millions.
Arla Foods has confirmed to BleepingComputer that it was targeted by a cyberattack that has disrupted its production operations.
The Danish food giant clarified that the attack only affected its production unit in Upahl, Germany, though it expects this will result in product delivery delays or even cancellations.
"We can confirm that we have identified suspicious activity at our dairy site in Upahl that impacted the local IT network," stated an Arla spokesperson.
"Due to the safety measures initiated as a result of the incident, production was temporarily affected."
Arla Foods is an international dairy producer and a farmer-owned cooperative with 7,600 members. It employs 23,000 people in 39 countries.
The firm has an annual revenue of €13.8 billion ($15.5 billion), and its products, including the brands Arla, Lurpak, Puck, Castello, and Starbucks, are sold in 140 countries worldwide.
The company told BleepingComputer that it is currently working to resume operations at the impacted facility, which should bring results before the end of the week.
"Since then, we've been working diligently to restore full operations. We expect to return to normal operations at the site in the next few days. Production at other Arla sites is not affected."
Considering that the first reports about a disruption at Arla's production operations surfaced on Friday, it is bound to cause shortages in some cases.
"We have informed our affected customers about possible delivery delays and cancellations," explained Arla's spokesperson.
BleepingComputer has asked the firm if the attack involved data theft or encryption, both staples of a ransomware attack, but Arla declined to share any additional information at this time.
Meanwhile, there have been no announcements about Arla on ransomware extortion portals, so the type of attack and the perpetrators remain unknown.
Microsoft’s head of security for AI, Neta Haiby, accidentally revealed confidential messages about Walmart’s use of Microsoft’s AI tools during a Build talk that was disrupted by protesters.
The Build livestream was muted and the camera pointed down, but the session resumed moments later after the protesters were escorted out. In the aftermath, Haiby then accidentally switched to Microsoft Teams while sharing her screen, revealing confidential internal messages about Walmart’s upcoming use of Microsoft’s Entra and AI gateway services.
Haiby was co-hosting a Build session on best security practices for AI, alongside Sarah Bird, Microsoft’s head of responsible AI, when two former Microsoft employees disrupted the talk to protest against the company’s cloud contracts with the Israeli government.
“Sarah, you are whitewashing the crimes of Microsoft in Palestine, how dare you talk about responsible AI when Microsoft is fueling the genocide in Palestine,” shouted Hossam Nasr, an organizer with the protest group No Azure for Apartheid, and a former Microsoft employee who was fired for holding a vigil outside Microsoft’s headquarters for Palestinians killed in Gaza.
Walmart is one of Microsoft’s biggest corporate customers, and already uses the company’s Azure OpenAI service for some of its AI work. “Walmart is ready to rock and roll with Entra Web and AI Gateway,” says one of Microsoft’s cloud solution architects in the Teams messages. The chat session also quoted a Walmart AI engineer, saying: “Microsoft is WAY ahead of Google with AI security. We are excited to go down this path with you.”
Akamai researcher Yuval Gordon discovered a privilege escalation vulnerability in Windows Server 2025 that allows attackers to compromise any user in Active Directory (AD).
The attack exploits the delegated Managed Service Account (dMSA) feature that was introduced in Windows Server 2025, works with the default configuration, and is trivial to implement.
This issue likely affects most organizations that rely on AD. In 91% of the environments we examined, we found users outside the domain admins group that had the required permissions to perform this attack.
Although Microsoft states they plan to fix this issue in the future, a patch is not currently available. Therefore, organizations need to take other proactive measures to reduce their exposure to this attack. Microsoft has reviewed our findings and approved the publication of this information.
In this blog post, we provide full details of the attack, as well as detection and mitigation strategies.
Une base de données contenant plus de 184 millions de logins et mots de passe uniques a été découverte en libre accès sur Internet. Derrière cette fuite massive, des identifiants liés à des services grand public, des comptes bancaires, des adresses gouvernementales… et aucun moyen d’en retracer l’origine.
Début mai, le chercheur Jeremiah Fowler a découvert une base de données Elastic non protégée contenant plus de 184 millions de logins et mots de passe uniques. Poids total de cette énorme fuite : 47,42 Go. L’ensemble était hébergé sur les serveurs du fournisseur World Host Group, sans mot de passe ni chiffrement, et donc librement accessible à quiconque connaissait son adresse IP.
Une attaque informatique touche le site de l'Etat du Valais depuis jeudi à la mi-journée. Celui-ci a été déconnecté par précaution. A ce stade, aucun dommage n'a été constaté.
"Les hackers ont disposé temporairement de droits étendus sur les sites internet et intranet de l'Etat du Valais, ce qui pourrait permettre de procéder à des modifications sur le site", précise Claude-Alain Berclaz, chef du Service cantonal de l'informatique. "Ceci n'a pour l'instant pas été constaté."
Cette attaque "est la première de cette ampleur", précise-t-il. L'acte malveillant n'a pas été revendiqué.
Opération de remise en état
Selon les autorités, ces attaques n'ont pas permis d’intrusion dans d’autres systèmes de l'Etat. Les activités opérationnelles internes ne sont pas touchées.
Le canton dit "avoir pris toutes les mesures pour sécuriser au maximum son environnement informatique" et appliquer "systématiquement les bonnes pratiques de cybersécurité". Il cherche à remettre ses systèmes en fonction "le plus rapidement possible".
Trend™ Research uncovered a campaign on TikTok that uses videos to lure victims into downloading information stealers, a tactic that can be automated using AI tools.
Threat actors are now using TikTok videos that are potentially generated using AI-powered tools to socially engineer users into executing PowerShell commands under the guise of guiding them to activate legitimate software or unlock premium features. This campaign highlights how attackers are ready to weaponize whichever social media platforms are currently popular to distribute malware.
This report details the observed tactics, techniques, and procedures (TTPs), indicators of compromise (IoCs), and the potential impact of this trend.
An unknown actor has been continuously creating malicious Chrome Browser extensions since approximately February, 2024. The actor creates websites that masquerade as legitimate services, productivity tools, ad and media creation or analysis assistants, VPN services, Crypto, banking and more to direct users to install corresponding malicious extensions on Google’s Chrome Web Store (CWS). The extensions typically have a dual functionality, in which they generally appear to function as intended, but also connect to malicious servers to send user data, receive commands, and execute arbitrary code.
This joint operation targeted the sophisticated ecosystem that allowed Europol’s European Cybercrime Centre has worked with Microsoft to disrupt Lumma Stealer (“Lumma”), the world’s most significant infostealer threat.
This joint operation targeted the sophisticated ecosystem that allowed criminals to exploit stolen information on a massive scale. Europol coordinated with law enforcement in Europe to ensure action was taken, leveraging intelligence provided by Microsoft.
Between 16 March and 16 May 2025, Microsoft identified over 394 000 Windows computers globally infected by the Lumma malware. In a coordinated follow-up operation this week, Microsoft’s Digital Crimes Unit (DCU), Europol, and international partners have disrupted Lumma’s technical infrastructure, cutting off communications between the malicious tool and victims. In addition, over 1 300 domains seized by or transferred to Microsoft, including 300 domains actioned by law enforcement with the support of Europol, will be redirected to Microsoft sinkholes.
The Head of Europol’s European Cybercrime Centre, Edvardas Šileris, said: “This operation is a clear example of how public-private partnerships are transforming the fight against cybercrime. By combining Europol’s coordination capabilities with Microsoft’s technical insights, a vast criminal infrastructure has been disrupted. Cybercriminals thrive on fragmentation – but together, we are stronger.”
La Digital Crimes Unit (DCU) de Microsoft, en collaboration avec des partenaires internationaux, s’attaque à l’un des principaux outils utilisés pour dérober massivement des données sensibles, qu’elles soient personnelles ou professionnelles, à des fines cybercriminelles. Le mardi 13 mai, la DCU de Microsoft a engagé une action en justice contre Lumma Stealer (« Lumma »), un malware spécialisé dans le vol d’informations, largement utilisé par des centaines d’acteurs de la menace cyber. Lumma vole des mots de passe, des cartes de crédit, des comptes bancaires et des portefeuilles de cryptomonnaies. Cet outil a permis à des criminels de bloquer des établissements scolaires afin de récupérer une rançon, de vider des comptes bancaires et de perturber des services essentiels.
Grâce à une décision de justice rendue par le tribunal fédéral du district nord de la Géorgie, la Digital Crimes Unit (DCU) de Microsoft a procédé à la saisie et à la mise hors ligne d’environ 2 300 domaines malveillants, qui constituaient l’infrastructure centrale de Lumma. Parallèlement, le département de la Justice américain (DOJ) a démantelé la structure de commande principale du malware et perturbé les places de marché où l’outil était vendu à d’autres cybercriminels. Europol, via son Centre européen de lutte contre la cybercriminalité (EC3), ainsi que le Centre de lutte contre la cybercriminalité du Japon (JC3), ont contribué à la suspension de l’infrastructure locale de Lumma.
Critical vulnerabilities in Versa Concerto that are still unpatched could allow remote attackers to bypass authentication and execute arbitrary code on affected systems.
Three security issues, two of them critical, were publicly disclosed by researchers at the vulnerability management firm ProjectDiscovery after reporting them to the vendor and receiving no confirmation of the bugs being addressed.
Versa Concerto is the centralized management and orchestration platform for Versa Networks' SD-WAN and SASE (Secure Access Service Edge) solutions.
Three security issues, two of them critical, were publicly disclosed by researchers at the vulnerability management firm ProjectDiscovery after reporting them to the vendor and receiving no confirmation of the bugs being addressed.
Versa Concerto is the centralized management and orchestration platform for Versa Networks' SD-WAN and SASE (Secure Access Service Edge) solutions.
Threat actors are advancing AI strategies and outpacing traditional security. CXOs must critically examine AI weaponization across the attack chain.
The integration of AI into adversarial operations is fundamentally reshaping the speed, scale and sophistication of attacks. As AI defense capabilities evolve, so do the AI strategies and tools leveraged by threat actors, creating a rapidly shifting threat landscape that outpaces traditional detection and response methods. This accelerating evolution necessitates a critical examination for CXOs into how threat actors will strategically weaponize AI across each phase of the attack chain.
One of the most alarming shifts we have seen, following the introduction of AI technologies, is the dramatic drop in mean time to exfiltrate (MTTE) data, following initial access. In 2021, the average MTTE stood at nine days. According to our Unit 42 2025 Global Incident Response Report, by 2024 MTTE dropped to two days. In one in five cases, the time from compromise to exfiltration was less than 1 hour.
In our testing, Unit 42 was able to simulate a ransomware attack (from initial compromise to data exfiltration) in just 25 minutes using AI at every stage of the attack chain. That’s a 100x increase in speed, powered entirely by AI.
Recent threat activity observed by Unit 42 has highlighted how adversaries are leveraging AI in attacks:
Discover how to intercept data stolen by cybercriminals via Telegram bots and learn to use it to clarify related threat landscape.
While analyzing malware samples uploaded to ANY.RUN’s Interactive Sandbox, one particular case marked as “phishing” and “Telegram” drew the attention of our security analysts.
Although this analysis session wasn’t attributed to any known malware family or threat actor group, the analysis revealed that Telegram bots were being used for data exfiltration. This led us to apply a message interception technique for Telegram bots, previously described on the ANY.RUN blog.
The investigation resulted in a clear and practical case study demonstrating how intercepting Telegram bot communications can aid in profiling the threat actor behind a relatively obscure phishing campaign.
Key outcomes of this analysis include:
Examination and technical analysis of a lesser known phishing campaign
Demonstration of Telegram API-based data interception techniques
Collection of threat intelligence (TI) indicators to help identify the actor
Recommendations for detecting this type of threat
KrebsOnSecurity last week was hit by a near record distributed denial-of-service (DDoS) attack that clocked in at more than 6.3 terabits of data per second (a terabit is one trillion bits of data). The brief attack appears to have been…
For reference, the 6.3 Tbps attack last week was ten times the size of the assault launched against this site in 2016 by the Mirai IoT botnet, which held KrebsOnSecurity offline for nearly four days. The 2016 assault was so large that Akamai – which was providing pro-bono DDoS protection for KrebsOnSecurity at the time — asked me to leave their service because the attack was causing problems for their paying customers.
Since the Mirai attack, KrebsOnSecurity.com has been behind the protection of Project Shield, a free DDoS defense service that Google provides to websites offering news, human rights, and election-related content. Google Security Engineer Damian Menscher told KrebsOnSecurity the May 12 attack was the largest Google has ever handled. In terms of sheer size, it is second only to a very similar attack that Cloudflare mitigated and wrote about in April.
After comparing notes with Cloudflare, Menscher said the botnet that launched both attacks bears the fingerprints of Aisuru, a digital siege machine that first surfaced less than a year ago. Menscher said the attack on KrebsOnSecurity lasted less than a minute, hurling large UDP data packets at random ports at a rate of approximately 585 million data packets per second.
“It was the type of attack normally designed to overwhelm network links,” Menscher said, referring to the throughput connections between and among various Internet service providers (ISPs). “For most companies, this size of attack would kill them.”
