#git

Summary The growing adoption of large language models (LLMs) in enterprise workflows has introduced a new class of adversarial techniques: indirect prompt injection. Indirect prompt injection can be used against systems that leverage large language models (LLMs) to process untrusted data. Fundamentally, the risk is that an attacker could provide specially crafted data that the LLM misinterprets as instructions.

Node-SAML loads the assertion from the (unsigned) original response document. This is different than the parts that are verified when checking signature. This allows an attacker to modify authentication details within a valid SAML assertion. For example, in one attack it is possible to remove any character from the SAML assertion username. To conduct the attack an attacker would need a validly signed document from the identity provider (IdP). In fixing this we made sure to process the SAML assertions from only verified/authenticated contents. This will prevent future variants from coming up. Note: this is distinct from the previous xml-crypto CVEs.

The education sector is haunted by a significant fraud problem where fake students impersonate celebrities and employ other identity techniques to steal resources and money from legitimate students.

A new report from Google's GTIG reveals how UNC3944 (0ktapus) uses social engineering to compromise Active Directory, then exploits VMware vSphere for data theft and direct ransomware deployment. Understand their tactics and learn vital mitigation steps.

In what's the latest instance of a software supply chain attack, unknown threat actors managed to compromise Toptal's GitHub organization account and leveraged that access to publish 10 malicious packages to the npm registry. The packages contained code to exfiltrate GitHub authentication tokens and destroy victim systems, Socket said in a report published last week. In addition, 73 repositories

### Summary An unauthenticated attacker is able to execute arbitrary JavaScript code in a victim's browser due to improper sanitization of multimedia tags in music files, including `m3u` files. ### Details Multimedia metadata is rendered in the web-app without sanitization. This can be exploited in two ways: * a user which has the necessary permission for uploading files can upload a song with an artist-name such as `<img src=x onerror=alert(document.domain)>` * an unauthenticated user can trick another user into clicking a malicious URL, performing this same exploit using an externally-hosted m3u file The CVE score and PoC is based on the m3u approach, which results in a higher severity. ### PoC 1. Create a file named `song.m3u` with the following content. Host this file on an attacker-controlled web server. ```m3u #EXTM3U #EXTINF:1,"><img src=x onerror=alert(document.domain)> - "><img src=x onerror=alert(document.domain)> http://example.com/audio.mp3 ``` ...

### Impact This vulnerability affects applications that: * Use the ImageMagick handler for image processing (`imagick` as the image library) * **AND** either: * Allow file uploads with user-controlled filenames and process uploaded images using the `resize()` method * **OR** use the `text()` method with user-controlled text content or options An attacker can: * Upload a file with a malicious filename containing shell metacharacters that get executed when the image is processed * **OR** provide malicious text content or options that get executed when adding text to images ### Patches Upgrade to v4.6.2 or later. ### Workarounds * **Switch to the GD image handler** (`gd`, the default handler), which is not affected by either vulnerability * **For file upload scenarios**: Instead of using user-provided filenames, generate random names to eliminate the attack vector with `getRandomName()` when using the `move()` method, or use the `store()` method, which automatically generates safe ...

Some risks don’t breach the perimeter—they arrive through signed software, clean resumes, or sanctioned vendors still hiding in plain sight. This week, the clearest threats weren’t the loudest—they were the most legitimate-looking. In an environment where identity, trust, and tooling are all interlinked, the strongest attack path is often the one that looks like it belongs. Security teams are

Sublime Security reveals a cunning romance/adult-themed scam targeting German speakers, leveraging Keitaro TDS to deliver an AutoIT-based malware loader. Learn how this sophisticated campaign operates, its deceptive tactics, and the hidden payload.

A list of topics we covered in the week of July 21 to July 27 of 2025