Cybersecurity experts from SafeBreach have revealed a series of vulnerabilities that could allow attackers to remotely delete files on a computer using Windows Defender, potentially leading to data loss and system instability.
Tomer Bar and Shmuel Cohen, seasoned security researchers at SafeBreach, presented the findings during their talk at the Black Hat conference.
Uncovering the Vulnerabilities
The research team’s goal was to trigger false positives in security systems, which led to the discovery of the vulnerability CVE-2023-24860.
.png
)
Free Webinar | Mastering WAAP/WAF ROI Analysis | Book Your Spot
By exploiting this vulnerability, the researchers demonstrated that it was possible to remotely delete critical files on Windows and Linux servers without authentication and bypass multiple security controls on fully patched servers.
The researchers employed a black box approach to extract byte signatures from Endpoint Detection and Response (EDR) systems, specifically targeting Windows Defender.
They developed a Python tool to minimize binaries into the smallest possible signature, which identified 130 unique signatures.
These signatures were then manually embedded into legitimate files to test the vulnerability.
Attack Vectors and Demonstrations
The team showcased several attack vectors, including the remote deletion of web server logs, local mailbox files in Mozilla Thunderbird, and Windows event log files.
They also demonstrated how Windows Defender could be tricked into deleting its own detection logs, a process they referred to as “self-cannibalism.”
Microsoft responded to the initial report by releasing a fix for CVE-2023-24860.
However, SafeBreach reported that the fix was incomplete and classified only as a “moderate DOS” by Microsoft, leaving several attack vectors unaddressed.
On the other hand, Kaspersky did not release a fix, stating that the issue could not be classified as a security vulnerability.
Further research led to the discovery of CVE-2023-36010, a bypass to the initial CVE-2023-24860 fix.
The researchers found that while some attack vectors were patched, others remained exploitable.
They also identified a method to bypass the patch for CVE-2023-36010 by manipulating MySQL storage engines and file formats.
Microsoft acknowledged the researchers’ disclosures and feedback, stating that they had implemented improvements to reduce the risk of false positives and data loss.
They also allowed customers to configure Defender to quarantine all remediation actions by default.
Looking to Safeguard Your Company from Advanced Cyber Threats? Deploy TrustNet to Your Radar ASAP.
