Telegram Desktop Bug Deletes Up to 800 GB of User Data on Windows Due to Spellcheck Library Error
A dangerous configuration error in Telegram Desktop for Windows led to the recursive deletion of nearly 800 GB of user data stored in the C:\custom directory. Home users reported that executable files began disappearing while folder structures remained, gradually emptying entire project and application directories.
The root cause was traced to the lib_spellcheck library responsible for handling user-added dictionary words in a file named custom. Following changes in version 7.1.0, the Windows code passed an empty path string to the spellchecker; Qt then appended /custom, resolving it to the absolute path C:\custom. When the library encountered a directory instead of the expected dictionary file, it executed removeRecursively(), systematically deleting all accessible contents.
Analysis with Process Monitor confirmed that Telegram Desktop was opening the path, traversing subdirectories, and requesting deletion rights without any malicious intent or updater malfunction. The flaw was limited to Windows builds containing the integrated spellchecker and existed in versions 7.1.0 and 7.1.1 for approximately 66 hours.
The researcher virtual-anvil submitted detailed logs and opened ticket #31170. Developers acknowledged the issue, implemented fixes that ensure the path is validated before the function is called and that the library is blocked from operating on an empty address, and released the corrected build in Telegram Desktop 7.1.2. Affected Windows users are strongly recommended to update without delay.
Related articles
Essential Active Directory Checks Before Launching Any Penetration Test
Internal penetration tests of Active Directory can quickly generate noise in logs or disrupt services if techniques are applied without proper context. The article stresses that testers must first map domain boundaries, identify controllers, clarify allowed segments, and establish escalation procedures with the client. Detailed inventory of user, service, and privileged accounts is required, including analysis of Kerberos pre-authentication settings that enable AS-REP Roasting. Testers should also examine DACL permissions, delegation relationships, and Active Directory Certificate Services templates before attempting any exploitation. The guidance covers NTLM relay risks, the importance of maintaining an action log, and the distinction between standard pentests and Red Team engagements. A demo version of the Red September course on Active Directory attacks is offered to practice these steps in a controlled environment.
Attackers Exploit Critical Sangoma Switchvox SQL Injection Flaw CVE-2026-9586 to Deploy Reverse Shells Without Credentials
A critical vulnerability in Sangoma Switchvox allows unauthenticated remote code execution through SQL injection in the /pa HTTP endpoint. The flaw, tracked as CVE-2026-9586 with a CVSS score of 9.3, affects internet-exposed VoIP deployments common in SMB environments. Attackers have already leveraged the issue in active campaigns since August 30, 2026, to open reverse shells and perform reconnaissance using Base64-encoded commands. The root cause is improper handling of the PhoneIP parameter in PostgreSQL queries, classified as CWE-89, which grants superuser database privileges. Sangoma released the fix in Switchvox 8.4.0.2 on July 14, 2026, following initial disclosure in April. Administrators are urged to patch immediately, restrict access, and monitor for indicators such as connections to 176.65.148.184 and suspicious entries in /var/log/switchvox/db-quirks.log.
Microsoft Defender Misclassifies Legitimate Google Search Links as Threats
Microsoft is investigating a false positive issue in Defender for Office 365 where legitimate Google search result links are incorrectly flagged as unsafe. Users attempting to open these links receive a warning stating the website may be insecure, and the problem persists even when copying and pasting URLs directly into browsers. The incident, tracked under advisory MO1465962, stems from erroneous classification by the Safe Links feature, which rewrites and rechecks URLs in email, Teams, and Microsoft 365 applications. The misclassification affects not only end users but also generates alerts for administrators in the Microsoft Defender Portal and Microsoft Sentinel SIEM platform. Safe Links is designed to protect against phishing by validating links at click time, but in this case it overreacted by blocking entire Google search URLs. Microsoft has not disclosed the number of impacted customers or affected regions and continues working on a fix for the erroneous classification.
NVD Disruptions Force Security Teams to Diversify Vulnerability Intelligence Sources
The vulnerability management ecosystem has undergone major shifts over the past two years, undermining reliance on a single authoritative database. NVD, long considered the central enrichment layer for CVE records, has experienced significant operational problems that affect timeliness and completeness of data. Russian organizations are increasingly turning to BDU FSTEC as the national vulnerability database that provides official descriptions, remediation guidance, and coverage of domestic software. CVE maintained by MITRE continues to serve as the universal identifier standard, yet it lacks the detailed scoring and product mappings previously supplied by NVD. Vendor security bulletins from Microsoft, Cisco, Astra Group, and Positive Technologies often deliver the earliest and most accurate information. The article explains why a multi-source approach has become essential for comprehensive vulnerability management programs.