“26,589 messages detected across 5,527 organizations,” reported email security vendor INKY — a number that frames a two-month campaign that relied on an overlooked file format and routine-looking voicemail lures to push executable code past native defenses.
Scale and timing: a broad spray, not a surgical strike
INKY, which is part of Kaseya, found the campaign running in waves from June 1 through August 4, 2026, largely pausing on weekends. The operation reached 26,589 messages across 5,527 organizations, and INKY said the activity was still running when its analysis closed. The largest single spike occurred on June 3, when 2,432 messages reached 1,149 organizations.
The distribution was blunt and wide rather than narrowly targeted: the median organization received two messages and 32% of organizations saw only one. The ten most-affected organizations accounted for just 6% of the total volume — consistent with broad “spray” delivery rather than a tightly focused spear-phish.
SVG smuggling: an added evasion layer
The campaign’s technical innovation was to weaponize scalable vector graphics (SVG) files disguised as benign voicemail attachments. Attachments used voicemail-style names, carried SVG and XML content, and — crucially — declared a MIME type of text/plain rather than image/svg+xml. As INKY phrased it, a scanner keying on the stated type would see an innocuous text file rather than active content.
SVG files can contain JavaScript; in the samples INKY analyzed a minimal graphic shell concealed obfuscated script. That script reconstructed strings at runtime, fetched a remote endpoint, and relied on deferred execution and runtime script injection to make static inspection less effective. In short, the SVG served as a bridge from social-engineering bait to executable browser content.
This site is the portfolio.
OSINTSights runs on Cloudflare Workers, D1, R2, and Vectorize, with an AI pipeline on Hetzner ARM. Nubivance designed, built, and operates it. We do the same for clients.
See what we buildAuthentication failures and native spam scoring
The attackers leaned heavily on internal spoofing: 95% of the messages claimed to come from the recipient’s own domain while arriving from an external sender that had never authenticated to the organization’s mail server. That mismatch between claimed origin and authenticated delivery was a dominant feature of the operation.
Native spam filters treated much of the traffic as harmless. Microsoft’s Spam Confidence Level (SCL) scores show the disparity: 19,994 messages (75%) received an SCL of 0 or 1 — scores Microsoft treats as not spam — while only 4,777 messages (18%) were assigned SCL 5. INKY noted that the campaign was built from a single template, yet the same message drew different native verdicts depending on the receiving mailbox.
What this means for technologists, mail administrators, and end users
- Technologists and email defenders: The campaign demonstrates how content-type mislabelling (text/plain for an SVG) and runtime-obfuscated JavaScript can defeat static scans; these defenders will be watching for attachments that carry image/XML payloads despite benign MIME labels and for script behavior that reconstructs code at runtime.
- Mail administrators and affected enterprises: The operation relied on claiming internal senders while using unauthenticated external delivery — a pattern that will draw scrutiny from administrators responsible for authentication and sender policies, particularly given the wide variation in native SCL outcomes across inboxes.
- End users: The social lure was simple and personal: 99.5% of subject lines incorporated the local part of the recipient’s own email address and presented as internal voicemail notifications. Recipients seeing single or rare voicemail-style messages with attachments should note the campaign’s use of familiar formatting and personalization.
The campaign is notable for combining low-cost personalization with a second-level technical trick: SVG smuggling that concealed runtime JavaScript and relied on MIME misclassification to mislead scanners. It did not need surgical targeting to be effective — breadth, simple personalization, and a deceptively benign attachment type together produced measurable reach. INKY’s findings leave one practical point clear: the operation was still active as the vendor closed its analysis, meaning the technique remained in play at that time.
