DISCARDED: Tales From the Threat Research Trenches

Send us fan mail!
Hello to all our cyber party pals!  
This week, Selena is joined by co-host Sarah Sabatka (and several very persistent, very uninvited pod crashers) to talk about a threat that's landed in a lot of listeners' personal inboxes lately: malicious event invitations.
Selena and Sarah walk through what these lures actually look like — polished, Evite- or Paperless Post-style invitations for birthdays, weddings, baby showers, housewarmings, and even generic "Celebrate with Me" surprises, sent from real, compromised accounts belonging to people the recipient actually knows. Sarah explains why the generic, low-detail invites may be the most effective: they don't require any specific knowledge of the target to feel plausible, and they lean on curiosity and FOMO rather than fear, a notable departure from traditional phishing psychology.
The conversation dives into ATO jumping, the mechanism that makes this threat spread so efficiently — once a threat actor compromises one inbox, they mine the contact list and send credential phishing to those contacts, who then get compromised themselves, expanding the reach with every hop. Sarah also connects this lure to established delivery chains like RMM tooling (ScreenConnect, Atera) and credential phishing panels such as TA4903 and "Evil Tokens," which use AI-assisted, sometimes vibe-coded automation to filter, organize, and scale compromised mailboxes rather than to improve the social engineering itself.
Selena and Sarah also dig into:
Why CAPTCHA and Cloudflare challenges are increasingly bundled into these lures — both to build a false sense of legitimacy and to slow down defenders and researchers trying to analyze the infection chain
Why Microsoft 365 and Google accounts are such high-value targets: they're the hub for cloud documents, financial access, social platforms, and a trusted contact list an attacker can impersonate
Practical recovery steps if you've already entered your credentials: revoke active sessions, change your password, enable MFA (passkeys where possible), freeze accounts if financial exposure is a concern, and give your contacts a heads-up
Why smart, security-aware people still fall for these — FOMO and curiosity are natural human responses, not a knowledge gap, and no lure is universal, but there's a lure for everyone
The FTC's May consumer alert on fake party invitations, and simple habits (a quick text or call to confirm an invite) that can blunt the whole scam
Plus: several unannounced podcast "guests" (Tim, Isaac, and Jared) who show up mid-episode having fallen for the very scam being discussed, a genuinely useful tangent on TA2725's shifting tool preferences, and a closing thought on why these lures work in the first place — people miss connecting with each other.
Resources Mentioned:
Asked to Enter Your Email Address and Password to Open a Party Invite? That's a Scam — Federal Trade Commission

https://www.proofpoint.com/us/blog/threat-insight/remote-monitoring-and-management-rmm-tooling-increasingly-attackers-first-choice

For more information about Proofpoint, check out our website.

Subscribe & Follow:
Stay ahead of emerging threats, and subscribe! Happy hunting!

Send us fan mail!
Hello to all our cybernauts!
This week Selena is joined by not one, not two, but three guests — Mark Kelly, Julia Paluch, and Dave Galazin — to unpack Proofpoint's latest research. Mark walks through how it started: on August 28, the China-aligned actor TA412 (aka Violet Typhoon), previously known mostly for device registration phishing, suddenly began delivering a Chrome exploit — kicking off a frantic cross-team investigation involving partners at Google and Microsoft. Within days, three more state-sponsored clusters adopted the same exploit chain, dubbed BlueMoon; by the time the blog published, that number had grown to seven distinct clusters, plus an eCrime actor and a brand-new espionage group picking it up since.
Julia breaks down the Chrome half of the chain — a type confusion bug in the V8 engine chained with a sandbox escape — and introduces the concept of the "patch gap": the roughly four-week window between when a fix is committed to the public Chromium codebase and when it's fully rolled out to stable Chrome releases. During that window, the regression test accompanying the fix effectively doubles as a roadmap for building an exploit, especially with an LLM doing the heavy lifting of interpreting it.
Dave then covers the Windows half: a kernel heap bug that lets an attacker escalate from Chrome's heavily sandboxed, untrusted-integrity renderer process straight to full privileges — without needing any file system access first. He and Mark also make the case that this particular Windows exploit had likely been sitting unused for over a year, based on consistent compilation timestamps across the payload's debug information.
Selena, Mark, Julia, and Dave also dig into:
Why the deployment looked rushed: exploit kits full of debug logging and comments, a downloader left behind indefinitely, and a Windows exploit that doesn't even check what OS the victim is running
Historical precedent for this kind of rapid multi-actor tool sharing among Chinese threat actors, including 2021's ProxyLogon and 2025's ToolShell activity
Evidence pointing toward AI-assisted exploit development — including leftover debugging comments and variable names lifted straight from the public regression test
Whether the open-source community's transparency (public commits, public regression tests) needs to change in light of this kind of patch-gap weaponization
Why Windows zero-days are much rarer to find "on the shelf" than browser bugs, and what that implies about how this one got used
Practical defensive takeaways: post-exploitation detection still matters most, since noisy, AI-assisted tradecraft is easier to catch than the initial exploit itself
Plus: a very enthusiastic Chrome-tab-hoarding discussion, Julia's advice to enable Chrome crash reporting, and a well-earned round of "you need a vacation, Mark."
Resources Mentioned:
Once in a BlueMoon: Multiple State-Aligned Threat Actors Rapidly Adopt Novel Exploit Chain Using Chrome and Windows Zero-Days


https://msrc.microsoft.com/update-guide/vulnerability/CVE-2026-85880

For more information about Proofpoint, check out our website.

Send us fan mail!
Hello to our Cyber Pals! 
This week, Selena hands the mic over to a recording of Proofpoint's Intercepted webinar — a monthly video series she co-hosts with Sarah Sabotka — for a deep dive into how cloud brute force attacks are quietly getting more sophisticated. Joining Sarah and webinar host Jason is Rachel Rabin, a threat researcher on Proofpoint's Cloud Threat Research Team, to unpack two spoofing techniques she's seen increasingly weaponized against Microsoft Entra ID environments.
Rachel opens with the state of brute forcing today: despite flashier techniques like device code phishing and MFA-resistant phishing kits, brute force remains the most common identity attack vector, with 99% of tenants experiencing some form of it in 2026 — even though success rates typically sit well under 1%. She walks through why first-party Microsoft apps like Azure CLI and Azure AD PowerShell remain the most heavily targeted (publicly known client IDs, gaps in MFA enforcement, and pre-consented scopes), and why the legacy ROPC authentication flow makes credential testing so convenient for attackers.
From there, Rachel breaks down two techniques that "level up" traditional brute forcing:
User agent spoofing, illustrated by a campaign (later tracked as UNK_CustomCloak) that hit Windows Live Custom Domains — a legacy, discontinued Microsoft app — impacting half a million users across 4,000 tenants using over 600,000 IP addresses and a staggering 7.4 million unique, machine-generated user agent strings.
OAuth client ID spoofing, a newer and stealthier technique where attackers submit a syntactically valid but fake client ID to infer whether a password is correct from Entra's error codes, without ever generating a successful sign-in event that would normally tip off defenders. Rachel compares two real-world campaigns (UNK_pyreq2323 and UNK_OutFlareAZ) that appear to have arrived at the technique independently.
Sarah, Jason, and the live chat also dig into:
Why over-engineering a spoofed user agent can make an attacker easier to spot, not harder
How conditional access policies scoped to the wrong signal (like operating system) can be silently bypassed by a spoofed user agent
Why a failed sign-in shouldn't be dismissed as a non-event. It can still mean a credential has been validated
Rachel's practical, "lowest hanging fruit" recommendations: MFA, blocking legacy authentication, layered conditional access, smart lockout policies, and the shift toward passkeys
Plus: a shoutout to next month's session on fraud and job scams ahead of Cybersecurity Awareness Month.
Resources Mentioned:
OAuth Client ID Spoofing: Why Fake Client IDs Are Gaining Traction for Stealthy Enumeration — Proofpoint

Send us fan mail!
Hello to all our cyber stars!
Host Selena Larson is joined by co-host Sarah Sabotka, along with returning guest Greg Lesnewich, Principal Threat Research Engineer, and Staff Threat Researcher Mark Kelly, for a follow-up to last episode's half-click deep dive — this time tracking how the technique is spreading beyond TA458 and TA488 into a brand-new Chinese cluster.
Greg picks up where last week left off with TA488 (aka Void Blizzard, aka Laundry Bear): after going quiet in February following exposure, the group resurfaced on July 22 — one day before Proofpoint's joint advisory with the NSA — with a new half-click exploit against Microsoft Outlook Web Access. The payload, a previously unseen browser-resident implant called OWAReaper, is built for persistence: it survives credential rotation, browser restarts, and even a full device reimage, and it phones home via GitHub commit messages and image requests proxied through legitimate CDNs like Slack and WordPress.
Mark then brings a new actor into the half-click conversation: UNK_MassTraction, a suspected China-aligned cluster exploiting a cross-site scripting flaw in Roundcube to target physics and engineering departments at North American universities — chosen, it turns out, specifically because they were running vulnerable Roundcube instances. The infection chain drops either a webshell or VShell, a Linux backdoor of murky, possibly commercially-developed origin that keeps popping up across Chinese state-sponsored intrusions.
The conversation covers:
How OWAReaper's persistence differs from anything Greg's seen before — including "poisoning" a user's other emails to reinfect the browser on every new tab
Why Greg's assessment of TA488 shifted from "less capable than TA458" to "neck and neck" after seeing this campaign
How Mark and Greg discovered UNK_MassTraction independently, within hours of each other
Why UNK_MassTraction's targets were all running vulnerable Roundcube — evidence of deliberate reconnaissance, not spray-and-pray
IceCube, the Roundcube stealer with telltale signs of LLM-assisted development
Why mail servers are functionally edge devices, and what that means for defenders who can't run EDR on them
Whether AI is accelerating adversary tradecraft on both the Russian and Chinese sides
The ethics of publishing offensive research on a live, evolving technique
The funniest (and most unsettling) lure email of the year: "Hi fellow, I really liked your vibe..."
Plus: Gus the dog's podcast debut, security awareness training in a world where opening an email can be enough, and shoutouts to Tom Lancaster (Volexity) and Microsoft's patch team.
Resources Mentioned:
Cleaning Out Inboxes: TA488 Comes for Outlook with Another Half-Click Exploit
TA488 Targets Zimbra Mailservers with Half-Click Exploits
One Email Closer to the Edge: UNK_MassTraction & the Physics of Exploitation
For more information about Proofpoint, check out our website.
Subscribe & Follow:
Stay ahead of emerging threats, and subscribe! Happy hunting!

Send us fan mail!
Hello to all our cyber roosters!
Host Selena Larson is joined by co-host Sarah Sabatka and returning guest Greg Lesnewich, Principal Threat Researcher at Proofpoint, to unpack two new reports on Russian aligned espionage actors abusing “half-click” exploits, which are vulnerabilities that compromise a target the moment they open an email in a vulnerable webmail viewer, no link or attachment required.
Greg breaks down the two distinct actors covered in Proofpoint's research: TA458, a mature, well-resourced group active across multiple webmail platforms (Zimbra, MDaemon, Roundcube, SOGo, and Horde) primarily targeting Ukraine, Eastern Europe, and militaries/ministries of foreign affairs; and TA488 (aka Void Blizzard, aka Laundry Bear), a comparatively scrappier contractor operation that leaned on a single Zimbra zero-day dubbed “ZimReaper” to steal credentials, full mailboxes, and up to 90 days of email history from Ukrainian government and U.S. targets.
The conversation covers how the half-click technique works under the hood (stored cross-site scripting via mishandled JavaScript event handlers in webmail HTML), how it differs from traditional phishing, the evidence pointing to a possible link between TA458 and GRU Unit 20728, why TA488's contractor status may explain its skittishness (burning down infrastructure after being outed) versus TA458's higher risk tolerance (including repeatedly firing exploits at targets that weren't even running vulnerable software), and what defenders running Zimbra, Roundcube, or similar platforms can actually do about it.
Resources Mentioned:
https://www.proofpoint.com/us/blog/threat-insight/ta458-roundpress-exploits
https://www.proofpoint.com/us/blog/threat-insight/ta488-targets-zimbra-mailservers-half-click-exploits
For more information about Proofpoint, check out our website.
Subscribe & Follow:
Stay ahead of emerging threats, and subscribe! Happy hunting!