Front Page

Network Infrastructure

The Router That Lied to Its Own Admins: Fire Ant Rewrites Cisco IOS XR's Command Line to Hide Its Own Tunnels

Sygnia disclosed this week that the China-nexus actor it tracks as Fire Ant — overlapping with Mandiant's UNC3886 — has expanded a long-running campaign beyond VMware hypervisors to compromise Cisco IOS XR routers, TACACS authentication servers, and the Linux hosts used to manage them. Reverse engineering of the router implant showed it modifies the IOS XR command-execution path itself, appending exclusion filters to "show" commands so that tunnel source, destination, interface, and VRF details never appear to an administrator reading the router's own CLI output — the compromise is invisible from inside the device it lives on.

The campaign introduced two new tools: BridgeAgent, a Linux backdoor for persistence on management hosts, and TacTap, a TACACS credential-harvesting toolkit that injects directly into the running authentication daemon to intercept credentials from inside the process rather than off the wire. Forensic priority: never trust router CLI output alone — pull configuration and state via an out-of-band management path or vendor API and diff it against CLI output for exclusion-filter tampering, hunt TACACS daemon processes for unexpected loaded modules, and treat any Cisco IOS XR device with anomalous VRF or tunnel activity invisible to "show" commands as compromised.

Zero-Day

Two Bugs, No Password: Attackers Chain a PaperCut Auth Bypass Into Unauthenticated Code Execution

PaperCut shipped a second emergency patch this week after researchers and then real-world attackers chained two flaws in its NG/MF print management software into pre-authentication remote code execution. CVE-2026-81578 (CVSS 8.8) lets an unauthenticated request trigger backend administrative actions in the web management interface before access validation completes; from there, an attacker edits a configuration file to reach CVE-2026-82078 (CVSS 9.4), an unsafe dynamic class-loading flaw in PaperCut's database connection utilities that loads driver classes by name with no allowlist. Honeypot telemetry shows exploitation in the wild beginning around August 29, with attackers using the chain to steal data from victim servers.

The class-loading flaw is the more durable lesson: any application that instantiates a class from a configurable, attacker-influenced string is one bypassed check away from remote code execution, regardless of language or framework. Forensic priority: review PaperCut Application Server logs for unauthenticated admin-function requests followed by configuration-file writes, hunt for unexpected JDBC driver class names in server configuration, and treat every unpatched pre-v24 through v26 PaperCut instance reachable from the network as breached until log review proves otherwise.

Threat Bulletin

ACTIVE EXPLOIT
CVE-2023-49105

ownCloud — an improper-authentication flaw first disclosed in 2023 is now under active exploitation, added to CISA KEV on August 27. Old, unpatched instances remain exposed years after a fix shipped. Forensic Note: Inventory every ownCloud deployment still in production regardless of age, confirm patch level, and review authentication logs for anomalous access predating the KEV addition — this flaw may already have been exploited quietly for years.

ACTIVE EXPLOIT
CVE-2026-66384

JFrog Artifactory — a path-traversal flaw allows a remote attacker to write files outside a restricted directory, potentially planting malicious artifacts into the software build pipeline. Added to CISA KEV August 27. Forensic Note: Review Artifactory access logs for path-traversal request patterns, audit recently modified artifacts for unauthorized changes, and treat any downstream build that consumed a suspect artifact as potentially poisoned.

Malware Spotlight

Nimbus Manticore — Iran's Espionage Toolkit Gets a New Backdoor

Researchers reported August 26 that the Iranian state-sponsored group Nimbus Manticore has expanded its toolset with a TWOSTROKE-like backdoor and a dedicated SSH tunneler, giving operators a durable channel into target networks that blends with legitimate remote-administration traffic. The addition signals a shift toward longer-term network persistence rather than smash-and-grab espionage. Forensic Note: Hunt unexpected outbound SSH tunnel establishment from hosts with no administrative reason to run one, review process ancestry for the new backdoor's loader chain, and treat any unsanctioned SSH tunneling tool as a persistence indicator rather than a misconfiguration.