Skip to content

Image: security.com · rights & removal

Executive Summary

An AI-assisted toolkit was used to scan an African stock exchange, testing credentials against its investment API and a securities depository gateway. The activity involved 41 PowerShell modules running on an IIS server for just over three minutes, which systematically hunted for secrets across shared folders and application directories, mapping cloud infrastructure, and probing financial APIs. The tools employed a unique DNS exfiltration method where output was hex-encoded and sent as DNS lookups to a public testing service. Six persistence implants were found on the server, attributed with high confidence to the China-linked espionage group Witchetty. The activity suggests a potential financial motivation, though evidence for espionage regarding specific investor data cannot be confirmed.

Facts Only

* A 41-module PowerShell toolkit was tested against an exchange's investment API and securities depository gateway.
* The modules ran under a web shell on an Internet Information Services (IIS) server for three minutes and twelve seconds.
* Modules hunted for credentials and secrets in shared folders, application directories, and infrastructure listings.
* Modules tested credentials, bearer tokens, and API keys against the exchange's investment API and depository gateway.
* The toolkit included modules for Host/Process Enumeration, Cloud Hunting (AWS), Web Server Profiling (IIS), Secret Scraping (GPP, AWS keys), Database Probing (AWS RDS, SQL Server), SharePoint enumeration, and API Exploitation.
* Output from the modules was sent as hex-encoded DNS lookups to a public out-of-band testing service.
* Six persistence implants on the server were attributed with high confidence to the China-linked espionage group Witchetty.
* The earliest evidence of malicious activity dated from April 2026, though attackers may have been present longer.
* The toolkit utilized a copied and byte-for-byte replicated Domain Name System (DNS) exfiltration loop across all modules.
* Persistence implants were attributed to Witchetty; the AI-authored toolkit attribution is unconfirmed.

Full Take

The pattern observed points toward a convergence of advanced, modular attack methodologies and generative AI capabilities targeting high-value financial infrastructure. The uniformity in the toolkit's structure—specifically the copied DNS exfiltration loop and standardized error handling across all 41 modules—strongly suggests an AI system was leveraged for tool generation rather than being the sole author of bespoke code. This indicates a shift where threat actors can rapidly prototype highly specific, multi-faceted reconnaissance tools designed to interface with complex systems, like financial APIs, in a manner that minimizes manual scripting effort.
The attribution uncertainty surrounding Witchetty versus an unknown actor highlights a critical difficulty in attributing cybercrime when state-sponsored espionage groups engage in financially motivated activities. The fact that the tooling is more advanced than previous attributed activity suggests either adaptation by established groups or the introduction of novel tooling by newer actors. The use of DNS exfiltration as an anomalous, low-noise channel for data exfiltration further exemplifies a sophisticated evasion technique intended to bypass traditional network monitoring focused on web requests.
The implications center on operationalizing machine learning in offensive security: if attackers can generate functionally complete, obfuscated tools rapidly, the barrier to entry for complex financial targeting is lowered significantly. The focus shifting from simple vulnerability exploitation (like the prior Exchange attacks) to automated credential harvesting and API testing suggests a maturation of threat actor objectives toward direct financial manipulation. What are the long-term implications when this capability becomes standard for financially motivated groups?
Bridge Questions:
How does the use of AI-assisted tooling change the defensive posture required for monitoring and detecting anomalous behavior in highly complex environments like exchange APIs?
If persistence implants remain attributed to espionage groups, how should defenders prioritize threat intelligence related to financial crime versus state-sponsored activity when operating within jurisdictions with known geopolitical sensitivities?
What systems are necessary to detect the *process* of AI-assisted tool generation rather than just the resulting artifact?

From the original · Symantec Threat Intelligence

Assisted Attack A 41-module PowerShell toolkit with all the hallmarks of AI authorship tested credentials and API keys against the exchange's investment API and a securities depository gateway.
Read the full story at security.com

Sentinel unavailable

The automated check of the source article's wording did not complete, so no result is shown.

This looks only at the wording of the original source article, not at this page's AI-written sections. A small local AI model made this estimate. It has not been checked against known human and machine texts, so treat it as provisional. It cannot show who wrote an article.

Major African Stock Exchange Targeted in AI | Huntaegis