African Stock Exchange Faces Credential Theft, Possibly by AI-Driven Attackers
In a striking report by Symantec researchers, it has come to light that cyber attackers likely employed an AI-generated toolkit to infiltrate an African stock exchange. According to the findings, the attackers executed a rapid series of operations, running an impressive 41 PowerShell modules in just 3 minutes and 12 seconds, from 08:44 to 08:47 UTC. This operation was conducted under a webshell hosted on an Internet Information Services (IIS) server, showcasing both the sophistication and speed of the assault.
The modules employed in this attack were designed to test various credentials, including a reused bearer token and multiple API keys, against the exchange’s investment API and a securities depository gateway. However, due to the limitations in their monitoring capabilities, Symantec was unable to ascertain whether any of the credentials successfully gained access to sensitive systems.
Targeted Operations and Types of Data Sought
The modules were particularly geared towards extracting Group Policy Preferences passwords, while also scouring critical directories such as SYSVOL and the web roots of the server. They were searching for crucial information like AWS keys and database connection strings, which are often vital for establishing connections to cloud services and databases. Additional modules concentrated on identifying cloud credentials and established a connection to an AWS-hosted database cluster, indicating a serious prioritization of cloud-related resources.
Moreover, the attackers seemed to have a clear strategy; several modules were specifically aimed at the investment API. They attempted to gain entry by sending a login request with a guessed password, reused the aforementioned bearer token, and tested various API keys, which intriguingly included a blank key suspecting it could be a control mechanism. The attackers utilized IP address-based connections, manipulated the Host header, and even disabled certificate validation, showing a blatant disregard for common security protocols.
Technical Insights into the Attackers’ Methodology
A detailed analysis revealed that every module launched with identical error-handling protocols and utilized consistent try/catch techniques. This level of uniformity in the programming suggests a potential automated origin. Notably, a DNS exfiltration loop, which typically would be written once as a common function by any skilled programmer, was instead duplicated byte-for-byte across all 41 scripts used in the attack.
Interestingly, while one module included a unique progress variable, another was designed to submit invalid credentials, indicating a deliberate testing strategy. The rapid deployment of new modules every few seconds adds to the impression that these operations were orchestrated with a high level of automation.
Symantec’s investigation did not yield a single conclusive signal indicating the nature of the authorship but pointed towards the likelihood of AI-assisted creation rather than traditional bespoke tooling. Although Symantec was unable to identify a specific AI model, the collective evidence leans towards a modern, sophisticated approach to cyber threats.
Historical Context and Attribution
The malicious activity on the server reportedly commenced as far back as April 2026, raising suspicions that the attackers might have had a prolonged presence before executing this recent operation. Researchers identified the installation of six persistence implants, leading to a high-confidence attribution to a cyber-espionage group known as Witchetty, also recognized by its alias, LookingFrog. This group is linked to activities associated with China.
Witchetty was previously documented by ESET in April 2022, primarily under the umbrella of TA410. Symantec had previously reported that this group targeted governmental organizations in the Middle East and an African stock exchange using recognized vulnerabilities to establish webshells.
Despite the advanced nature of the toolkit used in this latest assault, Symantec refrained from directly attributing it to Witchetty, suggesting it could either be the same group employing new tools or another equally skilled actor exploiting the same access points. The toolkit itself appears to be significantly more advanced than prior tools associated with Witchetty.
Uncertainty in Motive and Future Implications
Determining the attackers’ motivations remains elusive. The probing of the investment API and the banking gateway may indicate financial motivations, while additional interests—such as scrutinizing specific investor numbers and seeking source code or employee data—hints at potential espionage activities.
Symantec’s findings did not reveal any evidence of privilege escalation or data theft during this operation. Their Endpoint products are capable of blocking known malicious files effectively, and a Symantec Protection Bulletin outlines current protective measures against these types of threats.
As organizations continue to combat increasingly sophisticated cyber threats, the findings serve as a critical reminder of the evolving landscape of cybercrime, particularly the potential involvement of AI in crafting new and advanced attack strategies. This incident underscores the pressing need for enhanced security protocols, rigorous monitoring, and ongoing assessments of potential vulnerabilities to protect sensitive financial and investment infrastructures.
