--- title: "African Stock Exchange AI Toolkit Attack: What to Check" url: https://playciso.com/blog/african-stock-exchange-ai-powershell-credential-toolkit updated: 2026-11-20 section: Blog publisher: PlayCISO --- # African Stock Exchange AI Toolkit Attack: What to Check > Symantec found a likely AI-written, 41-module PowerShell toolkit hunting credentials at an African stock exchange. What is confirmed and what to check. Source: https://playciso.com/blog/african-stock-exchange-ai-powershell-credential-toolkit · Published: 2026-11-20 · Publisher: PlayCISO (https://playciso.com) Primary source: https://www.security.com/threat-intelligence/african-stock-exchange-ai-assisted-attack --- On **19 November 2026**, [Symantec's Threat Hunter Team](https://www.security.com/threat-intelligence/african-stock-exchange-ai-assisted-attack) reported that attackers ran a **41-module PowerShell toolkit** under a web shell on an Internet Information Services (IIS) server at **an unnamed stock exchange in an African country**. On **20 September 2026** the modules hunted for credentials and secrets and tested a password, a reused bearer token and API keys against the exchange's own investment API and a securities depository gateway. Symantec says the toolkit has "all the hallmarks of AI authorship". It found **no evidence of privilege escalation or data theft** and cannot say whether any credentials worked. It attributes six persistence implants on the same server to the China-linked group **Witchetty (aka LookingFrog)** with high confidence, but has not attributed the toolkit to anyone. **Who this is for:** security leaders at exchanges, banks, brokers, depositories and anyone running internet-facing IIS servers that hold application configuration files, cloud keys or API tokens. *Based on Symantec's post (read in full on 20 November 2026) and earlier research on Witchetty and TA410 from Symantec, ESET and Proofpoint. PlayCISO is not affiliated with Symantec or Broadcom. File hashes, the attacker's DNS token and internal host names are not repeated here; Symantec's post has what defenders need.* **Confirmed vs not yet known (20 Nov 2026)** - Reported by Symantec: 41 PowerShell modules ran under a web shell on an IIS server between 08:44 and 08:47 UTC on 20 September 2026; they searched for secrets, mapped cloud, database, SharePoint, SCCM and TFS systems, and tested credentials against the investment API and a securities depository gateway; output left as DNS lookups to oast.pro; six persistence implants were on the server; the earliest malicious activity found dates from April 2026. - Attribution, as Symantec words it: the implants are Witchetty's "with high confidence"; the toolkit is not attributed. Symantec says it "could be Witchetty using new tooling, a second and more capable operator using the same access, or an unrelated actor who has compromised the same server". - AI authorship: Symantec's assessment, not a certainty. It says "none of these signals is conclusive on its own" and it "could not identify which model may have been used". - Not found: "no evidence of privilege escalation or data theft". Some headlines say credentials were stolen; Symantec does not. - Not known: the exchange's name and country, how the attackers first got in (Symantec says a vulnerability "may have been exploited"), whether any credential or key worked, what the collector received, and what the attackers wanted. ## What the toolkit did Per Symantec, the 41 modules ran in "three minutes and twelve seconds, with a new module starting every few seconds", each a single-purpose script. Between them they: - Hunted for secrets in files. They searched web roots, a deployment directory, the domain's SYSVOL share and the application directories of six business applications for configuration files holding hardcoded passwords, Group Policy Preferences passwords, AWS access keys and database connection strings, and checked the host for cloud credentials. - Mapped the estate. They profiled IIS sites and application pools, opened a connection to a database cluster on Amazon Web Services, queried SQL Server, enumerated SharePoint sites for files mentioning words like "aws", "credential" and "password", listed Team Foundation Server projects, and asked System Center Configuration Manager (SCCM) which machine a given user last logged on to. - Tested what they found. Symantec says four modules targeted the exchange's investment API: "They sent a login request using a username and a guessed password, reused a bearer token, and tested several API keys against the depository gateway, including a blank key that may have served as a control." Those modules connected by IP address, set the Host header by hand and disabled certificate validation. Others queried endpoints for investor and market data, and Symantec notes the attackers "did also check a specific investor's number". ## How the output left: DNS to a testing service Symantec says each module "hex-encoded its output, split it into 50 hex character chunks and resolved a DNS name for each chunk". The names were unique subdomains of a single token under **oast.pro**, which Symantec calls "a public out-of-band testing service". oast.pro is one of the public default servers of [interactsh](https://github.com/projectdiscovery/interactsh), an open-source tool pen testers use to catch DNS and HTTP call-backs. Whoever holds the token sees every lookup, so the data never had to touch a web proxy. Symantec calls this "a rarely seen technique" and suggests it was an attempt to stay under the radar. For defenders, the pattern is the useful part: one server suddenly resolving many long, hex-looking, never-repeated subdomains under one parent domain, in a burst, with a short label per module (Symantec gives examples such as one for host enumeration and one for cloud hunting). That is visible in resolver logs even when the payload is not. ## Why Symantec thinks AI wrote it Symantec's case "rests on how uniform the modules are". Each opened with the same error-handling line, used the same outer try/catch, encoded output the same way and returned the same empty-output marker. The DNS exfiltration loop "was copied byte for byte into all 41 modules", where "a programmer building a toolkit would normally write that function once". Variable names were generic, one module had a progress variable no other module had "as if a model had improved a single script in isolation", and the pace matched automation. Symantec's own conclusion is careful: the signals "fit AI-assisted or AI-generated authorship better than they fit bespoke attacker tooling", but none is conclusive. It also points out the upside for defenders: "The uniformity that points to AI authorship also gives defenders something to look for." Whoever or whatever wrote the scripts, they still needed a web shell, readable secrets in files and open DNS egress. Those are the controls that matter. ## Witchetty, LookingFrog and TA410 Symantec calls Witchetty "the China-linked espionage group Witchetty (aka LookingFrog)". [ESET](https://www.welivesecurity.com/2022/04/27/lookback-ta410-umbrella-cyberespionage-ttps-activity/) described LookingFrog in April 2022 as one of three teams under TA410, which ESET calls "a cyberespionage umbrella group loosely linked to APT10"; ESET says LookingFrog uses the X4 backdoor and LookBack. [Proofpoint](https://www.proofpoint.com/us/blog/threat-insight/ta410-group-behind-lookback-attacks-against-us-utilities-sector-returns-new) first reported TA410 in 2019 for LookBack phishing against the US utilities sector, and in 2020 tied the FlowCloud malware to the same actor. In [September 2022](https://www.security.com/threat-intelligence/witchetty-steganography-espionage), Symantec reported Witchetty attacks between February and September 2022 on two Middle Eastern governments and, it now says, "this same African stock exchange", using ProxyShell and ProxyLogon flaws in Microsoft Exchange to plant web shells and a backdoor hidden in an image. This time Symantec says one Witchetty implant on the server "was first seen in 2020", but the toolkit was "a far more advanced toolkit than those previously attributed to Witchetty" and focused on cloud secrets and financial APIs, "so we treat the implants and the toolkit separately". It weighs a financial motive (the investment API and banking gateway tests) against espionage (the specific investor lookup) and does not settle it. To line up the group's other names, use [APT Rosetta](/tools/apt-rosetta). ## Timeline | Date | Event | Source | | --- | --- | --- | | Aug 2019 | Proofpoint reports LookBack attacks on the US utilities sector, later tracked as TA410 | Proofpoint | | 2020 | One of the Witchetty implants later found on the exchange's server is first seen | Symantec | | 27 Apr 2022 | ESET names LookingFrog as one of three TA410 sub-groups | ESET | | Feb-Sep 2022 | Witchetty targets two Middle Eastern governments and the same African stock exchange | Symantec (2022) | | Apr 2026 | Earliest malicious activity Symantec found on the server; six persistence implants present | Symantec | | 20 Sep 2026, 08:44-08:47 UTC | 41 PowerShell modules run under the web shell and test credentials against the investment API and depository gateway | Symantec | | 19 Nov 2026 | Symantec publishes "Major African Stock Exchange Targeted in AI-Assisted Attack" | Symantec | | 20 Nov 2026 | GBHackers reports the findings | GBHackers | ## What security leaders should do this week - Hunt for web shells on IIS. Look for files in web roots that are new or changed outside your release process, and above all for w3wp.exe (the IIS worker process) starting powershell.exe, cmd.exe or script hosts. On a normal web server that parent-child pair should almost never happen; alert on it. - Patch and inventory the internet-facing servers. Symantec thinks a vulnerability was likely exploited, and the 2022 intrusion came through unpatched Exchange. List every public IIS and Exchange server, its owner and its patch level. - Sweep files for credentials. Scan web roots, deployment folders, web.config, .json, .xml and .env files, file shares and SYSVOL for passwords, cloud keys, tokens and connection strings, and remove any Group Policy Preferences cpassword entries. Move what you find into a secrets manager. The [secret scanner](/tools/secret-scanner) shows what the patterns look like. - Rotate what a web server can read. If a public server could read an API key, bearer token or database password, assume it may be known. Rotate API keys and bearer tokens for financial and partner interfaces, shorten token lifetimes, and bind API keys to expected source hosts. Work through the [leaked credential response](/tools/leaked-credential-response) steps. - Watch DNS egress. Make servers use only internal resolvers and block direct outbound DNS. Alert on hosts that resolve many unique, long or encoded-looking subdomains under one domain, and on any lookups of out-of-band testing domains such as interactsh's oast.* servers from production systems. - Block OAST services at the resolver where you do not need them. Your pen testers may use them; production servers should not. Allow them only for named test hosts and windows. - Protect the money-adjacent APIs. Alert on failed logins, reused tokens and bursts of API-key attempts against investment, settlement and depository interfaces, and on calls that arrive by IP address with a hand-set Host header from an internal web server. - Do not change the detections because of AI. AI-assisted scripting made the toolkit faster to build, not harder to see. Web shell, credential-in-files and DNS egress detections all still apply; the copied scaffold even gives you a signature to hunt. ## Put it to work - [DNS tunneling detection lab](/tools/dns-tunneling-detection-lab): find encoded subdomains in DNS logs. - [Secret scanner](/tools/secret-scanner): see the key, token and connection-string patterns to sweep for. - [Leaked credential response](/tools/leaked-credential-response): rotate and contain exposed keys in the right order. - [APT Rosetta](/tools/apt-rosetta): map Witchetty, LookingFrog, TA410 and related names. For other recent attacks on financial organisations, see the [Binance hack explainer](/blog/binance-hack-aave-loan-usdt-freeze), which Symantec cites as part of the same trend, and [bank cyber incidents in November 2026](/blog/bank-cyber-incidents-november-2026-whats-confirmed). ## FAQ ### Which African stock exchange was attacked? Symantec does not name it. Its report refers only to "a stock exchange in an African country" and says it is the same exchange Witchetty targeted in 2022, which Symantec also did not name at the time. Any name attached to this story elsewhere is not from Symantec. ### Was data stolen from the stock exchange? Symantec says: "We found no evidence of privilege escalation or data theft." The toolkit did send its output out of the network as DNS lookups, but the results went to a collector Symantec cannot see, so it does not know what was received or whether any credentials or keys worked. Headlines saying credentials were "stolen" go further than Symantec does. ### Was the toolkit written by AI? Symantec says it has "all the hallmarks of AI authorship" and calls it "likely AI-authored", but adds that "none of these signals is conclusive on its own". The signs it cites: every module shared the same scaffold, error handling and output format, a DNS exfiltration loop was copied byte for byte into all 41 scripts instead of being written once, variable names were generic, and modules started every few seconds. Symantec could not identify which model may have been used. ### Who is Witchetty? Symantec calls Witchetty (aka LookingFrog) a China-linked espionage group. ESET first described LookingFrog in April 2022 as one of three teams under the TA410 umbrella, which ESET calls "loosely linked to APT10". Proofpoint first reported TA410 in 2019 for LookBack attacks on the US utilities sector. Symantec attributes the six persistence implants on the exchange's server to Witchetty with high confidence, but says it has not attributed the AI-authored toolkit to any group. ### What is oast.pro and why was it used? oast.pro is one of the public default servers of interactsh, an open-source out-of-band application security testing (OAST) tool that pen testers use to see whether a target makes DNS or HTTP calls back to them. Symantec says each module hex-encoded its output into 50-character chunks and looked up a DNS name for each chunk under a single token on oast.pro, so the data left as DNS queries rather than web requests. Symantec calls this "a rarely seen technique". ### How did the attackers get into the stock exchange? Symantec does not say for certain. It says the toolkit ran from a web shell under the IIS worker process, "which suggests that a vulnerability may have been exploited to gain access". The earliest malicious activity it found dates from April 2026, and it says the attackers may have been present for longer. In 2022 Witchetty got into the same exchange through ProxyShell and ProxyLogon flaws in Microsoft Exchange. ### How do you detect DNS exfiltration like this? Look for hosts, especially web servers, that resolve many unique, long, hex- or base32-looking subdomains under one parent domain in a short time, and for any lookups of OAST services such as interactsh domains from production servers. Servers should normally use only your internal resolvers, so direct DNS to the internet from a web server is itself worth an alert. Our DNS tunneling detection lab walks through the signals. ## Sources - [Symantec Threat Hunter Team (Broadcom): Major African Stock Exchange Targeted in AI-Assisted Attack](https://www.security.com/threat-intelligence/african-stock-exchange-ai-assisted-attack) (19 Nov 2026; Thibaut Passilly, Brigid O'Gorman). Primary source, including the indicators of compromise. - [Symantec: Witchetty: Group Uses Updated Toolset in Attacks on Governments in Middle East](https://www.security.com/threat-intelligence/witchetty-steganography-espionage) (September 2022) - [ESET: A lookback under the TA410 umbrella: Its cyberespionage TTPs and activity](https://www.welivesecurity.com/2022/04/27/lookback-ta410-umbrella-cyberespionage-ttps-activity/) (27 Apr 2022) - [Proofpoint: TA410, the group behind LookBack attacks against US utilities sector, returns with new malware](https://www.proofpoint.com/us/blog/threat-insight/ta410-group-behind-lookback-attacks-against-us-utilities-sector-returns-new) (2020) - [ProjectDiscovery: interactsh](https://github.com/projectdiscovery/interactsh) (lists oast.pro among its public servers) - [GBHackers: AI-Assisted Hackers Deploy 41 PowerShell Scripts to Steal Credentials From African Stock Exchange](https://gbhackers.com/african-stock-exchange-credential-theft/) (20 Nov 2026) *This article is educational guidance for security leaders, not a threat-intelligence report. Indicators are in Symantec's post; check it for updates before acting.*