
StrikeShark is a newly discovered intrusion campaign that primarily uses the custom SharkLoader malware loader to deploy Cobalt Strike Beacon on victim systems. What initially appeared to be an isolated incident at an Indonesian diplomatic organization was later found by researchers to be part of a much broader campaign spanning multiple countries and industries, indicating a larger and ongoing attack operation.
This campaign combines widespread opportunistic threats with potentially strategic targeted attacks. Researchers observed that attackers have been exploiting vulnerabilities in a range of Internet-facing enterprise applications and network devices, including Microsoft Exchange, SharePoint, Openfire, GeoServer, Fortinet, Cisco IOS XE, F5 BIG-IP, Zimbra, Apache Shiro, and Hikvision products. The attackers likely rely on publicly available proof-of-concept (PoC) exploits, allowing them to rapidly scale their attacks and sweep the Internet for exposed, unpatched systems. Internet-wide scanning activity linked to the attack infrastructure further confirms their intent to hunt for vulnerable organizations at scale.
At the same time, victimology shows the attackers have a strong, repeated interest in government, diplomatic, and software development organizations in Indonesia and Taiwan, with additional victims spread across Hong Kong, Lebanon, Syria, Colombia, North Macedonia, Nepal, and Serbia. This pattern has raised concerns that the campaign may support espionage or intelligence-gathering activity, although its true intent cannot yet be fully confirmed.
Affected organizations may face significant impact. Once the attackers gain access, they establish persistence and deploy SharkLoader to launch Cobalt Strike Beacon, giving them a flexible foothold for reconnaissance, credential theft, lateral movement, and long-term access. Post-compromise activity includes enumerating Active Directory information, using WebShells and post-exploitation tools, and attempting to dump credentials from LSASS and the NTDS database. These actions allow the attackers to escalate privileges, move deeper into the environment, and compromise additional systems and accounts.
For affected industries, the risk is never limited to a single compromised host. Government and diplomatic entities may face risk of exposing sensitive internal communications, network architecture, and privileged credentials; software development firms may face threats to intellectual property, source code, internal documents, and trusted development infrastructure. Because this campaign also attacks opportunistically, any organization that exposes unpatched services to the public Internet is at risk, regardless of industry.
The true scale of this campaign is likely far larger than what is currently known. The research team's visibility is limited to observed telemetry, and the attackers' frequent exploitation of public-facing application vulnerabilities suggests there may be many more undiscovered victims. Even where data exfiltration has not yet been confirmed, if the attackers retain Cobalt Strike access they keep the ability to launch further attacks, including data theft, remote command execution, and deeper network compromise.
Overall, StrikeShark is a high-risk campaign because it combines large-scale exploitation of known vulnerabilities with post-exploitation tactics suited for long-term persistence. The threat is especially significant for organizations that operate Internet-facing enterprise services, and for industries where the loss of data, credentials, orstrategic intelligence could trigger major operational or geopolitical consequences.
StrikeShark's intrusion chain combines opportunistic attacks with a technically mature, multi-stage malware loader. Atthe core of the campaign is the custom SharkLoader loader, which deploys CobaltStrike Beacon directly in memory, minimizing physical files left on disk and significantly increasing the difficulty of detection. In observed cases, attackers exploit Internet-facing application vulnerabilities and deliver custom droppers disguised as legitimate software to gain access. They then use WebShells, DLL Side-Loading, encrypted payload stages, reflective loading, API hooking, and stealth-focused memory operations to establish execution and maintain control.
Reconnaissance
Both before and after compromisinga system, the attackers gather information to identify viable targets and understand the victim environment.
Before compromise, the most notable behavior is Internet-wide scanning launched from IPs linked to the attackers' infrastructure. This strongly suggests the attackers are broadly searching for vulnerable enterprise software and devices, which is entirely consistent with the campaign's wide victim distribution and its repeated exploitation of public-facing application vulnerabilities.
After gaining access,the attackers use the following native commands to perform host and network reconnaissance:

In more thorough intrusions (particularly in Active Directory environments), they expand their investigation using the following commands:

The attackers also use PowerShell and ADtools (including Get-ADGroupMember, Get-ADUser, dsquery, and dsget) to enumerate privileged groups and users. This shows the attackers are not just collecting local system details, but also mapping trust relationships, privileged accounts, and a potential map of systems for later use.
Resource Development
This campaign appears to rely heavily on external, off-the-shelf offensive resources rather than developing its own exploits. The research team assesses with moderate confidence that the attackers use public PoC exploits hosted on platforms such as GitHub to attack known vulnerabilities in enterprise systems.
The attackers also prepare related operational resources, including:
In addition, the attackers incorporate several open-source tools into their workflow, including FScan, Searchall, Pillager, SharpGPOAbuse, Microsoft Detours, and MinHook — indicating they assemble and adapt a practical tool chain from public resources suited for stealthy post-exploitation.
Initial Access
This campaign primarily uses two initial access vectors.
The first and primary vector isexploitation of Internet-facing application vulnerabilities. Observed vulnerable products include:
This set of vulnerabilities shows that the root cause of many intrusions is not a Zero-day or rare weakness, but organizations exposing unpatched systems directly to the Internet.
The second vector is delivery via droppers. Custom droppers disguise themselves as legitimate installers or documents, with filenames such as:
Some samples also use document-themed lures, such as naming executables to look like PDFs. In one Taiwan-related sample, the filename was in Chinese and used a .pdf.exe double extension, combining social engineering with executable delivery.
Execution
How execution begins varies by access vector, but all paths ultimately converge on the same SharkLoader execution chain.
In exploitation cases, the attackers upload components via a WebShell, then launch a legitimate Windows binary to trigger DLL Side-Loading. A common approach is copying and executing SystemSettings.exe:

This technique works because the attackers place a malicious SystemSettings.dll in the same directory. When SystemSettings.exe runs, Windows loads the attacker-controlled DLL instead of only trusted components.
In dropper cases, the execution process is more complex: one sample embedded a legitimate Cisco AnyConnect MSI installer in its resource section. The dropper operates as follows:
This dual-purpose behavior makes the malware appear completely legitimate to the victim, while secretly deploying the real malicious payload.
Other droppers extract a decoy PDF from a resource named TELEMETRY, write it to %TEMP% or %TEMP%\aswerf, and launch it via cmd.exe /c. This social-engineering layer distracts the user while installation is complete.
Execution then proceeds into SharkLoader's multi-stage architecture:
One key technical detail is that SharkLoader does not simply load a DLL and spawn the Beacon; instead, it uses an advanced process centered on “Perfect DLL Hijacking.”
Once SystemSettings.dll is loaded, SharkLoader manipulates undocumented internal Windows loader mechanisms in ntdll.dll, including:
It uses LeaveCriticalSection to forcibly release the loader lock, InterlockedDecrement64 to decrement the loader state, and SetEvent to signal completion, before safely creating a new thread from the DllMain path. This technique is critical, because calling thread-creation APIs from within DllMain normally risks deadlock due to the Windows loader lock; SharkLoader deliberately manipulates the loader's internal state specifically to bypass this restriction and continue execution.
Persistence
Persistence mechanisms vary by intrusion vector.
After exploiting public-facing application vulnerabilities, the attackers establish persistence through a WebShell on the compromised server. Although researchers did not recover the actual file, command telemetry and detection results both indicate the attackers used this method.
In dropper infection cases, persistence is built directly into the installer logic. The dropper creates two scheduled tasks via the Task Scheduler COM interface:
Both point to a copy of SystemSettings.exe located in an attacker-controlled directory (such as %APPDATA%\xwreg or %APPDATA%\xgdf).
The first task runs every 5 minutes, providing stable persistence.
The second task runs every second, ensuring immediate execution right after installation; it is deleted after about 1.5 seconds, leaving only the long-term task.
In other observed incidents, the attackers also established persistence manually:
These mechanisms ensure the Side-Loading chain restarts SharkLoader after a system reboot or user login.
Privilege Escalation
While the current intelligence does not show a dedicated local privilege escalation exploit, the attackers use mechanisms that provide high-privilege execution or otherwise help obtain elevated privileges.
For example, configuring a scheduled task to run with SYSTEM privileges:

This allows SharkLoader to execute in a highly privileged context.
The attackers also target high-value credentials by dumping LSASS memory and the NTDS database. In practice, these operations let the attackers obtain and reuse administrator- or domain-level credentials, supporting privilege escalation.
Defense Evasion
Defense evasion is one of the most technically notable features of this incident.
First, the malware abuses legitimate binaries for DLL Side-Loading, particularly SystemSettings.exe, but also variant targets involving msedge.dll, PrintDialog.dll, andmiracastview.dll. This lets the malicious code execute under the guise of normal Windows or application behavior.
Second, the attackers make extensive use of masquerading:
Third, SharkLoader uses encryption and in-memory loading techniques to hide its individual stages:
Once decrypted, both payloads are written directly into memory via reflective loading rather than back to disk. Both also erase the MZ header to hinder static analysis and signature-based detection.
The reflective loading logic manually performs the following steps: parsing the PE header, allocating memory, aligning sections, applying relocations, resolving the import table, setting memory protections, and finally transferring execution to the in-memory image. This loader design significantly reduces forensic artifacts left ondisk.
Another particularly notable evasion feature is SharkLoader's use of “Perfect DLL Hijacking” to manipulate loader internal state, allowing it to safely continue execution from DllMain without exposing the malware through obvious instability.
In addition, the malware installs extensive API hooks using Microsoft Detours and MinHook.
The decrypted SyncRes.dat DLL hooksa large set of Windows APIs, covering the following functionality:
Particularly notable:
SharkLoader also registers a Vectored Exception Handler (VEH) specifically to monitor for memory access violations (0xC0000005). When triggered, it uses VirtualProtect to restore the faulting memory page to RWX (read/write/execute) permissions and resumes execution. Although no such exceptions were observed during analysis, the mechanism is clearly designed to maintain runtime stability for protected or tampered memory regions.
Finally, DscCoreR.mui decompresses and loads MinHook to hook VirtualAlloc and Sleep. These hooks directly support the Beacon's stealthy execution:
This is a sophisticated anti-memory-scanning technique that effectively reduces the exposure ofexecutable, writable memory during idle periods.
Credential Access
The attackers attempt to steal credentials using native and legitimate tools.
Researchers observed two keycommands:

The first command creates an Install From Media (IFM) copy of the NTDS database, which may expose ActiveDirectory password hashes. The second command dumps LSASS memory, a common method for recovering credentials from memory.
These actions are technically significant, showing the attackers moving from an initial foothold to credential harvesting to expand control within the domain.
Discovery
After entering the system, the attackers conduct extensive internal discoveries, using native commands and PowerShell to enumerate:
Examples include:

Using tools such as FScan, Searchall, and Pillager further extends this capability, showing the attackers want both broad situational awareness and targeted discovery of sensitive information.
Lateral Movement
While the current intelligence provides limited direct evidence of remote code execution on other hosts, itclearly shows the attackers' preparation for lateral movement.
The attackers enumerate remote administrative shares and directories (such as \\c$) and collect information on domain users, computers, and privileged groups — common preparatory steps for lateral movement using harvested credentials.
Credential dumping plays a keyrole here. The research team explicitly notes that dumping LSASS and NTDS supports lateral movement, privilege escalation, and deeper compromise.
The use of SharpGPOAbuse also shows the attackers' interest in domain-level mechanisms, which can facilitate propagation or remote execution by manipulating Group Policy.
Collection
Collection activity in this campaign focuses on gathering system, domain, and credential data rather than confirmed document theft.
The attackers use:
This shows the collection phase focuses mainly on operational intelligence gathering and expanding access, rather than immediate data theft.
Command and Control (C2)
SharkLoader's ultimate purpose is to deploy Cobalt Strike Beacon as the campaign's primary C2 implant.
Once the final stage is ready, DscCoreR.mui performs the following:
This allows the Beacon to run entirely in memory.
Associated attack infrastructure includes the following domains:
The malware's memory hooks on VirtualAlloc and Sleep are specifically designed to help the Beacon stay stealthy while maintaining its C2 lifecycle, particularly during sleep intervals.
Impact
Although this campaign shows no destructive behavior, some actions have directly impacted system integrity and security controls.
Using SharpGPOAbuse to modify Group Policy Objects can alter enterprise configurations and grant broader control over the victim environment.
At the malware level, SharkLoaderalso manipulates normal process and memory behavior, thereby undermining system integrity:
While these actions are not destructive in the traditional sense, they do alter system behavior to serve the attackers' interests and undermine trust in normal operating system and logging mechanisms.
In summary, the technical core ofthe StrikeShark incident is a carefully engineered, layered execution chain: opportunistic exploitation or social-engineering delivery triggers SharkLoader installation; SharkLoader abuses DLL Side-Loading and loader-lock manipulationto run safely; encrypted stages are decrypted and reflectively loaded inmemory; API hooks and memory protection tricks skillfully evade detection; and Cobalt Strike Beacon is ultimately launched as the operator's interactive foothold. It is this combination of commodity intrusion techniques with advanced in-memory execution that makes this campaign both easy to replicate at scale and difficult to detect.
Mitigating the StrikeShark threat requires focusing on the two most exposed links in the attack chain: initial access via unpatched vulnerabilities, and post-compromise persistence and credential theft.
Reduce Exposure of Internet-Facing Services
This campaign repeatedly abuses Internet-facingenterprise applications and devices, so the top priority is identifying, patching, and hardening all externally exposed systems. Based on the incidents described, organizations should promptly review:
Because the attackers rely heavily on known vulnerabilities with public exploit code, timely patching and risk management are two of the most effective defenses. This directly affects infrastructure teams responsible for:
If patches cannot be applied immediately, organizations should implement interim hardening measures, such as:
Monitor WebShells and Anomalous Server-Side Activity
After successful exploitation, the attackers reportedly use WebShells for follow-on activity. Defenders should therefore inspect compromised or exposed servers for:
This mitigation mainly affects:
Detect DLL Side-Loading and Living-off-the-Land Binary Abuse
A notable feature of this campaign is abusing legitimate executables such as SystemSettings.exe to side-loadmalicious DLLs. Organizations should monitor for:
This may require tuning EDR, application control, and threat-hunting content to flag:
These controls mainly affect:
Hunt Persistence via Scheduled Tasks and Run Keys
StrikeShark operators use scheduled tasks and registry Run keys to maintain access. Organizations should review systems for:
This measure affects:
Strengthen Defenses Against Credential Theft and Lateral Movement
Post-compromise activities include Active Directory enumeration, LSASS dumping, and attempts to access the NTDS database. To reduce the impact after initial compromise, organizations should:
Special attention should be paid to:
This will affect:
Improve Detection of In-Memory Malware and Cobalt Strike Activity
SharkLoader ultimately deploys CobaltStrike Beacon and uses reflective loading, API hooking, and memory protection changes to evade detection. Since much of this activity may leave no traditional file-on-disk artifacts, organizations should ensure their EDR andSOC workflows can detect:
This mitigation mainly affects:
Guard Against Malicious Droppers and Disguised Installers
Some infections use custom droppersdisguised as Google Update, Cisco AnyConnect, AutoUpdate, or document-themed files. To reduce this risk:
This mainly affects:
Use IOCs and Threat Hunting to Support Response
Organizations should also operationalize published threat indicators, including:
IOC matching alone is not sufficient, but it helps quickly identify known infections and prioritize response.
Practical Response Priorities
For security defenders, the most urgentactions right now are:
In short, mitigating this incident cannot be limited to removing malware from endpoints; it requires a fully coordinated response across Internet-facing infrastructure, Windows persistence controls, endpoint detection, Active Directory protection, and threat hunting.
Malicious Domains
Files
CVEs
File Paths
CyCraft Technology(7823.TW) is a Taiwan-listed cybersecurity company, dedicated to automatingcybersecurity with AI technology and safeguarding AI models. CyCraft’s productsuite encompasses XecART, the AI Red Teaming, and XecGuard, the Guardrail APIfor LLMs and AI Agents. The XCockpit AI platform integrates EASM, IASM, andEndpoint protection, providing preemptive and real-time defense-in-depth. Witha proven track record in the government, finance, and semiconductor sectors,and recognition from international institutions, CyCraft continues to safeguardenterprise digital resilience.