top of page

The Thief Behind the Frozen Screen: How Bazuka Turned Android Phones Into Houses With Invisible Intruders

hace 9 minutos
11 min de lectura

The house was quiet.

Its owner had installed security cameras, reinforced the doors and protected the most valuable possessions.

Every evening, he checked the surveillance monitor.

The living room appeared empty.

The windows were closed.

The furniture was exactly where he had left it.

Nothing seemed suspicious.

But what the owner didn’t know was that a thief had already entered the house.

And this thief had discovered a remarkably simple way to hide.

He took a photograph of the empty living room.

Then he placed it directly in front of the security camera.

From the owner’s perspective, nothing had changed.

The room still looked empty.

The furniture still appeared untouched.

But behind that photograph, the thief was opening drawers, collecting valuables and searching for the keys to the safe.

The security system wasn’t showing what was happening. It was showing what the intruder wanted the owner to believe.

On November 20, 2026, cybersecurity researchers disclosed details of Bazuka, a previously unknown Android malware framework discovered by ThreatFabric in late October.

Its capabilities include credential theft, remote device control, authentication-code interception and cryptocurrency fraud.

But one of its most distinctive features is a deceptive screen-overlay technique.

Bazuka can capture an image of the victim’s home screen or lock screen and display that image over the real interface.

While the victim sees an apparently ordinary screen, malicious activity can continue underneath.

The malware also includes AI-assisted victim management, illustrating how mobile threats are combining traditional banking Trojan capabilities with increasingly sophisticated automation.


Phase 1 — The Thief Chooses His House


Before entering the house, the thief needed to persuade its owner to open the door.

He couldn’t simply announce that he intended to steal everything.

He needed a disguise.

Bazuka’s distribution framework includes malicious applications that appear to offer legitimate or attractive functionality.

One observed dropper masqueraded as a premium, adult-oriented version of TikTok.

Its Android package name was:

com.tiktok.premium

The researchers identified two possible infection paths.

In one, the initial application requested permission to install software from third-party sources before downloading another APK, identified as:

In another, the downloader retrieved the main malicious payload more directly.

The framework also includes tools for managing traffers — criminals responsible for directing victims toward malicious applications through deceptive websites, advertising or other lures.

Its infrastructure supports the creation of different droppers and permission configurations.

This suggests that operators could distribute Bazuka through multiple channels, including attempts to place malicious applications in official app stores.

However, the report does not establish that the observed TikTok-themed sample was successfully distributed through Google Play.

The thief’s first disguise wasn’t a mask. It was an application that looked worth installing.


Phase 2 — The Owner Invites the Intruder Inside


The owner believed he was installing an ordinary application.

But once the software entered the house, it began preparing the way for something more dangerous.

Bazuka uses a modular architecture consisting of a dropper or downloader, a core component and 23 additional modules.

The initial application is not necessarily responsible for every malicious action.

Its purpose is to establish the infection chain and retrieve the components required for a particular campaign.

The core then acts as the central coordinator.

It communicates with the attackers’ infrastructure, loads additional DEX modules, manages permissions, collects device information and dispatches commands.

This architecture allows the malware’s capabilities to be assembled according to the operator’s objectives.

A criminal interested in banking credentials may require different modules from one seeking location data, messages or remote control.

The owner thought he had invited one harmless visitor. In reality, the visitor was preparing rooms for an entire criminal operation.


Phase 3 — The Photograph in Front of the Camera


The thief had entered the house.

Now he needed to make sure the owner wouldn’t notice what happened next.

He photographed the empty living room.

Then he placed the photograph in front of the security camera.

The owner continued checking the monitor.

Everything appeared normal.

This is the central deception behind Bazuka’s distinctive screen-overlay capability.

The malware can take a screenshot of the device’s home screen or lock screen.

It can then display that screenshot as an overlay, concealing activity taking place underneath.

Unlike a conventional phishing overlay that imitates a banking login page, this technique can reproduce the victim’s own familiar interface.

That makes the deception especially convincing.

The user may believe the device is simply showing its normal home screen or lock screen.

Meanwhile, the attacker may be interacting with the underlying system.

The technique does not mean that every infected phone is continuously displaying a fake screen.

It is a capability that the malware can activate when needed.

The thief didn’t turn off the camera. He replaced reality with a convincing image of normality.


Phase 4 — The Thief Learns to Control the House


Once hidden, the intruder began examining the building’s control systems.

He discovered switches that opened doors.

He found mechanisms that controlled the lights.

And he learned how to operate parts of the house without the owner’s involvement.

Bazuka attempts to obtain similar control through Android’s Accessibility Services and debugging capabilities.

Accessibility Services are legitimate Android features designed to help users interact with their devices.

However, malicious applications can abuse powerful accessibility permissions to observe interface elements and automate interactions.

Bazuka’s core can use accessibility functionality to navigate settings and approve certain requests.

The framework also supports Android Debug Bridge, or ADB, which is normally used for development and device administration.

ThreatFabric reported that Bazuka can navigate Android settings, enable Developer options and activate Wireless Debugging.

It can then capture the pairing PIN required to establish a connection with the ADB daemon.

This creates additional opportunities for command execution and device control.

The researchers warned that ADB-based capabilities are becoming increasingly important in modern banking Trojans.

They also cautioned that Bazuka’s current architecture had not yet reached the more advanced independent-module persistence seen in some other threats.

The thief wasn’t merely hiding in the house. He was learning to operate its doors, switches and security controls.


Phase 5 — The Criminal Opens Every Drawer


The owner kept his most valuable possessions in different rooms.

Financial records were stored in one cabinet.

Passwords were kept in another.

Private correspondence remained locked in a desk.

The thief moved from room to room.

Bazuka’s 23 modules provide a broad collection of surveillance, theft and device-control capabilities.

Among them are:

  • input_recorder, which captures text entered into applications, including potential passwords, payment-card numbers and cryptocurrency addresses.

  • smsmessage, which accesses SMS messages, extracts possible verification codes and can attempt to delete messages.

  • twofa, which targets one-time authentication codes.

  • screen_control, which can capture lock-screen PINs, passwords and unlock patterns.

  • contacts, which collects contact information.

  • location, which obtains device-location information.

  • file_manager, which accesses files.

  • media, which supports photo capture and camera or microphone streaming.

  • optics, which enables screen capture, remote interaction and phishing overlays.

These capabilities extend beyond ordinary credential theft.

They can expose private communications, financial information and sensitive device data.

The modules do not prove that every infected device experienced every form of theft.

They describe what the framework is designed to support.

The thief had entered for valuables. But once inside, he discovered that almost every room contained something worth stealing.


Phase 6 — The Thief Replaces the Safe’s Destination


The criminal discovered another opportunity.

Sometimes the owner transferred money or valuables to another location.

Rather than stealing the contents directly, the thief could change the destination.

Bazuka includes a clipboard-monitoring module called:

clip_monitor

Its purpose is to identify cryptocurrency wallet addresses copied to the clipboard and replace them with addresses controlled by the attacker.

This is commonly known as cryptocurrency address swapping or clipper behavior.

A victim might copy the intended recipient’s wallet address.

The malware could substitute another address before the victim pasted it into a transaction.

If the user failed to notice the change and completed the transfer, the cryptocurrency could be sent to the criminal.

Because blockchain transactions are generally difficult or impossible to reverse, the consequences can be significant.

The framework also includes functionality for interacting with NFC-based contactless banking data.

The report establishes the presence of these capabilities, not a confirmed financial-loss total attributable to Bazuka.

The thief didn’t always need to open the safe. Sometimes he only needed to change where its contents were being sent.


Phase 7 — The Criminal Silences the Alarm


The house had security mechanisms intended to detect intruders.

The thief understood that these protections could interfere with his work.

So he attempted to disable them.

Bazuka includes functionality that abuses accessibility permissions to interfere with Google Play’s protective components.

The framework also includes an uninstall_shield module intended to resist attempts to uninstall or force-stop the malicious application.

Other components inspect the runtime environment and perform anti-analysis checks.

These capabilities can make detection and removal more difficult.

However, their presence does not mean the malware can defeat every Android security control on every device.

Success depends on factors such as the Android version, permissions, device configuration and the effectiveness of security protections.

The key concern is that Bazuka does not merely attempt to steal information.

It also includes mechanisms intended to preserve its operational access.

The thief wanted more than a hiding place. He wanted the alarm system to stop interfering with his work.


Phase 8 — The Thief Communicates With His Accomplices


The intruder was not working entirely alone.

Outside the house, other criminals were waiting for information.

They needed to know what he had discovered.

They also wanted to give him new instructions.

Bazuka communicates with its command-and-control infrastructure through two WebSocket channels.

The first is the control channel.

It allows operators to issue commands to the infected device.

The second is the data channel.

It supports the transfer of stolen information and screen-capture streams.

Separating control and data functions helps organize the malware’s operations.

The framework can receive instructions, collect information and send results back to the attackers.

This architecture also enables remote interaction with compromised devices.

The presence of command-and-control communication is central to understanding Bazuka.

It is not simply an application that steals information once and disappears.

It is designed as a remotely managed malware platform.

The thief was inside the house, but his accomplices could watch the operation and tell him which room to search next.


Phase 9 — The Criminal Hires an Artificial Intelligence Assistant


The operation was becoming difficult to manage.

There were potentially many houses.

Some contained little of value.

Others belonged to wealthy owners.

The criminal needed to decide where to focus his attention.

So he introduced an AI assistant into the control room.

Bazuka’s management panel supports integration with large language models.

Operators can use locally hosted models or access models through an API.

ThreatFabric identified prompts suggesting that the AI can help evaluate victims and assign scores reflecting their potential value.

This could allow criminals to prioritize particular devices or victims.

The panel also supports the management of malware distributors and the creation of droppers.

These features illustrate how AI can be integrated into the operational management of cybercrime.

Importantly, the researchers did not establish that Bazuka’s AI independently carries out every stage of an intrusion.

The demonstrated role is associated with victim management, prioritization and operator assistance.

The thief still chose to commit the crime. But now he had an assistant helping him decide which houses deserved the most attention.


Phase 10 — The Thief Erases His Footprints


After searching the house, the criminal needed a way to disappear.

He didn’t want investigators to find his equipment.

He didn’t want the owner to discover evidence of his movements.

Bazuka includes a self-destruct mechanism.

According to ThreatFabric, the malware can clear local storage, delete its database and disable the application.

Such functionality may help attackers reduce the evidence left on an infected device.

It can also complicate forensic investigation.

But self-destruction should not be confused with guaranteed elimination of every trace.

Network logs, external telemetry, backups and other artifacts may still provide evidence of malicious activity.

Nor does the presence of this feature prove that it was activated in every observed infection.

It is one of the capabilities available to operators.

The thief had prepared not only an entrance and a disguise, but also a way to make his own equipment disappear.


Phase 11 — Another Criminal Uses a Black Screen


ThreatFabric’s findings appeared alongside separate research into another Android malware family.

That malware, ArcticRAT, was analyzed by Zimperium.

It also combines mobile surveillance, remote control and AI-related capabilities.

However, its concealment technique differs from Bazuka’s.

While Bazuka can display a captured image of the home or lock screen, ArcticRAT can display a completely black overlay and reduce screen brightness to zero.

To the victim, the phone may appear locked or switched off.

Behind that black screen, the malware can perform remote interactions and attempt to approve permissions.

ArcticRAT has also evolved to support a broad command set and uses language-model APIs to help process stolen information.

Its reported capabilities include examining SMS content and identifying potentially valuable financial data.

The two malware families should not be treated as the same operation.

The article does not establish a shared operator or common development team.

But their concealment strategies illustrate a common objective.

One criminal showed the owner a photograph of an empty room. Another simply made the security monitor appear switched off. Both wanted to conceal what was happening inside.


Phase 12 — How to Protect the House


Defending against malware such as Bazuka requires more than checking whether a phone appears to be working normally.

Organizations and individuals should focus on preventing installation, restricting dangerous permissions and identifying suspicious device behavior.

1. Avoid unofficial application modifications.

Be particularly cautious with supposedly premium, unlocked or modified versions of popular applications.

2. Restrict installation from unknown sources.

Do not enable third-party APK installation unless there is a clear, legitimate reason.

3. Review Accessibility permissions.

Only trusted applications with a genuine need should receive powerful accessibility access.

4. Keep debugging features disabled.

Developer options, USB debugging and Wireless Debugging should remain off when not required.

5. Maintain Android security updates.

Install operating-system and security updates promptly.

6. Keep Google Play Protect enabled.

Investigate unexpected attempts to disable or interfere with protective services.

7. Watch for suspicious overlays.

Unexpectedly frozen screens, unusual lock-screen behavior or applications requesting broad screen-control permissions deserve investigation.

8. Protect financial transactions.

Verify cryptocurrency wallet addresses carefully before confirming transfers.

9. Use mobile threat defense where appropriate.

Organizations managing Android fleets should monitor risky permissions, suspicious applications and unusual device-control activity.

10. Respond carefully to suspected compromise.

If a device may be infected, avoid entering sensitive credentials or conducting financial transactions on it. Seek trusted technical assistance, review account activity from a clean device and consider a properly managed device reset when warranted.

A factory reset can remove many forms of application-based malware, but recovery should include investigation of compromised accounts and credentials.

The best defense is to prevent the thief from entering. The next best is to recognize that a perfectly normal-looking security monitor may not be showing the truth.


Phase 13 — The Danger of Trusting What the Screen Shows


The most striking aspect of Bazuka is not simply the number of modules it contains.

It is the way the malware attacks the user’s perception of the device.

People rely on visual feedback.

They expect the screen to show what the phone is doing.

They interpret familiar interfaces as signs that the system is operating normally.

Bazuka exploits that assumption.

By displaying a screenshot over the real interface, it can separate what the user sees from what the device is actually doing.

That is a powerful form of deception.

The technique also illustrates a broader cybersecurity principle.

An attacker who gains sufficient control over an endpoint may be able to manipulate the interface through which the victim observes that endpoint.

The user may not have a reliable view of the underlying activity.

This makes independent controls particularly important.

Device-management telemetry, security monitoring, strong permission boundaries and trusted recovery procedures can provide information beyond the potentially compromised interface.

When the attacker controls the screen, the screen itself can become part of the attack.


The Final Scene — The Empty Room That Wasn’t Empty


The owner returned home.

He checked the surveillance monitor.

The living room looked exactly as it had that morning.

The furniture was untouched.

The doors were closed.

The lights were off.

Everything seemed normal.

But when he entered the room, he discovered the truth.

The drawers had been opened.

The safe had been emptied.

Private documents were missing.

The thief had been working inside the house while the owner watched a photograph of an empty room.

And somewhere beyond the property, the criminal’s accomplices had been receiving information and issuing instructions.

Bazuka demonstrates how modern Android malware can combine traditional information theft with deceptive interface manipulation, modular remote-control capabilities and AI-assisted victim management.

Its 23 modules provide the building blocks for a wide range of malicious operations.

Its control infrastructure allows operators to direct infected devices.

Its screen-overlay technique can hide activity behind a familiar image.

And its AI integration may help criminals identify the victims they consider most valuable.

ThreatFabric has not publicly attributed Bazuka to a known threat actor, and the article does not establish the total number of victims or confirmed financial losses.

But the framework’s capabilities are a warning about the evolution of mobile malware.

The thief no longer needs the owner to look away.

He can manipulate what the owner sees.

The photograph remained in front of the camera.

The monitor still displayed an empty room.

And behind it, the house was being robbed.


The most dangerous thing about the thief wasn’t that he had entered unnoticed.

It was that he had convinced the owner he was never there.


 
 
 

Comentarios


bottom of page