This week's ThreatsDay Bulletin highlights a disturbing trend in 2026: adversaries are increasingly disguising their malicious activities as useful or benign functionality. From Android applications masquerading as safety tools while exfiltrating data, to AI image prompts containing hidden commands that bypass traditional security controls, the threat landscape continues to evolve in sophistication and impact. For security practitioners, this means expanding detection capabilities beyond traditional indicators of compromise to include behavioral analysis of trusted applications, AI-powered systems, and operational technology environments.
Technical Analysis
1. Android Spyware Masquerading as Safety Applications
Malicious Android applications appearing as legitimate safety or security tools represent a significant threat to mobile endpoints. These apps request excessive permissions upon installation, including access to contacts, messages, microphone, camera, and location data. The collected data is transmitted to command and control (C2) servers using encrypted channels to evade detection. Some variants exploit Android's accessibility services to maintain persistence and evade uninstallation, resulting in complete compromise of mobile device privacy and potential infiltration of corporate networks through VPN profiles.
2. PLC (Programmable Logic Controller) Attacks
Attacks targeting Industrial Control Systems (ICS) and Operational Technology (OT) environments continue to pose critical infrastructure risks. Adversaries exploit vulnerabilities in PLC firmware and communication protocols (Modbus, DNP3, OPC UA) to manipulate controller logic. These attacks often exploit weak authentication or protocol vulnerabilities, potentially causing disruption of critical infrastructure, physical damage to equipment, and safety hazards. Detection challenges are significant as traditional security tools often lack visibility into OT protocol communications.
3. AI Image Prompt Injection
A sophisticated attack vector where images containing embedded text instructions are processed by AI agents. Malicious actors encode instructions within images that bypass text-based input filters, causing AI agents to execute unauthorized commands, potentially accessing restricted data or performing unwanted actions. These attacks target AI-powered customer service bots, automated assistants, and image analysis systems. Traditional input validation fails against these multi-modal attacks that separate visual and textual content.
4. Fake Browser Extensions Opening Remote Access
Malicious browser extensions distributed through third-party repositories exploit browser APIs to establish remote access connections. These extensions utilize legitimate update mechanisms to maintain persistence, resulting in remote code execution, credential theft, and browser session hijacking.
Detection & Response
---
title: Android Spyware Data Exfiltration
id: a7b3c9d4-1e5f-4a2b-8c6d-9e1f2a3b4c5d
status: experimental
description: Detects potential Android spyware behavior including excessive permission requests and data exfiltration patterns.
references:
- https://attack.mitre.org/techniques/T1119/
- https://attack.mitre.org/techniques/T1560/
author: Security Arsenal
date: 2026/07/14
tags:
- attack.collection
- attack.exfiltration
- attack.t1119
- attack.t1560
logsource:
product: android
category: network_connection
detection:
selection:
DestinationPort|endswith:
- '443'
- '8443'
Initiated: 'true'
AppName|contains:
- 'safety'
- 'security'
- 'emergency'
- 'protect'
condition: selection and not filter
falsepositives:
- Legitimate safety applications
level: high
---
title: PLC Unusual Protocol Activity
id: b8c4d0e5-2f6a-5b3c-9d7e-0f2a3b4c5d6e
status: experimental
description: Detects anomalous protocol activity in ICS/OT environments that may indicate PLC manipulation attempts.
references:
- https://attack.mitre.org/techniques/T0885/
- https://attack.mitre.org/techniques/T0869/
author: Security Arsenal
date: 2026/07/14
tags:
- attack.ics
- attack.t0885
- attack.t0869
logsource:
product: industrial_control_system
category: network_connection
detection:
selection:
Protocol|contains:
- 'modbus'
- 'dnp3'
- 'opcua'
SourceUser|re: 'Admin|Operator|Engineer'
BytesOut|gt: 10000
condition: selection
falsepositives:
- Legitimate engineering updates
- Scheduled maintenance activities
level: high
---
title: Browser Extension Remote Access Behavior
id: c9d5f1a6-3g7b-6c4d-0e8f-1g3b4c5d6e7f
status: experimental
description: Detects browser extensions exhibiting remote access behaviors such as unauthorized WebSocket connections and unusual API calls.
references:
- https://attack.mitre.org/techniques/T1219/
- https://attack.mitre.org/techniques/T1071/
author: Security Arsenal
date: 2026/07/14
tags:
- attack.command_and_control
- attack.t1219
- attack.t1071
logsource:
product: browser
category: extension
detection:
selection:
EventType: 'websocket'
ExtensionPermissions|contains:
- 'idle'
- 'background'
- 'webRequest'
DestinationHostname|contains:
- '.duckdns.org'
- '.no-ip.org'
- '.ddns.net'
condition: selection
falsepositives:
- Legitimate remote access tools
level: high
---
title: AI Agent Prompt Injection via Image
id: d0e6g2b7-4h8c-7d5e-1f9g-2h4c5d6e7f8g
status: experimental
description: Detects potential prompt injection attacks through image inputs to AI agents.
references:
- https://attack.mitre.org/techniques/T1566/
- https://attack.mitre.org/techniques/T1204/
author: Security Arsenal
date: 2026/07/14
tags:
- attack.initial_access
- attack.t1566
- attack.t1204
logsource:
product: ai_platform
category: prompt_injection
detection:
selection:
InputType: 'image'
ImageTextExtraction|contains:
- 'ignore previous instructions'
- 'override'
- 'execute'
- 'run'
- 'system:'
condition: selection
falsepositives:
- Legitimate testing and development
level: medium
KQL (Microsoft Sentinel / Defender)
// Detect Android spyware data exfiltration patterns
DeviceNetworkEvents
| where ActionType in ("ConnectionAccepted", "ConnectionInitiated")
| where RemotePort in (443, 8443, 9999)
| where Timestamp >= ago(7d)
| where InitiatingProcessFolderPath contains "/data/app/" and (
InitiatingProcessName contains "safety" or
InitiatingProcessName contains "security" or
InitiatingProcessName contains "emergency" or
InitiatingProcessName contains "protect"
)
| where SentBytes > 1000000
| project Timestamp, DeviceName, InitiatingProcessName, RemoteUrl, RemoteIP, SentBytes
| order by Timestamp desc
// Detect anomalous PLC protocol activity
CommonSecurityLog
| where DeviceVendor contains "Siemens" or DeviceVendor contains "Rockwell" or DeviceVendor contains "Schneider"
| where Protocol in ("Modbus", "DNP3", "OPC-UA")
| where Timestamp >= ago(7d)
| where SentBytes > 10000
| summarize count(), max(SentBytes) by Protocol, SourceUserName, DestinationIP
| where count_ > 100
| project Protocol, SourceUserName, DestinationIP, count_, max_SentBytes
// Detect browser extension remote access behaviors
DeviceNetworkEvents
| where ActionType in ("ConnectionInitiated", "DnsQuery")
| where InitiatingProcessName endswith ".exe"
| where InitiatingProcessFolderPath contains "Extensions"
| where RemoteUrl contains "duckdns.org" or RemoteUrl contains "no-ip.org" or RemoteUrl contains "ddns.net"
| project Timestamp, DeviceName, InitiatingProcessName, RemoteUrl, RemoteIP, LocalPort
| order by Timestamp desc
// Detect AI prompt injection via images
let suspiciousKeywords = dynamic(["ignore previous instructions", "override", "execute", "run", "system:", "ignore all", "new instructions"]);
AuditLogs
| where Timestamp >= ago(7d)
| where OperationName contains "AI" and (OperationName contains "Prompt" or OperationName contains "Generation")
| where RequestResources contains "image" or RequestResources contains "Image"
| extend isSuspicious = coalesce(array_index_any(RequestResources, (item: string) => item has_any(suspiciousKeywords)), false)
| where isSuspicious == true
| project Timestamp, OperationName, Caller, RequestResources, ObjectId
| order by Timestamp desc
Velociraptor VQL
-- Hunt for Android applications with suspicious permission combinations
SELECT
PackageName,
AppName,
VersionName,
Permissions,
InstallTime,
LastUpdateTime,
SourceDir,
UID
FROM android_packages()
WHERE Permissions =~ 'android.permission.READ_SMS'
AND Permissions =~ 'android.permission.ACCESS_FINE_LOCATION'
AND Permissions =~ 'android.permission.RECORD_AUDIO'
AND (AppName =~ 'safety' OR AppName =~ 'security' OR AppName =~ 'emergency' OR AppName =~ 'protect')
-- Hunt for browser extensions with remote access capabilities
SELECT
Name,
ID,
Version,
Enabled,
Path,
Permissions,
InstallTime
FROM chrome_extensions()
WHERE Permissions =~ 'background'
AND Permissions =~ 'webRequest'
AND Permissions =~ 'tabs'
-- Hunt for processes with network connections to suspicious domains
SELECT
Pid,
Name,
Username,
CommandLine,
Cwd,
Exe,
CreateTime
FROM pslist()
JOIN ON Pid = Pid
(
SELECT
Pid,
RemoteAddress,
RemotePort,
State
FROM netstat()
WHERE RemoteAddress =~ 'duckdns.org'
OR RemoteAddress =~ 'no-ip.org'
OR RemoteAddress =~ 'ddns.net'
)
-- Hunt for recent AI prompt processing logs
SELECT
Timestamp,
Service,
EventType,
Prompt,
InputType,
Output,
SourceIP,
User
FROM file(files="/var/log/ai_service/*")
WHERE EventType =~ "prompt_processing"
AND InputType =~ "image"
AND (Prompt =~ "ignore" OR Prompt =~ "override" OR Prompt =~ "execute" OR Prompt =~ "run")
Remediation Script (PowerShell for Windows systems)
# Script to check for and remediate suspicious browser extensions
# Run this on endpoints with browser extensions installed
# Function to check Chrome extensions
function Check-ChromeExtensions {
$chromeExtensionsPath = "$env:LOCALAPPDATA\Google\Chrome\User Data\Default\Extensions"
if (Test-Path $chromeExtensionsPath) {
$extensions = Get-ChildItem -Path $chromeExtensionsPath -Directory
foreach ($ext in $extensions) {
$manifestPath = Join-Path -Path $ext.FullName -ChildPath "*\manifest."
if (Test-Path $manifestPath) {
$manifest = Get-Content -Path $manifestPath | ConvertFrom-Json
# Check for suspicious permissions
$suspiciousPermissions = @("background", "webRequest", "idle")
$hasSuspiciousPerms = $false
if ($manifest.permissions) {
foreach ($perm in $manifest.permissions) {
if ($perm -in $suspiciousPermissions) {
$hasSuspiciousPerms = $true
}
}
}
# Report suspicious extensions
if ($hasSuspiciousPerms) {
Write-Output "Suspicious extension found: $($manifest.name), ID: $($ext.Name), Permissions: $($manifest.permissions -join ', ')"
# Option to remove extension
$remove = Read-Host "Do you want to remove this extension? (Y/N)"
if ($remove -eq "Y") {
Remove-Item -Path $ext.FullName -Recurse -Force
Write-Output "Extension removed: $($ext.Name)"
}
}
}
}
}
}
# Function to check for suspicious network connections
function Check-SuspiciousConnections {
$suspiciousDomains = @("duckdns.org", "no-ip.org", "ddns.net")
$connections = Get-NetTCPConnection -State Established -ErrorAction SilentlyContinue
foreach ($conn in $connections) {
try {
$process = Get-Process -Id $conn.OwningProcess -ErrorAction SilentlyContinue
$remoteAddress = [System.Net.Dns]::GetHostEntry($conn.RemoteAddress).HostName
foreach ($domain in $suspiciousDomains) {
if ($remoteAddress -like "*$domain*") {
Write-Output "Suspicious connection found: Process: $($process.ProcessName), PID: $($conn.OwningProcess), Remote: $remoteAddress"
# Option to kill the process
$kill = Read-Host "Do you want to terminate this process? (Y/N)"
if ($kill -eq "Y") {
Stop-Process -Id $conn.OwningProcess -Force
Write-Output "Process terminated: PID $($conn.OwningProcess)"
}
}
}
} catch {
# Skip errors in resolving hostnames
continue
}
}
}
# Main execution
Write-Output "Starting security check..."
Check-ChromeExtensions
Check-SuspiciousConnections
Write-Output "Security check completed."
Remediation
Android Spyware
- Uninstall suspicious applications immediately
- Perform factory reset on affected devices
- Enforce application allowlisting for BYOD environments
- Implement Mobile Device Management (MDM) solutions with app vetting capabilities
- Educate users about the risks of downloading safety/security apps from unofficial sources
- Regular security audits of installed applications and their permissions
PLC Attacks
- Implement network segmentation to isolate OT networks from IT networks
- Deploy specialized ICS/OT monitoring solutions
- Regularly update PLC firmware following validated change management procedures
- Implement strict access controls for engineering workstations
- Conduct regular security assessments of ICS environments
- Implement anomaly detection for industrial protocols
AI Image Prompt Injection
- Implement multi-modal input validation for AI systems
- Apply OCR processing to images to detect embedded text commands
- Implement sandboxed environments for AI agent execution
- Monitor AI system logs for unexpected command patterns
- Implement rate limiting and anomaly detection on AI prompt submissions
- Regularly test AI systems against adversarial inputs
Fake Browser Extensions
- Restrict extension installation to approved repositories
- Implement extension allowlisting policies
- Regularly audit installed extensions and their permissions
- Deploy browser security configurations that limit extension capabilities
- Educate users about extension security risks
- Implement endpoint detection for suspicious extension behavior
General Security Measures
- Maintain an updated asset inventory across IT, OT, and mobile environments
- Implement centralized logging and monitoring
- Conduct regular security awareness training for all users
- Establish incident response playbooks for each threat type
- Implement vulnerability management for open-source components
- Regular penetration testing and security assessments
Related Resources
Security Arsenal Penetration Testing Services AlertMonitor Platform Book a SOC Assessment vulnerability-management Intel Hub
Is your security operations ready?
Get a free SOC assessment or see how AlertMonitor cuts through alert noise with automated triage.