November 22, 2026
By Burak Oktenli
Key Takeaways:
By Burak Oktenli
Key Takeaways:
AI-assisted tools such as ARTEX and residential proxy networks can route attacks through ordinary household internet connections, so a logged IP address may identify only the final relay rather than the actual operator.
Investigations of IPIDEA, NetNut and similar services show that free VPNs and third-party apps can silently enroll devices as exit nodes, with hundreds of threat groups observed using those addresses in short periods.
The article argues that investigators, platforms and banks must treat residential IPs as investigative starting points, not proof of guilt, and must notify users, disrupt proxy businesses and distinguish victims of concealed relaying from the attackers who control them.
AI-led intrusions and hidden residential proxy networks expose an uncomfortable distinction: a digital trail can identify the household that relayed an attack without identifying the attacker.
A bank receives an automated warning about an attempted account takeover. Its logs point to an ordinary home internet connection in another country. The residents are asleep, their router is on, and nobody in the household has tried to enter the bank. The address is real. The apparent suspect may be entirely wrong.
This is a hypothetical scenario, but the machinery that makes it possible is documented. Cybercriminals can route requests through residential proxy networks that borrow the internet connections of homes and small businesses. Some people knowingly rent out their bandwidth. Others are enrolled through software that does not clearly disclose what it is doing. For an investigator, the visible source of the traffic may be the last relay rather than the person controlling the operation.
That distinction has become newly urgent as AI-assisted intrusion tools spread. A November 20 report in the New York Post described how a phone could be used to direct ARTEX, a framework that delegates parts of hacking activity to AI agents running on other infrastructure. It did not demonstrate that ARTEX secretly installs itself on iPhones. A phone acting as an operator’s remote screen is a different proposition from an unsuspecting household device being used as a network relay. Mixing the two turns a real warning into a misleading one.
The infrastructure behind the borrowed address
The immediate connection is a finding from Check Point Research’s November 5 investigation. Researchers examined exposed ARTEX operator consoles and found that some operators configured their agents to use residential proxies to obscure traffic and work around blocking. That shows a tactic being used. It does not identify which homes or phones supplied those proxies, or prove that their owners were unaware of the arrangement.
Separate investigations explain how unwitting participation can occur. In January, Google reported disrupting the IPIDEA network, whose operators used free VPN products and code embedded in third-party applications to recruit devices as proxy exit nodes, frequently without adequate disclosure. Some of the applications still performed the service users expected. That is part of the deception: an app can look useful while quietly selling access to its customer’s internet connection.
Google reported observing more than 550 tracked threat groups using IPIDEA exit addresses during a single seven-day period that month, including espionage and criminal actors from several regions. In July, Google and partners, including the FBI, announced further action against a different network, NetNut. Its researchers observed 316 distinct threat clusters using suspected NetNut exits during a week in June. These are the investigators’ observed counts, not a census of all attacks or proof that either network carried the ARTEX traffic documented by Check Point.
The FBI warned the public in March that residential proxies can conceal the origins of cybercrime by using residential and small-business internet networks. That warning is global in consequence: a relay in one country may carry activity planned, financed and directed from several others.
An internet address is not a human identity
Residential proxies have lawful uses, including privacy protection and legitimate network testing. The policy problem is concealed enrollment and criminal use, not the mere existence of proxy technology. A public IP address establishes where traffic appeared to exit. It does not, on its own, establish who typed the instruction, whether the connection was shared, or whether software secretly relayed the request.
That should change how banks, employers and investigators handle suspicious traffic. Blocking an active attack may be necessary immediately. Accusing a customer, employee or household member should require corroboration: account and device evidence, session history, relevant application activity and, where lawful and proportionate, provider records. An IP address is a starting point for investigation, not a verdict.
The risk also runs in the other direction. Google’s IPIDEA investigation found that proxy software could expose consumer devices and other equipment on the home network to further abuse. A household recruited as a relay is not automatically innocent of every security problem, but it may simultaneously be a victim whose connection is being exploited. Making that possibility visible is as important as tracing the intrusion.
Make the hidden trade visible
The first responsibility belongs to software distributors and app stores. Google Play’s published policy permits apps to facilitate third-party proxy services only when doing so is their primary, user-facing purpose. Apple’s review rules also restrict unrelated background tasks and undocumented app behavior. Neither policy is a guarantee that covert proxy code will never be distributed. Both show that the industry already recognizes a basic principle: users should not have to infer that their connectivity is being resold from an obscure software dependency.
When concealed relaying is discovered, platform operators should notify affected users in plain language, identify the app and affected versions, explain how to remove it and say whether their connection may have been used by others. A generic “security improvement” notice is inadequate. Internet providers and security vendors should share confirmed indicators through lawful channels, without treating every residential address associated with a proxy network as a criminal address.
Cross-border enforcement also needs to follow the business rather than just the last IP address: the firms distributing software kits, the operators selling proxy access and the resellers obscuring who supplied it. Google’s January and July interventions offer a model of technical disruption, platform enforcement and law-enforcement cooperation. The limit is equally clear: taking down one network does not automatically remove every reseller or every device already enrolled.
For ordinary users, the practical precautions are modest: avoid VPNs and “earn money from unused bandwidth” applications from untrusted suppliers, examine whether a service shares your connection with third parties, keep device protections active and investigate unexpected account warnings. For authorities, the obligation is harder: protect victims of abuse without confusing the address they use with the actor who chose the attack.
The most important new digital divide may not be between those who can use AI and those who cannot. It may be between the people who control an automated operation and the people whose connections quietly carry it. Global cybersecurity will be more credible when it can tell those two groups apart.
About Burak Oktenli
Burak Oktenli holds an MBA and a Master of Professional Studies in Applied Intelligence from Georgetown University. His research addresses the governance of authority in autonomous and AI-enabled systems, and his writing has appeared at the Modern War Institute at West Point, RUSI, RealClearDefense, RealClearMarkets, and Geopolitical Monitor. He is the author of Authority Architectures for Autonomous Systems, a ten-volume series on how authority in autonomous systems is delegated, monitored and recovered, at authority-architecture.me.
View all posts by Burak Oktenli →

No comments:
Post a Comment