An update on residential proxies and the scraper situation

An Update on Residential Proxies and the Scraper Situation: Two Major Takedowns, One Persistent Problem

More than a year after LWN.net first documented the AI scraper bot scourge tearing through independent web publishers, the situation has both worsened and—for the first time—begun to attract the kind of coordinated law-enforcement response that might actually move the needle. The update on the scraper situation published by LWN in July 2026 lands in the same week that the FBI and Google announced the takedown of the NetNut residential proxy network. The timing is not coincidental: the two stories are essentially the same story.

This comprehensive update on residential proxies and the scraper situation examines how these networks operate at the infrastructure level, what the back-to-back disruptions of IPIDEA and NetNut actually mean in practice, what defenders are doing to protect themselves, and why the underlying economic incentives that created this ecosystem remain largely unchanged. For publishers, security teams, and policymakers trying to understand the residential proxy problem, this represents the most current picture of a landscape that is actively transforming.

How Residential Proxy Networks Turn Your Devices Into Weapons

To understand why the scraper problem is so difficult to stop, you need to grasp what a residential proxy network actually is at the infrastructure level—because it is not what the marketing materials claim.

The basic mechanism is straightforward and predatory. Software running on ordinary consumer devices—smart TVs, Android streaming boxes, mobile phones, Windows PCs—receives orders from operator-controlled command-and-control (C2) nodes. It fetches web pages on behalf of those operators and sends the retrieved data back. The device owner typically has no idea this is happening. Either the proxy code was bundled silently into another application, the terms of service were deliberately obscured, or the installation was not consensual at all. From the perspective of the target website, the request looks like it came from a legitimate residential or mobile IP address, which makes it nearly impossible to distinguish from a real human visitor.

The IPIDEA Architecture: A Case Study in Proxy Infrastructure

Google's Threat Intelligence Group (GTIG) documented the IPIDEA network's two-tier C2 architecture in exhaustive detail, providing a template for understanding how residential proxy networks actually work at scale. Enrolled devices first connect to a small set of stable Tier One domains at startup. These domains deliver configuration instructions, initialization parameters, and botnet version updates. Devices then poll approximately 7,400 rotating Tier Two servers daily for proxy tasking—that is, instructions on which websites to scrape, which URLs to fetch, and where to send the retrieved data. This two-tier design provides operational resilience: even if Tier Two servers get disrupted or blocked, the C2 relationship survives as long as Tier One keeps running.

The network enrolled devices through four distinct SDKs—Castar, Earn, Hex, and Packet—reaching across Android, Windows, iOS, and WebOS platforms. This multi-platform approach was deliberately engineered to maximize device diversity and geographic spread. Google found over 600 Android applications with IPIDEA C2 connections baked into their code, plus 3,075 unique Windows PE file hashes making DNS requests to Tier One domains. The sheer number of entry points into the network meant that blocking a single malicious app had virtually no impact on the botnet's overall capacity.

Attack Patterns and Evasion Techniques

The attack pattern that emerges from this infrastructure is specifically designed to defeat traditional blocking mechanisms. Scraping jobs get distributed across millions of unique IP addresses over hours or days, with each individual address hitting a target site just two or three times before getting cycled out. By the time a site operator has enough evidence to identify an address as malicious, that address will never be used again. As LWN's analysis puts it plainly: blocking is ineffective against this pattern. Rate-limiting per IP is equally useless when the attacker has millions of addresses at their disposal.

The attacker-controlled metadata—user-agent strings, accept headers, referrer information—is crafted to look like a human browser. Bots typically skip fetching images and CSS files, which is detectable but easily circumvented. A sophisticated scraper operator can add image and CSS requests to their scraping jobs to further hide the automated nature of the traffic. Cookie handling, JavaScript execution simulation, and behavioral fingerprinting can all be mimicked by modern proxy infrastructure, making client-side detection unreliable.

"What isn't clear is who is using the residential proxies; somebody is paying them to run these attacks on web sites." — LWN, "An update on the scraper situation" (July 2026)

That economic observation is the heart of the whole problem. These are not hobbyist scrapers or passive infrastructure providers. They are paid services operating at industrial scale, with multiple revenue streams: they charge clients for bandwidth and IP diversity; they receive payments from app developers for embedding SDK code; and in some cases they operate affiliate programs or reseller networks that spread the proxy code even further. Money flows from clients—often AI data brokers, competitive intelligence firms, and credential-stuffing operations—through proxy operators, through SDK embedding deals with app developers, all the way down to the consumer device in someone's living room, whose owner has no idea their device is being rented to attackers.

The Operator Landscape: From Criminal to "Quasi-Legitimate"

LWN's analysis draws a useful distinction between two categories of residential proxy operator, though the line between them is blurrier than either side would like to admit. Often the difference comes down to disclosure and consent rather than technical capability.

Criminal Operators

At the overtly illegal end, networks enroll devices without any consent whatsoever—pure malware, without even the pretense of disclosure. The IPIDEA and NetNut networks both fall into this category, despite NetNut's origins as a publicly traded company and despite the marketing claims made by both entities about user consent.

IPIDEA operated 13 distinct proxy and VPN brands as a reseller front, creating the illusion of competitive market choice where none actually existed. Google's GTIG documented all 13 brands: 360 Proxy, 922 Proxy, ABC Proxy, Cherry Proxy, Door VPN, Galleon VPN, IP2World, Ipidea, Luna Proxy, PIA S5 Proxy, PY Proxy, Radish VPN, and Tab Proxy. This reseller strategy served multiple purposes: it obscured the true scale of the operation from enforcement; it created compartmentalization so that disruption of one brand did not immediately wipe out customer visibility of others; and it allowed IPIDEA to price-discriminate across customer segments and geographies. When Google obtained a court order in January 2026 to take down dozens of IPIDEA-owned domains, those brands lost their upstream supply overnight, leaving resellers scrambling to pivot to alternative suppliers. NetNut subsequently gained significant market share from the fallout—only to be seized itself six months later by the FBI.

Quasi-Legitimate Operators and the Disclosure Problem

At the other end sit operators that LWN characterizes as "quasi-legitimate"—companies that operate with some degree of legal entity registration, published terms of service, and claimed user consent. Their actual consent mechanisms are questionable, however, and their marketed use cases directly target website security bypass.

Companies like Bright Data openly advertise helping customers bypass website access controls and anti-bot measures. They offer "free" VPN services where the price of admission is granting the mobile app permission to route third-party traffic through the user's device. This model creates several layers of deception: users downloading the VPN app to protect their own privacy may not realize they are simultaneously becoming proxy exit nodes for unknown third parties; the app's terms of service may technically disclose this in legal language, but the consent is not truly informed; and the operator maintains plausible deniability by claiming users consented and that the operator is merely a neutral platform.

Others offer SDKs to app developers who get paid for embedding the proxy code—a model that ranges from arguably GDPR-compliant (if consent is clear and revocation is easy) to, as LWN puts it, "overtly sleazy" (if disclosure is buried in the terms of service and users cannot disable the proxy functionality without uninstalling the app entirely). In either case, the technical outcome for the target website is identical: traffic that originates from residential IP addresses, arrives at scale, and is effectively unstoppable at the IP level.

LWN notes there is no evidence that frontier-model AI companies are directly using these criminal networks, stating: "There is no evidence (that I am aware of) that the frontier-model companies are using those networks." The client base appears to be a layer of data brokers and intermediaries operating beneath the headline names, purchasing scraped data in bulk or using proxy services for targeted data extraction projects.

The IPIDEA and NetNut Takedowns: What Actually Happened

The six months between January and July 2026 saw the two largest single disruptions of residential proxy infrastructure ever executed. Understanding both is essential for grasping any comprehensive update on the residential proxies scraper situation.

IPIDEA: January 2026 Coordinated Disruption

Google obtained a court order in January 2026 against IPIDEA, targeting the C2 domains underpinning its entire multi-brand operation. The legal action came paired with technical measures: Google configured Play Protect to warn, remove, and block IPIDEA SDK applications on certified Android devices. The company partnered with Cloudflare to disrupt IPIDEA's domain resolution through DNS filtering. This combination of legal, regulatory, and technical pressure created a cascading failure: new device enrollments became difficult as domain resolution failed; existing devices could no longer reach C2 servers for tasking; and users opening the Google Play Store would see warnings about the proxy app, creating friction for the operator's user acquisition.

The scope of GTIG's investigation is instructive. Google identified over 550 distinct threat groups from China, North Korea, Iran, and Russia using IPIDEA exit nodes in a single seven-day window in January 2026—a figure that underscores how deeply embedded the network was in nation-state-adjacent activity, not just commercial scraping. This co-mingling of use cases—commercial data brokers and state-sponsored threat actors using the same infrastructure—reflects the reality that residential proxy operators do not care who their clients are, only that they pay.

LWN observed a meaningful reduction in scraper traffic following the IPIDEA action, offering one of the first concrete signals that supply-side disruption can produce measurable effects for individual publishers. This was not theoretical; it was quantifiable harm reduction that LWN's engineers could measure in real time.

NetNut / Popa: July 2, 2026 FBI Seizure

The NetNut action was larger in scale, more complex in its technical and financial implications, and coordinated across a broader coalition than the IPIDEA case. The FBI and IRS Criminal Investigation division seized hundreds of NetNut domains on July 2, 2026, replacing the netnut.io homepage with an official seizure notice. Google's GTIG simultaneously disabled Google accounts and services used for NetNut's C2 infrastructure, revoked OAuth tokens, and disabled apps bundling NetNut's SDKs from the Google Play Store. Industry partners—Lumen Technologies (which operates abuse-fighting infrastructure), Shadowserver Foundation (which tracks botnet communications), Synthient (a proxy-tracking research service), and Black Lotus Labs (an abuse response team)—contributed technical intelligence, domain registration information, and network telemetry that supported the enforcement action.

The network's scale was staggering and unprecedented. Google's analysis identified at least 2 million compromised devices scattered worldwide, spanning residential broadband connections, mobile networks, and business-class ISP allocations. In a single week in June 2026—before the seizure—GTIG observed 316 distinct clusters of threat actors using suspected NetNut exit nodes. That figure is not merely academic: it means at least 316 distinct criminal or state-sponsored organizations had active NetNut subscriptions and were actively conducting malicious activity through the network.

GTIG also identified NetNut botnet plugin components inside Badbox 2.0—a large-scale Android and IoT-device botnet that Google had pursued with litigation in July 2025 and that is primarily associated with large-scale advertising fraud and residential proxy abuse. This overlap extended NetNut's reach beyond commercial scraping and into broader cybercrime infrastructure, making it a node in multiple attack chains simultaneously.

Alarum Technologies and Market Consequences

NetNut was operated by publicly traded Israeli firm Alarum Technologies (NASDAQ: ALAR), a company that had publicly marketed its residential proxy services and claimed to operate in compliance with legal and ethical standards. Following the seizure, the company's legal counsel stated it would "fully cooperate with law enforcement," but the market reaction was immediate and severe: Alarum's stock fell approximately 67% within a week of the seizure, trading at $2.62 per share as investors recognized the company faced criminal liability, asset seizure, and likely civil litigation from harmed publishers and device owners.

Benjamin Brundage, founder of proxy-tracking service Synthient, described the market dynamics directly to Krebs on Security: NetNut had become the dominant replacement for IPIDEA's traffic in the six months following that earlier takedown. "I think this takedown is going to have a big impact, because NetNut gained significant popularity after the IPIDEA takedown," Brundage told Krebs. "Also NetNut has been incredibly common among resellers, and they were on par with IPIDEA in terms of their daily traffic, quality, size, price per gigabyte, all of it." The proxy ecosystem, in other words, had simply rerouted around the first disruption, with surviving competitors absorbing the displaced traffic within weeks.

Network	Operator / Parent	Action Date	Devices Compromised	Threat Clusters Observed	Action Taken By	Known Brands / Reseller Fronts
IPIDEA	Chinese entity (unspecified)	January 2026	Millions (scale reduced by Google's action per GTIG reporting)	550+ (7-day window, Jan 2026)	Google (court order), Cloudflare, Play Protect	360 Proxy, 922 Proxy, ABC Proxy, Cherry Proxy, Door VPN, Galleon VPN, IP2World, Luna Proxy, PIA S5 Proxy, PY Proxy, Radish VPN, Tab Proxy, Ipidea (13 total)
NetNut / Popa	Alarum Technologies (NASDAQ: ALAR)	July 2, 2026	2 million+	316 (single week, June 2026)	FBI, IRS-CI, Google GTIG, Lumen, Shadowserver, Synthient, Black Lotus Labs	NetNut (primary), multiple regional resellers

Why the Problem Keeps Growing Despite the Wins

The back-to-back takedowns are genuinely significant—the largest coordinated enforcement actions against residential proxy infrastructure ever executed—but Google itself acknowledged the structural fragility of the approach in its public statements: "Creating a lasting disruption in this fluid ecosystem means we must scale our efforts to target the infrastructure of several interconnected providers." That is a careful way of saying that seizing one network simply pressurizes the others, causing demand to shift and revenue to flow through alternative channels.

The Reseller Problem and Infrastructure Fungibility

The reseller model is the core reason for this resilience. Google noted that many popular residential proxy brands are in fact whitelabeling the NetNut botnet—meaning dozens of apparent competitors are actually the same underlying infrastructure, just repackaged under different brand names, different TLDs, and different payment processing chains. When the upstream gets seized, those resellers can and do pivot to other upstream networks within days. IPIDEA itself demonstrated this pattern: after Google's January 2026 action, the operator began reselling competitors' services and rebuilt its revenue model by functioning as a pure reseller broker, taking a percentage of traffic revenue from other proxy operators.

The LWN community discussion surfaced a parallel insight worth emphasizing: even community-level defensive responses are insufficient because the incentive structure is systemic and repeated at multiple layers. Individual site operators defending themselves against scrapers is analogous to individual users defending against spam—tactically necessary, but insufficient to address the underlying economics. As long as there is demand for scraped data (for training AI models, building competitive intelligence databases, conducting account takeovers, or verifying stolen credentials), operators will continue to provide it.

Traffic Pattern Analysis and Intensity Escalation

A subscriber in the LWN discussion thread noted a 50% traffic drop on June 25, 2026—several days before the NetNut seizure—attributed to traffic carrying a distinctive "sleepbot" user-agent string going silent. This suggests some proxy operators may have received advance warning or detected enforcement activity and voluntarily took infrastructure offline. Following the July 2nd seizure, the same commenter observed a 30% drop in total scraper traffic, indicating that NetNut's seizure did produce measurable supply-side reduction.

However, LWN also reported experiencing its heaviest scraper attack ever in the period leading up to the July article—that is, in the weeks between the IPIDEA disruption in January and the NetNut seizure in July. This suggests that the actors who remain operational are intensifying their campaigns, possibly absorbing traffic from disrupted competitors and consolidating market share. The net effect is ambiguous: total attack volume may have decreased modestly, but the attacks that do occur are more focused and potentially more damaging.

The App Store SDK Vetting Gap

The app store problem deserves particular attention as a structural failure point in platform security architecture. Major platform vendors—Apple App Store, Google Play Store, device OEMs, and smart TV manufacturers—nominally prohibit non-user-authorized residential proxies in their app stores. These policies are well-intentioned and genuinely aim to prevent malware. But enforcement has been reactive rather than proactive: a malicious app gets published, spreads to hundreds of thousands or millions of devices, accumulates enough reports or gets discovered through security research, gets flagged for review, gets removed (after weeks or months), and by that time the damage is done.

LWN notes that major vendors remain conspicuously silent about why publishing residential-proxy SDK-bundled apps remains so easy in the first place. Why is there not an automated or AI-driven scanning system that examines submitted apps for known proxy SDKs before publication? Why do app stores not require explicit permission categories for "will route third-party traffic through this device"? Google Play Protect's after-the-fact scanning is an improvement over manual review, but it addresses the symptom rather than the SDK vetting gap. An attacker or proxy operator can push out a new SDK variant, get it into apps, publish to the store, and compromise millions of devices before Play Protect's ML models flag it as malicious.

What Defenders Are Actually Doing at the Application Layer

For individual publishers, the options are a set of uncomfortable trade-offs. LWN's experience illuminates why there is no clean answer that preserves both openness and defensibility.

• Proof-of-work (Anubis): Now widespread among independent sites, Anubis imposes a JavaScript-based computational challenge before serving content—requiring the client to perform tens of thousands of hash computations to generate a valid token. This approach is cryptographically sound and does not rely on blocking algorithms that can be evaded. LWN rejected it on two grounds: it introduces real latency for legitimate users (particularly on mobile devices or in regions with slower internet), and it is ultimately a delay tactic, not a solution—sufficiently motivated scraper operators will build workarounds by implementing GPU-accelerated proof-of-work solvers or simply accepting the computational overhead as a cost of doing business. Anubis also creates accessibility barriers for users on older devices or with bandwidth constraints.
• CAPTCHA / "prove you are human" services: Commercial services with challenge buttons (reCAPTCHA, hCaptcha, Cloudflare Turnstile) solve the immediate bot problem by requiring human interaction to proceed. They introduce friction for every anonymous reader, however, and outsource trust decisions to third-party vendors, with all the privacy implications that entails. These services collect behavioral and interaction data that can be used for profiling or sold to data brokers. They also create accessibility challenges for users with visual or cognitive disabilities.
• Data poisoning (iocaine): Tools like iocaine serve subtly corrupted content to detected scrapers—for example, injecting errors into data, shuffling fields, or returning plausible-but-false data—attempting to degrade the quality of training data derived from a site. This starts an arms race within an arms race: as model builders detect poisoned training data, they improve their filtering algorithms; as defenders improve their poisoning, attackers improve their detection. The economic question is whether the cost of poisoning is lower or higher than the benefit to the attacker of scraped data quality loss.
• Login gates and paywalls: LWN currently applies defensive measures to anonymous readers while logged-in subscribers pass through unimpeded. This is a reasonable pragmatic division that protects the audience that contributes revenue while accepting some scraper traffic from free tier readers. But this strategy exerts pressure toward gating previously open content, reducing publisher accessibility and potentially eroding the open web as a public resource.
• Server-side optimization and rate-limiting: LWN aggressively optimized expensive site operations and minimized computational overhead triggered during attacks. This includes caching frequently-accessed content, deferring non-critical background tasks, and compressing responses. In an ironic side effect, response times during peak scraper attacks are often better than during calm periods, because the optimizations benefit all traffic proportionally. This does not stop the scraping, but it reduces the operational cost to the publisher.
• IP reputation and behavioral analysis: Some publishers use third-party IP reputation services (Akamai, Cloudflare, Imperva) that maintain databases of known proxy exit nodes, data center IP ranges, and VPN providers, blocking or rate-limiting traffic from these ranges. This is imperfect—residential proxy networks intentionally use legitimate residential IP space, and blocking too aggressively can block legitimate users behind ISP-shared infrastructure—but it can provide modest protection against less sophisticated scrapers.

Beyond these tactical measures, the broader security and publishing community has discussed more structural approaches that remain at early, conceptual stages. One frequently raised idea is that AI companies and researchers wanting web data could draw from shared, rate-limited, policy-governed public corpora—such as the Common Crawl nonprofit archive, which maintains a continuously updated public web crawl freely available to researchers—rather than running their own or purchased scraping infrastructure. This would require AI companies to commit to shared norms around data sourcing; some already rely on Common Crawl, but a formalized legal and technical framework preferring it over direct scraping does not yet exist. A second class of proposals involves cryptographic mechanisms for distinguishing live human users from automated clients—where a device could generate a verifiable attestation of interactive use without revealing the user's identity or location—as a system-level complement to application-layer defenses. Both categories remain underdeveloped and would require coordination among AI companies, publishers, and platforms at a policy level that has not yet materialized.

Why it matters: The defenders' dilemma is asymmetric. A scraper operator needs to succeed once per IP per session—they need to fetch the page once from a million different IPs, which modern proxy networks enable trivially. A defender needs to correctly classify every single request, distinguish legitimate users from bots, and do so without creating friction that harms the user experience. As long as the underlying proxy infrastructure can generate millions of fresh residential IPs on demand, the defensive surface is effectively unbounded. Rate-limiting, IP blocking, and device fingerprinting all operate at a layer below the proxy infrastructure's capability to generate new addresses.

The Accountability Gap: Who Is Actually Paying for This?

One of the most important questions in the LWN analysis is one it cannot definitively answer: who are the end clients purchasing scraped data or paying for proxy access? The residential proxy operators themselves are not building AI models or deploying the scraping results. They are selling bandwidth and IP diversity to downstream customers, who may include AI data brokers, competitive intelligence firms, ad fraud operators, account takeover services, or nation-state threat actors.

The takedowns of IPIDEA and NetNut reveal that a single network serves all of these use cases simultaneously. GTIG observed state-sponsored threat groups from four countries (China, North Korea, Iran, Russia) using IPIDEA alongside what were presumably commercial scraping clients buying data in bulk. This co-mingling of use cases creates a legal and regulatory ambiguity that has served the industry well for years: a proxy operator can credibly claim it does not control what its clients do with the bandwidth it sells—the same argument historically made by bulletproof hosting providers and currently made by mainstream cloud providers when hosting illegal content.

The NetNut case may establish a useful legal precedent that operating a network whose primary documented uses include advertising fraud, account takeover, and mass unauthorized scraping constitutes participation in those harms, not merely facilitation of them. The distinction matters legally: a facilitator argues they are neutral infrastructure; a participant is liable for the harms committed through that infrastructure. Alarum Technologies' stock collapse reflects the market's read on that legal exposure and on whether regulators and courts will accept the "neutral platform" defense.

The participation of the IRS Criminal Investigation division in the NetNut seizure is a notable signal that deserves attention. IRS-CI typically investigates financial crimes—wire fraud, money laundering, proceeds of crime flows, tax evasion. The agency's involvement suggests investigators found financial crime theories—wire fraud, money laundering, or proceeds-of-crime flows—distinct from the Computer Fraud and Abuse Act theories typically applied to botnet operators. That broader legal theory, if it holds in subsequent litigation and appeals, has implications for the entire ecosystem of companies selling bandwidth sourced from non-consenting device owners: it suggests they can be held liable not just for operating malware, but for receiving proceeds of fraud.

Key Takeaways and Current State

• Two of the largest residential proxy networks ever documented—IPIDEA and NetNut—were disrupted within six months of each other, in January and July 2026 respectively, through coordinated action by Google GTIG, the FBI, IRS Criminal Investigation, and industry partners. These represent the most significant enforcement actions against proxy infrastructure ever undertaken.
• NetNut's Popa botnet comprised at least 2 million consumer devices, with 316 distinct threat clusters observed using its exit nodes in a single week in June 2026; IPIDEA had over 550 distinct threat clusters in a comparable period and controlled 13 separate proxy brands as reseller fronts. Both networks served commercial and state-sponsored clients simultaneously.
• The ecosystem is structurally resilient to single-network disruptions: after IPIDEA's takedown in January, NetNut absorbed much of the displaced traffic within months and achieved price-parity with IPIDEA before being seized itself. Proxy operators can rebuild by reselling competitors' networks, and new entrants can enter the market by building reseller fronts that obscure the upstream operator's identity.
• Individual publishers face asymmetric defensive conditions: Anubis, CAPTCHAs, iocaine, login gates, and server-side optimization all have real trade-offs in terms of latency, accessibility, and effectiveness. LWN's heaviest scraper attack ever occurred in the same period as these major enforcement actions, suggesting that remaining operators are intensifying campaigns.
• The app store SDK vetting gap remains unaddressed: platform vendors forbid unauthorized residential proxy apps in policy but have relied on reactive enforcement through Play Protect scanning and user reports. Proactive scanning for known proxy SDKs before app publication could prevent millions of device enrollments, but platform vendors have not implemented this at scale.
• Financial and legal exposure may be broadening for publicly traded entities: IRS-CI involvement in the NetNut seizure signals a money-flows theory of liability (wire fraud, money laundering) distinct from CFAA violations. Alarum Technologies' ~67% stock drop reflects market recognition of that expanded liability risk. This may discourage public companies from entering the residential proxy business.
• System-level proposals—shared corpora such as Common Crawl, cryptographic interactive attestation—remain underdeveloped but represent the only class of solution that addresses the structural incentive problem rather than individual attack vectors. These would require coordination among AI companies, publishers, and platforms at a policy level that has not yet materialized.
• No evidence implicates frontier AI companies in direct use of criminal residential proxy networks, but data brokers and intermediaries operating below the headline layer clearly monetize scraped data and sell it to model builders. Regulatory attention should focus on this intermediary layer and on supply-side controls rather than prosecuting individual scraping consumers.

What Comes Next: Emerging Enforcement Patterns and Regulatory Implications

Google has signaled that its enforcement posture is not a one-time effort, but rather the beginning of a sustained campaign. The language in the NetNut announcement explicitly frames it as a continued disruption effort, and public statements from GTIG indicate further actions against other providers in the interconnected residential proxy ecosystem are in preparation. The Play Protect integration, which now provides ongoing protection against future NetNut SDK installations and can be updated in real time without waiting for app store review processes, is a durable infrastructure change that outlasts any single domain seizure or court order.

The path forward involves escalation on multiple fronts:

• Platform enforcement: Apple, Google, and device OEMs could implement pre-publication SDK scanning to detect known proxy SDKs before apps reach users. This would close the current gap where apps are published, gain millions of installations, and are only removed weeks or months later.
• Financial chilling: If the IRS-CI precedent of wire fraud and money laundering liability holds in subsequent prosecutions and civil cases, venture capital and investors will become more cautious about funding residential proxy companies, even quasi-legitimate ones. Alarum Technologies' stock collapse may deter other public companies from entering the space.
• ISP and telecom cooperation: Residential proxy networks rely on residential ISP infrastructure. ISPs could implement filters, behavioral analysis, or device-side monitoring to detect and remediate proxy malware at scale. This is technically feasible but requires coordination with device manufacturers and changes to consumer terms of service.
• Regulatory clarification: The Federal Trade Commission (FTC) has authority under Section 5 of the FTC Act to regulate "unfair" practices that harm consumers. Using consumer devices without consent to host proxy servers and make them accessible to criminals could potentially be regulated as an unfair practice—though no formal FTC proceeding on residential proxies has been announced to date. Similar frameworks exist under GDPR and other jurisdictions.
• Civil liability expansion: Publishers harmed by scraping could pursue claims not just against proxy operators (who often have limited assets), but against data brokers and intermediaries who purchase and use scraped data. This would create liability further up the supply chain.

Whether the surviving quasi-legitimate players like Bright Data face escalating scrutiny—for advertising bypass of access controls and for their SDK-embedding business models—may be the defining regulatory question of the next twelve months. For independent publishers like LWN, the honest assessment is that the structural problem is being acknowledged at the right levels of government and industry for the first time, enforcement tools are improving, and the problem is still getting worse. That is genuine progress. It is not yet a solution.