Behind the Proxy: How Real Attacks Use Residential Proxy Networks
Dave Shackleford11.17.202618 minute read
Editor’s Note: This guest post was written by Dave Shackleford, owner and principal consultant of Voodoo Security, senior instructor with SANS, and faculty member at IANS Research.
— A login from a home broadband connection might look perfectly normal. But what if that connection is actually carrying traffic for an attacker who has never been anywhere near the house?
That's the problem with residential proxy networks. Attackers can route malicious activity through consumer IP addresses, making the originating connection appear more trustworthy than it really is.
And these aren't hypothetical scenarios. Microsoft documented residential proxy use in Midnight Blizzard's intrusion and Quad7's password-spraying operations. Google has investigated the commercial infrastructure supplying these networks, while the Justice Department has tied similar services to financial fraud.
I've discussed the limitations of IP reputation-based detection before. These investigations show what those limitations look like in practice, and why defenders need to understand what's happening behind the IP address.
Residential Proxies Have Become a Commercial Attack Resource
Residential proxies route traffic through IP addresses assigned to homes and small businesses. Some services use legitimate bandwidth-sharing arrangements. Others enroll devices through poorly disclosed software, compromised routers, or backdoored consumer equipment.
The distinction matters. A residential proxy exit might belong to an informed participant, or to somebody who has no idea their device is carrying another person's traffic.
The commercial ecosystem is what makes this particularly interesting from a security perspective.
Attackers don't necessarily need to compromise and manage thousands of devices themselves. Proxy operators and resellers can provide access to large pools of residential connections, often letting customers select geographic locations and rotate between available exits.
In that model, acquiring anonymized infrastructure becomes a purchasing decision rather than a major technical undertaking.
This creates a key consequence for defenders: different threat actors may use the same underlying proxy network, while a single actor may move between several providers.
That makes infrastructure attribution more complicated than identifying the IP address involved in a particular event.
The investigations that follow show how this plays out.
Quad7 and Storm-0940: Password Spraying Without the Noise
Microsoft's October 2024 investigation into CovertNetwork-1658, also known as Quad7, revealed how attackers used compromised small-office and home-office (SOHO) routers to conduct distributed password spraying.
The network consisted primarily of compromised TP-Link routers. Attackers installed backdoors and SOCKS5 proxy software, allowing authentication attempts to originate from the infected equipment.
Microsoft historically observed an average of approximately 8,000 compromised devices actively participating in the network at any given time. About 20% were conducting password spraying.
The attacks weren't particularly noisy at the individual account level. In about 80% of observed cases, CovertNetwork-1658 made only one sign-in attempt per account per day.
Consider what that means for a typical authentication detection rule.
A rule watching for repeated failures from one IP address may see very little. A rule focused on repeated attempts against a single account may encounter the same limitation.
Meanwhile, the campaign continues across many targets and sources.
Microsoft also observed Chinese threat actor Storm-0940 using credentials obtained through CovertNetwork-1658, sometimes on the same day they were compromised.
In several instances, successful access enabled further intrusion activity, including reconnaissance, lateral movement, and attempts to establish persistence.
The proxy network did more than hide traffic sources. It formed part of a broader operation that separated credential acquisition from follow-on compromise.
The detection lesson: Evaluate authentication attempts across accounts, networks, and time rather than relying exclusively on per-IP thresholds. Low-volume activity can be significant when it forms part of a coordinated pattern.
Midnight Blizzard: Residential Proxies Throughout an Identity Compromise
Midnight Blizzard, the Russian state-linked threat actor also known as APT29, demonstrated a related problem during its compromise of Microsoft's corporate environment, disclosed in early 2024.
The attackers initially gained access through password spraying against a legacy, nonproduction test tenant account that did not have multifactor authentication (MFA) enabled.
That was the opening.
From there, the actor compromised a legacy OAuth application with elevated access, created malicious applications, and exploited application permissions to access corporate email accounts.
Residential proxy infrastructure helped conceal activity during both the initial compromise and subsequent access to Exchange Online.
Microsoft specifically noted that the high turnover of residential proxy IP addresses made traditional indicator-of-compromise tracking insufficient.
Blocking an individual address might have interrupted one connection. It wouldn't have addressed the compromised identity, excessive application permissions, or access established through malicious OAuth applications.
There's a broader lesson here about the assumptions organizations make after authentication succeeds.
Once an attacker has valid credentials or application tokens, subsequent activity may appear legitimate unless the organization is examining what those identities and applications are actually doing.
The network connection is only one part of that picture.
The detection lesson: Eliminate legacy authentication exceptions, audit OAuth application privileges and consent grants, and monitor unusual identity and application behavior after successful authentication.
In July 2026, Google Threat Intelligence Group, working with the FBI, Lumen, and other partners, announced action against NetNut, also known as Popa.
Google estimated that the residential proxy network included at least two million devices worldwide, including consumer equipment such as smart TVs and streaming boxes.
During one week in June 2026, Google observed 316 distinct threat clusters using suspected NetNut exit nodes, including cybercriminal and espionage groups.
That's an important finding. While it does not establish that all 316 clusters successfully compromised a victim, it shows how widely threat actors relied on infrastructure associated with a single commercial proxy network.
Google also described NetNut's reseller ecosystem, in which other providers could market access to its infrastructure under their own brands.
Disrupting one service may degrade its operations, but customers and resellers can potentially move to other providers.
The available pool of proxy exits can change without eliminating the demand for access.
Google's NetNut investigation illustrates why tracking an individual service name or a fixed set of exit IPs provides only part of the picture.
XCPE: When the Backdoor Is Already on the Device
In November 2026, Google and industry partners reported disrupting XCPE, a firmware-level backdoor that had enrolled more than 1.3 million Android-based set-top boxes into a botnet.
Google assessed with high confidence that the backdoor had been intentionally developed and operated for financial gain by device manufacturer Videostrong Technology.
The operator monetized compromised devices by deploying additional malware modules from affiliated organizations.
Those modules supported advertising fraud, residential proxy enrollment, and further malware deployment. Google identified NetNut among the proxy-related components associated with the ecosystem.
The device owner might see a functioning streaming box while software running in the background provides access to the household's internet connection.
That's a difficult problem for consumers, and it has implications for organizations operating unmanaged equipment.
These devices may have limited security controls, inconsistent firmware updates, and little practical telemetry available to the people responsible for them.
Google's XCPE investigation describes the firmware backdoor, affiliate relationships, and monetization mechanisms in detail.
While NetNut and XCPE involved distinct investigations, both exposed connected parts of the same commercial ecosystem: compromised devices, proxy enrollment, resale, and downstream abuse.
The infrastructure isn't necessarily built for one attacker or one campaign. It can become a reusable resource for many.
According to the DOJ, the service had offered access to approximately 369,000 different IP addresses since 2020.
As of February 2026, its application listed approximately 8,000 infected routers available for customers to access.
Those figures describe different things: addresses offered over the service's history and infected routers listed at a particular point in time.
Criminal customers purchased access to these connections to conceal their originating IP addresses and locations while conducting fraud.
The DOJ documented activity involving bank and cryptocurrency account takeovers and fraudulent unemployment claims. Identified victims included a cryptocurrency customer who lost $1 million and a manufacturing company that lost $700,000.
The proxy network didn't necessarily provide the credentials used in every fraud. Its role was to make criminal activity appear to originate from ordinary residential connections.
For fraud teams, that's the problem.
An account takeover attempt could come from the same geographic region as the legitimate customer. A fraudulent transaction could originate from a residential address with no obvious history of abuse.
The connection may look familiar even when the activity is not.
What These Attacks Reveal About Traditional Detection
Although these incidents had different objectives, they expose recurring weaknesses in how organizations evaluate network activity.
Quad7 demonstrates how distributing authentication attempts can defeat thresholds based on individual accounts or IP addresses.
Midnight Blizzard shows how residential proxies can conceal multiple stages of an identity compromise, particularly when attackers exploit legacy accounts and application permissions.
NetNut and XCPE reveal the commercial supply chain behind these networks, while SocksEscort demonstrates their use in financial fraud.
Across the cases, the central issue is the same: the IP address a defender observes may identify the infrastructure carrying an attack without revealing the person responsible.
That doesn't make IP reputation, geolocation, or ASN information useless. It changes how much confidence we should place in those signals on their own.
Watch the Webinar: Behind the Proxy: Real Attacks, Hidden Adversaries
How are attackers using residential proxy networks to hide their activity? In this on-demand session, Dave Shackleford examines real-world attacks, the infrastructure behind them, and what security teams should look for when traditional IP-based detection falls short.
There isn't a single detection rule that solves this problem. The cases point toward several practical priorities.
Strengthen authentication and eliminate legacy exceptions.
Phishing-resistant MFA should be the standard wherever supported, particularly for privileged identities. Older accounts, test tenants, and application permissions deserve the same scrutiny as production environments. Exceptions can become attractive entry points.
Look for distributed patterns.
Correlate authentication failures across accounts, source IPs, network providers, and time. One failed attempt may be unremarkable. Thousands of accounts receiving similar attempts across a rotating pool of sources tell a different story.
Understand the infrastructure behind the connection.
Current IP intelligence can identify known residential proxy exits, associated services, and relevant network attributes. That context can help analysts distinguish potentially coordinated activity from ordinary residential access.
Monitor what happens after login.
Valid credentials don't establish that subsequent behavior is authorized. New OAuth grants, unusual mailbox access, unfamiliar resources, and unexpected account changes can reveal compromises that authentication controls alone missed.
Include transaction and device context.
For fraud teams, a suspicious account change or payment may be more meaningful when considered alongside proxy involvement, device information, and previous customer behavior. For security teams, vulnerable edge equipment and unmanaged devices deserve attention because they can become part of the proxy infrastructure itself.
Rather than labeling every residential connection as malicious, security teams should stop treating residential network ownership as proof that a connection is trustworthy.
Five Questions to Ask After an Anonymized Attack
One useful exercise after an incident is to identify where anonymization changed what the security team could see.
I recommend asking five questions:
How did the attacker get in? Identify the initial access mechanism, whether it involved compromised credentials, an exposed device, or another weakness.
Where did anonymization enter the attack chain? Determine whether proxy infrastructure supported reconnaissance, authentication attempts, post-compromise activity, or multiple stages.
What security assumptions were defeated? Did the organization trust a residential IP, familiar geography, low authentication volume, or successful login?
What evidence could have flagged the activity earlier? Examine identity logs, network intelligence, application permissions, device information, session activity, and transaction records.
What control changes would have altered the outcome? Identify improvements to authentication, correlation, monitoring, or enforcement that would have exposed or interrupted the activity.
Beyond simply expanding a blocklist with new indicators, the goal is to understand why an attack appeared legitimate and determine what evidence could have challenged that assumption.
The Address Is Only Part of the Story
Residential proxy networks have made ordinary internet connections useful infrastructure for sophisticated attackers and financially motivated criminals alike.
The IP address may be accurate. What it tells us about the person initiating the activity is another matter.
Security teams need to evaluate infrastructure alongside identity, device, session, and behavioral evidence. That includes recognizing when apparently unrelated events may be part of a coordinated attack.
Spur provides residential proxy detection, service attribution, and other explainable IP intelligence signals that security and fraud teams can incorporate into their existing investigations and decision workflows.
Because the next attacker may appear to be connecting from an ordinary residential broadband address, that alone shouldn't be enough to convince us they're legitimate.