The new DDoS threat landscape
Monetized, weaponized, inside your network
The DDoS threat landscape has changed
The DDoS threat landscape has changed. For more than two decades, attacks have mostly come from shady but bulletproof hosting companies. They were at a gigabit scale and telecommunications providers could address them with a mature, predictable playbook. Today, ISPs and hosting providers face a new generation of threats — and many of the largest now come from their own subscribers. More than 200 million compromised subscriber devices worldwide, funded at industrial scale by AI scraping demand, and sub-leased to criminal organizations and nation-state actors, generate terabits per second of malicious traffic from inside Tier-1 carrier networks. The result is degraded performance, poisoned IP reputation, and the collapse of legacy DDoS response timing — most attacks now end within five minutes.
Internal threats
Attacks no longer come from outside the network — they come from your subscribers, weaponized without their knowledge.
AI proxies
AI scraping demand is funding the underlying residential proxy market at a hundred-million-dollar scale, sustaining the compromised installed base.
Rapid attacks
Most attacks now peak in 1–3 minutes and end within 5 — outpacing legacy detect-and-divert response cycles.
How DDoS got monetized and weaponized
Attacks now come from inside the network
For 20+ years, DDoS attacks have come from outside the network, for example from Eastern European bulletproof hosting environments, using amplification or spoofing, and resulting in tens or hundreds of gigabits of malicious traffic. At some point in 2025, that pattern broke. Tens of thousands or hundreds of thousands of subscriber handsets and home gateways in major Tier-1 networks worldwide are now compromised. They generate multi-terabit attacks against targets – globally. If marshaled simultaneously, the aggregate firepower across affected carriers would exceed hundred Tbps levels — enough to disrupt internet connectivity for many countries and even continents. DDoS has become a country-scale concern.
AI is the demand-side funder
The hidden engine is the residential proxy market — middlemen who pay app developers to embed proxy SDKs, and device manufacturers who even ship malware in factory firmware. Up to 2024, this was a tens-of-millions-of-dollars niche market, selling residential IPs to fraudsters. Today it is a multi-hundred-million-dollar business, with one leading vendor publicly stating that 14 of the top 20 LLM companies are its customers. A US Tier-1 mobile IP now retails for about $95 for two weeks. AI training and inference demand has, in effect, paid for the world's largest compromised installed base.
The same devices are weaponized for crime and conflict
Twenty to thirty distinct criminal organizations and nation-state actors have stepped into the same installed base, running cybercrime operations and contributing to state-sponsored DDoS. The compromised handsets in major Western markets have been simultaneously monetized for AI scraping and weaponized for cyber operations.
DDoS at country scale
If every compromised device in Tier-1 US mobile and fixed networks were used simultaneously for DDoS, the aggregate firepower would exceed many hundreds of Tbps — enough to disrupt the entire US internet and affect global internet traffic. The DDoS threat has moved from a single-operator concern into national security territory.
of all attacks end within five minutes
of all attacks end within two minutes
of all attacks are below 50Gbps; modest individually, lethal collectively
of all attacks are multi-vector
of all attacks are multi-target
exploitable devices
These data points have been collected by the Nokia Deepfield Emergency Response Team (ERT) from samples obtained from our global deployments and from the Nokia Threat Intelligence Report 2025.
What's at stake for ISPs and hosting providers
When the threat comes from inside, the consequences hit the operator before they hit the target. Three layers of impact — operational, reputational, and customer-experience — combine to make this a P&L issue, not just a security issue.
RAN and backhaul absorb terabits of unsolicited outbound traffic
Compromised devices generate multi-terabit flows that sit on top of RAN, backhaul, and peering capacity — driving undiagnosed performance variation and silently consuming engineered headroom.
IP reputation is poisoned across the subscriber base
Tier-1 mobile CGNAT IPs now score 100 on third-party reputation engines. Subscribers face CAPTCHAs, payment-card declines, and content throttling — without knowing why or that their carrier is the cause.
Attacks end before legacy mitigation engages
78% of attacks now end within 5 minutes; 37% within 2. Time-to-peak is 1–3 minutes. Detect-and-divert response cycles inherited from a 100 Gbps world cannot meet that clock.
A national-scale problem hiding in plain sight
If marshaled simultaneously, compromised devices across Tier-1 US carriers could generate aggregate firepower in the hundreds of terabits — enough to disrupt the US internet. This is a national-security issue, not just an operator issue.
Most attacks are small, but lethal in swarms
82% of attacks are below 50 Gbps; 52% are carpet-bombing patterns hitting multiple prefixes; 58% are multi-vector. Modest individually, devastating collectively, invisible to threshold-based systems.
Anatomy of a 2025–2026 DDoS attack
Four structural traits define today's attacks. Each is documented across more than 140 operator deployments worldwide and reflected in the Nokia Threat Intelligence Report 2025.
Multi-vector attacks
58% of attacks now combine two or more vectors — for example, a SYN flood paired with a DNS amplification and an HTTP layer-7 attack. Defenders cannot tune for a single attack signature, and single-vector mitigations leave entire flanks exposed.
Under-5-minute bursts
78% of attacks now end within 5 minutes, up from 44% in 2024. 37% end within 2 minutes, with time-to-peak typically 1–3 minutes. By the time legacy alert systems escalate to mitigation diversion, the attack is over and the damage is done.
Attacks from inside the network
The largest attacks now originate from compromised subscriber devices within Tier-1 networks. Outbound floods consume RAN, backhaul and peering capacity and poison the operator’s IP reputation before any external target is hit. Perimeter-only defenses never see them forming.
Carpet bombing
52% of attacks now spread across multiple destination prefixes simultaneously — hitting many hosts at once with traffic levels per host that stay below conventional triggers. The aggregate is devastating; the slice on any one prefix looks benign.
Want the stage-by-stage version?
Follow a botnet DDoS attack from device recruitment through command and control to launch — and see where the chain breaks.
Terms shaping the new threat landscape
What is a botnet?
A network of compromised devices — handsets, routers, IoT gear, streaming boxes — remotely controlled by an attacker via command-and-control (C2) servers. Aisuru, Kimwolf, and Eleven11bot each span hundreds of thousands to millions of devices.
What is a residential proxy?
Residential proxies are IP addresses sold to look like ordinary household subscribers — not data-center IPs. Delivered via Android SDKs and Chinese factory firmware, they are bought by AI scrapers and exploited by attackers using the same devices.
Inbound vs outbound DDoS
Inbound DDoS arrives from outside the network, aimed at customers. Outbound DDoS originates inside the network from compromised subscribers, aimed at external targets — degrading the operator's RAN, peering, and reputation along the way.
Frequently asked questions
Terabit-scale attacks are now routine. Nokia’s Threat Intelligence Report 2025 found daily DDoS peaks of 5–10 Tbps, and a 33 Tbps attack against a gaming provider was observed in October 2025 — with terabit-scale attacks occurring roughly five times more often than a year earlier. At the same time, sheer size is only half the story: 82% of DDoS floods stay below 50 Gbps, relying on speed, repetition and precision rather than raw volume.
Most are over before a human can respond: in 2025, 78% of DDoS attacks ended within five minutes (up from 44% in 2024), and 37% lasted less than two minutes, with attacks often reaching peak intensity in one to three minutes. Campaigns may also recur in bursts over hours or days. Short does not mean harmless — fast, repeated hits interrupt payments, logins and communications, and they defeat any defense that needs minutes to divert traffic.
Botnets weaponize the enormous base of insecure IoT devices — IP cameras, DVRs, home and business gateways — now connected over gigabit and multi-gigabit broadband. Devices attack using real IP addresses, from inside and outside service provider networks, and once enrolled, they typically remain in the botnet permanently, ready to be triggered without warning. The largest botnet families have spanned hundreds of thousands to millions of devices, and their traffic can resemble normal consumer activity — defeating threshold- and reputation-based defenses.
Residential proxies route attack traffic through ordinary household connections, using constantly rotating consumer IP addresses, making malicious requests appear to be from genuine subscribers and defeating IP blocklists, geo-blocking, and rate limiting. Nokia identified more than 100 million residential endpoints as potentially exploitable and observed active residential-proxy endpoints associated with DDoS grow from about 1 million to 8–9 million in one year — with the United States accounting for 75% of both attack volume and compromised endpoints in 2025. Symmetric fiber uplinks and industrial-scale AI web scraping keep the proxy economy growing.
Attackers use automation and AI to orchestrate campaigns that shape-shift within seconds — rotating vectors from TCP carpet bombing to UDP floods to DNS attacks — while monitoring defenders’ thresholds, alert cadence and tactics and adapting in real time. In one observed 2025 campaign, four distinct attacks were launched within three minutes, each adjusted in response to the defensive response. Evidence of fully autonomous AI-driven attacks remains limited, but algorithmic orchestration is already the norm in sophisticated campaigns.
A DDoS tsunami is a hyper-volumetric flood — a massive wave of attack traffic, now reaching multiple terabits per second, that aims to overwhelm links and infrastructure through sheer scale. Nokia contrasts tsunamis with “piranha” attacks: swarms of small, short, precise strikes that stay under volumetric radar. Defending against both requires two kinds of intelligence — network-scale visibility for the tsunamis and behavioral, context-aware detection for the piranhas.
Carpet bombing spreads attack traffic across many addresses or prefixes — sometimes an operator’s entire address block — rather than concentrating on a single host, keeping per-destination volumes below detection thresholds while the aggregate load still saturates the infrastructure. It is one facet of the broader multi-target trend: in 2025, 52% of DDoS attacks targeted multiple systems simultaneously.
Inbound DDoS is malicious DDoS traffic arriving from outside a network toward targets inside it — the classic case. Outbound DDoS attacks originate from compromised devices within the network, targeting other customers within the network or external targets. When it goes outside the network, it consumes the operator’s own upstream capacity, degrades service for legitimate subscribers, damages the network’s IP reputation, and can create regulatory and legal exposure. Because botnet and residential-proxy endpoints live inside subscriber networks, modern operators must detect and mitigate in both directions.
Nokia Deepfield: Next-gen DDoS security
Discover Nokia's cutting-edge DDoS security, featuring AI-powered Deepfield Defender and Secure Genome for robust protection against evolving cyber threats.
DDoS security
Defend your network against botnets and application-layer attacks with an AI-driven DDoS security for fast and accurate detection and real-time mitigation
Deepfield Defender
Next-gen, big data and AI-driven DDoS detection and mitigation solution.
Deepfield Genome Shield
Deepfield Genome Shield delivers proactive, always-on, network-wide DDoS protection for the AI era.
Learn more
Blog
Article
Blog
Blog
Blog
Blog
Blog
Blog
Contact us
Please complete the form below.
The form is loading, please wait...
Thank you. We have received your inquiry. Please continue browsing.