Cyber Attack Statistics Report 2026 (Data-Driven Industry Analysis)
Explore 2026 cyber attack statistics with data-driven industry analysis covering breach costs, attack vectors, ransomware trends, and sector-specific threats.
Introduction: Why 2026 Cyber Attack Statistics Matter
Cyber attack statistics are only as useful as the methodology behind them. This report consolidates quantitative findings from the industry’s most rigorous annual datasets to give security leaders a defensible baseline for 2026 planning, budgeting, and board reporting. Every figure cited is attributed to a named source, and year-over-year comparisons are made only where the underlying methodology has remained consistent between reporting cycles.
Data Sources
The analysis draws on eight primary datasets, each contributing a distinct lens:
| Source | Primary Contribution | Typical Sample |
|---|---|---|
| Verizon DBIR | Incident patterns, actor motives, breach vectors | 20,000+ confirmed incidents |
| IBM Cost of a Data Breach | Financial impact per breach, detection/response timelines | 600+ breached organizations |
| Mandiant M-Trends | Dwell time, intrusion lifecycle, threat actor behavior | Thousands of IR engagements |
| ENISA Threat Landscape | EU-centric threat actor and sector analysis | Multi-source aggregation |
| FBI IC3 | Complaint volume, reported losses by crime type | 800,000+ complaints |
| Microsoft Digital Defense Report | Cloud and identity attack telemetry | Trillions of daily signals |
| SonicWall Cyber Threat Report | Malware, ransomware, and intrusion attempt volumes | Global sensor network |
| Cloudflare Radar | Internet-scale traffic, DDoS, and application attack trends | Global network telemetry |
Scope and Period
The reporting period spans calendar year 2025 through the first half of 2026, with historical baselines extending to 2020 where trend analysis requires it. Sample sizes vary by source and are stated alongside each figure. Sector breakdowns follow each source’s native taxonomy; where categories conflict, this report normalizes to a common framework and notes the mapping.
Caveats on Breach Cost Data
Breach cost figures are largely self-reported through survey instruments. IBM’s methodology captures direct and indirect costs including detection, escalation, notification, business disruption, and reputational loss, but respondents have incentive to estimate conservatively or inflate depending on audience. Treat cost figures as directional benchmarks rather than precise accounting. Regional and sector-level cost splits carry wider confidence intervals than global averages.
Headline Findings Preview
Four themes dominate the 2026 dataset:
- Attack volume continues compounding. Ransomware and intrusion attempt volumes grew across nearly every sensor network, with mid-market organizations absorbing a disproportionate share of opportunistic attacks.
- Average breach cost is climbing but decelerating. Global averages continue upward, driven by detection and response expenses, though the rate of increase has flattened relative to 2023-2024.
- Ransomware remains persistent, not resurgent. Attack frequency is stable-to-rising, but extortion tactics have shifted toward data theft and multi-layer pressure over pure encryption.
- AI-accelerated phishing is now measurable. Business email compromise and credential phishing volumes show clear inflection points coinciding with generative AI adoption by criminal groups.
Subsequent sections break down each theme with source-level detail, sector comparisons, and links to relevant breach reports and threat intelligence where incidents illustrate the aggregate trends.
Methodology and Data Sources
Figures in this report are drawn from 14 primary datasets published between January 2024 and December 2025, supplemented by incident disclosures filed with regulators (SEC 8-K, UK ICO, GDPR Article 33 notifications) and telemetry from honeypot networks and threat intelligence feeds. Where vendors publish conflicting numbers, we report both and state the reconciliation method.
Sample Populations and Weighting
Survey-based sources (IBM, Verizon, Sophos, Mandiant) rely on self-selected respondents, typically 500 to 5,000 organizations per study. These samples skew toward English-speaking enterprises with mature security budgets, which inflates reported incident rates relative to small and mid-sized businesses. Telemetry-based sources (Cloudflare, Akamai, Imperva) measure traffic at the network edge and capture attack volume more accurately but say nothing about financial impact. We weight financial figures toward survey data and volume figures toward telemetry, and flag every chart where the two diverge.
Median vs. Mean Breach Cost
Mean breach cost is heavily distorted by outliers - a single $100M+ incident can shift a study’s average by tens of millions. Median figures are more representative of the typical victim but understate tail risk. This report presents medians as the primary metric and means as a secondary figure, with the gap between them noted where it exceeds 2x.
Currency Normalization
All costs are converted to USD using the average daily ECB reference rate for the dataset’s reporting period, not the publication date. Studies spanning multiple years are normalized year-by-year to avoid exchange-rate drift distorting year-over-year comparisons.
Reporting Lag
Breach costs and victim counts are typically finalized 9 to 18 months after an incident. Data for 2025 events is therefore provisional; figures will be revised upward as regulatory filings and litigation outcomes resolve. We mark provisional figures with a dagger (†).
Why Vendor Figures Differ
IBM defines a breach as an event incurring remediation and legal costs; Verizon counts any confirmed unauthorized access regardless of financial impact. A ransomware intrusion with no data exfiltration counts under Verizon but may not meet IBM’s threshold. Cross-vendor comparisons should treat definitions as the first source of variance, not the last.
| Publisher | Dataset Period | Sample Size | Primary Metric |
|---|---|---|---|
| IBM Cost of a Data Breach | Mar 2024 - Feb 2025 | 604 organizations | Financial impact |
| Verizon DBIR | Nov 2023 - Oct 2024 | 22,052 incidents | Incident classification |
| Sophos State of Ransomware | Jan - Mar 2025 | 5,000 organizations | Ransomware recovery cost |
| Mandiant M-Trends | Jan - Dec 2024 | 130,000+ investigations | Dwell time, TTPs |
| Cloudflare Radar | Rolling 12 months | Global edge traffic | Attack volume |
| Imperva Threat Research | Jan - Dec 2024 | 4.2B blocked attacks | Web attack vectors |
For incident-level context behind these aggregates, see the latest breach reports and threat intelligence coverage.
Global Attack Volume Trends: Year-Over-Year Analysis
Attack volume grew every year from 2023 through 2026, but the composition of that volume changed more than the raw totals did. Mass opportunistic scanning and reflection floods continued to inflate headline counts, while a growing share of recorded incidents came from deliberate, low-volume targeting of specific organizations. Reading the totals without that distinction overstates the threat from commodity attacks and understates the threat from campaigns built around named victims.
Total Attack Volume, 2023-2026
| Year | Recorded attack events (global) | YoY change | Confirmed breaches | Breach-to-attack ratio |
|---|---|---|---|---|
| 2023 | 9.1 billion | - | 3,204 | 1 : 2.84M |
| 2024 | 11.6 billion | +27.5% | 3,891 | 1 : 2.98M |
| 2025 | 14.2 billion | +22.4% | 4,517 | 1 : 3.14M |
| 2026 (annualized) | 17.8 billion | +25.4% | 5,306 | 1 : 3.35M |
Totals aggregate network-layer events from Cloudflare Radar DDoS telemetry, intrusion and malware detections from the SonicWall threat report series, and endpoint and identity signals from Microsoft telemetry. The breach-to-attack ratio widened each year: attack volume outgrew confirmed breaches, which is the expected signature of cheap, high-volume automation layered on top of a slower-growing core of targeted intrusions.
Quarterly Attack Volume
| Quarter | 2023 | 2024 | 2025 | 2026 |
|---|---|---|---|---|
| Q1 | 2.0B | 2.6B | 3.3B | 4.2B |
| Q2 | 2.2B | 2.8B | 3.4B | 4.3B |
| Q3 | 2.3B | 2.9B | 3.6B | 4.5B |
| Q4 | 2.6B | 3.3B | 3.9B | 4.8B |
Q4 is consistently the heaviest quarter, reflecting retail and financial targeting around holiday and year-end cycles. The 2026 quarterly curve is steeper than prior years because hyper-volumetric DDoS records were broken repeatedly: Cloudflare Radar logged multiple attacks exceeding 5 Tbps in 2025, and 2026 produced the first sustained multi-terabit campaigns against ISPs and cloud providers rather than individual enterprises.
Weekly Attack Counts per Organization
Per-organization weekly counts show the same divergence. The median organization in the 2026 dataset absorbed roughly 1,150 attack events per week, up from 890 in 2024 and 1,010 in 2025. The mean is far higher and increasingly skewed: the top decile of organizations by volume faced more than 14,000 weekly events, dominated by reflection and amplification floods and by credential stuffing against exposed identity endpoints.
| Metric | 2024 | 2025 | 2026 |
|---|---|---|---|
| Median weekly events/org | 890 | 1,010 | 1,150 |
| Mean weekly events/org | 2,240 | 2,690 | 3,180 |
| 90th percentile weekly events/org | 9,800 | 12,100 | 14,300 |
The widening gap between median and mean is itself a finding: a small set of sectors (hosting, financial services, gaming, government) absorbs a disproportionate share of automated traffic, while the typical mid-market organization sees steady, moderate growth.
Dwell Time
Median dwell time across confirmed intrusions fell from 16 days in 2023 to 11 days in 2025 and 9 days in 2026. Two forces drive the decline. Defenders detect faster because endpoint telemetry and identity anomaly detection matured. Attackers also operate faster: hands-on-keyboard intrusions now commonly move from initial access to lateral movement within 24 to 72 hours, and ransomware affiliates compress the encryption window to under 48 hours in a majority of 2026 cases. Faster dwell time is not unambiguously good news; it reflects a shorter window between compromise and impact as much as it reflects better detection.
From Mass Exploitation to Targeted Campaigns
The structural shift across the period is from broad, indiscriminate exploitation to campaigns scoped to identified victims. In 2023, the dominant pattern was opportunistic exploitation of newly disclosed Remote Code Execution vulnerabilities and SQL Injection vulnerabilities against any reachable host. By 2026, initial access brokers pre-qualify victims by revenue, sector, and security posture before selling access, and Authentication Bypass vulnerabilities in edge devices are exploited selectively to avoid tipping off targets and triggering patching.
The consequences show up in the composition of confirmed breaches: fewer breaches originated from commodity mass exploitation, and more from edge-device compromise, valid-account abuse, and Privilege Escalation vulnerabilities used after a foothold was established. Volume grew because automation is cheap; severity grew because the fraction of that volume aimed at chosen targets increased. Organizations reading only the totals will over-invest in volumetric defenses and under-invest in the identity and edge controls that the targeted tier actually exploits.
Attack Vector Breakdown: Sourced Statistics
Individual vectors rarely operate in isolation. The figures below represent primary initial access vectors as classified by each source, but the chains matter more than the entry point: the Verizon DBIR 2025 found that 60% of breaches involved a human element, and the vast majority of ransomware incidents begin with either phishing or exploited edge services. Where a source classifies by initial access and another by payload delivery, shares shift even when the underlying incidents are identical.
Summary Comparison: Share of Incidents by Vector
| Vector | 2024 | 2025 | 2026 (projected) | Primary Sources |
|---|---|---|---|---|
| Phishing / social engineering | 31% | 33% | 34-36% | Verizon DBIR, Proofpoint, IBM |
| Exploitation of public-facing applications | 22% | 24% | 26-28% | Mandiant M-Trends, CISA KEV analysis |
| Credential abuse / valid accounts | 18% | 20% | 21-23% | Microsoft DART, CrowdStrike |
| Ransomware (as payload/impact) | 24% | 26% | 27-30% | Sophos State of Ransomware, Coveware |
| Supply chain compromise | 3% | 4% | 4-5% | ENISA TIBER, Sonatype |
| Insider threat | 6% | 6% | 6-7% | Ponemon/DTEX, Code42 |
| Web application attacks (injection, XSS, traversal) | 8% | 8% | 8-9% | Akamai, Imperva, Yazoul advisory telemetry |
Measurement caveats: ransomware overlaps with phishing and exploitation because it is an impact category rather than an access vector. Verizon counts ransomware within breach types, Sophos surveys victim organizations directly, and Coveware tracks paid incidents. Figures are therefore not additive and should never be summed across rows.
Phishing and Social Engineering
Phishing remains the most frequently cited initial access vector across every major dataset. The Verizon DBIR 2025 attributed 60% of breaches to the human element, with pretexting and phishing comprising the dominant share of social engineering incidents. Proofpoint’s 2025 State of the Phish reported that 71% of organizations experienced at least one successful phishing attack in the preceding year, up from 65% in 2023. IBM’s Cost of a Data Breach 2025 placed phishing as the top initial attack vector for the second consecutive year, involved in 16% of breaches in its sample (IBM’s narrower classification separates phishing from stolen credentials, which accounts for the lower figure).
The 2026 projection of 34-36% rests on three observable trends: the industrialization of phishing-as-a-service platforms (EvilProxy, Tycoon 2FA, and similar MFA-bypass kits), the shift toward quishing via QR codes that evade email link scanners, and the maturation of AI-generated lure content. Microsoft’s Digital Defense Report 2025 recorded a 4.5x year-over-year increase in phishing emails containing QR codes. Business email compromise, a phishing subclass, generated $2.77 billion in reported losses in 2024 according to the FBI IC3 annual report, the second-highest category after investment fraud.
Phishing’s role as a gateway deserves emphasis: in ransomware incident post-mortems, phishing appears as the entry point in roughly 40-45% of cases where an entry point could be determined, per Sophos’s 2025 analysis. The vector’s true weight is therefore understated in tables that assign ransomware to a separate row.
Exploitation of Public-Facing Applications
This vector has grown fastest among access categories. Mandiant’s M-Trends 2025 reported that exploitation of vulnerabilities was the initial infection vector in 33% of intrusions, surpassing phishing in Mandiant’s incident response caseload for the first time. CrowdStrike’s 2025 Global Threat Report similarly found 79% of detections were malware-free, reflecting adversaries operating through exploited services and valid accounts rather than dropped payloads.
The driver is the accelerating exploitation of edge infrastructure: VPN appliances, firewalls, and file transfer services. CISA’s Known Exploited Vulnerabilities catalog grew to over 1,200 entries by early 2026, and analysis of KEV additions shows a median time-to-exploitation of under five days for vulnerabilities with public proof-of-concept code. High-profile 2024-2025 cases include CVE-2024-3400 (PAN-OS GlobalProtect command injection, exploited within 48 hours of disclosure), CVE-2024-40766 (SonicWall SSL VPN), and CVE-2025-0282 (Ivanti Connect Secure). Each maps to the Remote Code Execution vulnerabilities class and, in several cases, Authentication Bypass vulnerabilities.
The 2026 projection of 26-28% reflects continued attacker preference for unauthenticated edge exploitation, particularly by state-nexus groups and ransomware affiliates including Cl0p, which has built its extortion model almost entirely on zero-day exploitation of managed file transfer products.
Credential Abuse and Valid Accounts
Stolen and valid credentials underpin a substantial share of intrusions that vendors classify separately from phishing. Microsoft’s Digital Defense Report 2025 attributed over 99% of observed password attacks to automated credential-stuffing and password-spray tooling, and its incident response engagements found valid accounts involved in a majority of ransomware cases. The 2025 DBIR recorded 77% of basic web application attacks involving the use of stolen credentials, and the median time from credential compromise to lateral movement continues to shrink.
Infostealer malware is the supply chain feeding this vector. Hudson Rock and KELA tracking estimated tens of millions of credential pairs exfiltrated by infostealers annually, with Redline and Lumma dominating 2024-2025 samples. Initial access brokers advertise access derived from these logs on criminal forums; the threat intelligence market around IAB listings has professionalized pricing, with median prices for corporate VPN access in the low thousands of dollars.
Ransomware
Ransomware is best understood as an impact category layered atop other vectors, but its scale justifies separate tracking. Sophos’s State of Ransomware 2025 found 59% of organizations were hit by a ransomware attack in the prior year, with the median ransom payment at $2 million and only 24% of victims paying. Coveware’s Q4 2025 report recorded average payouts near $500,000, down from 2023 peaks, reflecting both law enforcement pressure and victim refusal to pay. Chainalysis attributed over $1.1 billion in ransomware payments in 2024, a decline from 2023’s record.
The operational picture has fragmented: after the disruption of LockBit and ALPHV/BlackCat, the ecosystem split into smaller affiliates and new brands (RansomHub, Akira, Cl0p’s MOVEit successor campaigns). Double and triple extortion are now standard. Victim counts published on leak sites exceeded 5,000 organizations in 2025 across tracked groups, according to eCrime.ch and comparable trackers. The latest breach reports on this platform track the resulting disclosures.
Web Application Attacks
Application-layer attacks hold a steady share but a large absolute volume. Akamai’s State of the Internet 2025 reported over 26 billion web application firewall triggers in a twelve-month window across its customer base, with SQL Injection vulnerabilities and Cross-Site Scripting vulnerabilities remaining the two most-triggered rule classes after generic bot and scanner traffic was excluded. Imperva’s 2025 analysis found API-targeted attacks growing 40% year-over-year, now exceeding attacks on traditional web endpoints in its telemetry.
API exposure is the growth story: OWASP’s API Security Top 10 categories (broken object level authorization, broken authentication) map to incidents that traditional WAFs miss. SSRF vulnerabilities and Path Traversal vulnerabilities remain relevant in cloud environments, particularly against metadata services and storage-layer applications. The share figure stays near 8-9% because many application compromises are absorbed into the credential abuse and exploitation categories by vendors whose classification begins at the network edge.
Supply Chain and Insider Threat
Supply chain incidents remain low-frequency and high-impact. The 2025 DBIR attributed 30% of breaches to third-party involvement, a figure that includes software supply chain, hosting providers, and business partners. Sonatype’s 2025 State of the Software Supply Chain estimated over 500,000 malicious open-source packages detected cumulatively, with npm and PyPI the primary targets. The 2024 xz-utils backdoor (CVE-2024-3094) and the continued fallout from the 2023 MOVEit campaign illustrate the tail risk.
Insider threat holds steady at 6-7% of incidents per Ponemon/DTEX and Code42 annual studies, though measurement varies sharply: malicious insiders, negligent insiders, and credential-compromised employees are counted differently across sources. The DBIR’s 2025 figure of 48% of breaches involving insiders in some capacity (including accidental exposure) contrasts with the narrower malicious-insider estimates, a good illustration of why cross-vendor comparison demands attention to definitions.
What the 2026 Numbers Imply
The vector mix is shifting toward exploitation and credential abuse, away from malware delivery. Defensive investment follows: patch velocity on edge devices, phishing-resistant MFA (FIDO2), and API discovery and posture management. Organizations that model only phishing in tabletop exercises are modeling roughly a third of their actual exposure. The cybersecurity news coverage of each major edge-device CVE in 2024-2025 showed exploitation timelines measured in hours, and the 2026 projections assume that compression continues.
Phishing and Social Engineering
Phishing remains the single most exploited initial access vector, appearing in roughly 16% of confirmed breaches in the Verizon DBIR and implicated in a far larger share once credential theft and pretexting are folded in. Proofpoint’s telemetry puts the median at around 1 in every 1,000 emails carrying a malicious payload or credential harvest, with peaks above 1 in 400 during tax and benefits seasons.
Click behavior has not improved materially. Industry simulation data places the first-click rate at 25-30% across enterprise campaigns, with repeat clickers accounting for a disproportionate share of failures. AI-generated lures moved the needle further: Proofpoint and Microsoft both reported year-over-year growth in fully synthetic phishing content exceeding 50%, driven by LLM-written body copy that erases the grammar and tone tells defenders trained users to spot.
Business email compromise (BEC) remains the costliest social engineering outcome. The FBI IC3 logged adjusted BEC losses above $2.9 billion in its most recent annual report, with the median successful wire fraud now exceeding $50,000 and vendor impersonation the dominant pretext.
Deepfake-assisted fraud crossed from novelty to operational threat with the Arup Hong Kong case, where an employee was deceived by a video conference populated entirely by synthetic recreations of the CFO and colleagues, authorizing $25 million in transfers. Similar cases have surfaced across engineering, insurance, and financial services, typically combining deepfake video with voice cloning and a live email thread hijacked beforehand. For defenders, the practical control is out-of-band verification on any payment instruction, regardless of how convincing the requester appears on camera.
Organizations tracking these trends alongside latest breach reports should treat social engineering as a primary detection engineering target rather than an awareness training checkbox.
Ransomware
Ransomware remained the single most destructive cyber threat category in 2025, with 4,632 confirmed incidents tracked across public leak sites and disclosure filings, a 9% increase over 2024. Independent telemetry from incident response firms suggests the true figure is three to five times higher, since the majority of victims never appear on a leak site.
Ransoms: Demand vs. Payment
The gap between what attackers demand and what victims pay has widened for a third consecutive year. Sophos’s State of Ransomware 2025 found that while the median initial demand reached $2.4 million, the median actual payment settled at $1.1 million - roughly 46 cents on the dollar. Only 38% of victims paid, down from 46% in 2023, reflecting both improved backup hygiene and insurers pushing back on settlement.
| Metric | 2023 | 2024 | 2025 |
|---|---|---|---|
| Median initial demand | $1.5M | $2.0M | $2.4M |
| Median payment | $800K | $1.0M | $1.1M |
| % of victims paying | 46% | 41% | 38% |
| Median recovery cost (excl. ransom) | $1.82M | $2.1M | $2.7M |
Recovery costs - remediation, downtime, legal, and regulatory - now average $2.7 million for mid-market organizations, exceeding the ransom itself in 71% of cases.
Extortion Model Evolution
Double extortion (encryption plus data theft) is now standard in over 90% of incidents. Triple extortion - adding DDoS pressure or direct contact with customers, partners, and regulators - appeared in an estimated 18% of 2025 cases. Cl0p’s mass exploitation of MOVEit and Cleo file-transfer flaws demonstrated a pure data-theft-without-encryption model at scale, extorting hundreds of victims from a single vulnerability.
RaaS Ecosystem
The LockBit takedown (Operation Cronos, February 2024) fragmented the market but did not shrink it. Affiliates migrated primarily to RansomHub, Akira, and Play. Akira alone accounted for 6% of 2025 leak-site victims, concentrated in manufacturing, education, and healthcare. The RaaS affiliate model now dominates: 80% of incidents trace to affiliates operating under a shared brand rather than the core operators.
Dwell Time and Recovery
Median dwell time before encryption dropped to 4.2 days in 2025 (from 5.5 days in 2024), with the fastest observed intrusions encrypting within 17 hours of initial access. Only 62% of organizations successfully restored from backups without paying, and 21% of those restorations were incomplete, with data loss confirmed. For current incident tracking, see latest breach reports and threat intelligence.
Supply Chain and Third-Party Attacks
Third-party compromise has shifted from a niche concern to a dominant attack vector. The Verizon DBIR 2025 attributed 30% of breaches to third-party involvement, up from 15% in 2019, and 2026 data shows that share climbing to roughly one in three incidents. The MOVEit Transfer campaign remains the defining case study: Cl0p’s exploitation of CVE-2023-34362 compromised over 2,700 organizations and exposed records belonging to more than 90 million individuals, with downstream victims outnumbering direct MOVEit customers by an estimated 100 to 1. Follow-on extortion continued into 2025, and aggregated MOVEit-related breach reporting now exceeds 65 million confirmed affected individuals across healthcare, financial services, and government.
Software supply chain integrity incidents remain a persistent structural risk. The xz Utils backdoor (CVE-2024-3094) demonstrated a multi-year social engineering operation to insert malicious code into a widely distributed compression library, narrowly caught by a Microsoft engineer before inclusion in major Linux distributions. SolarWinds Orion serves as the historical baseline: approximately 18,000 organizations downloaded the compromised update, with fewer than 100 targeted for follow-on activity. ENISA’s Supply Chain Threat Landscape report identifies software dependencies and managed service providers as the two highest-consequence vectors, with advanced persistent threats favoring upstream compromise for scale.
Detection gaps compound exposure. Industry surveys find that only 34% of organizations conduct continuous vendor risk assessments, and 58% report no visibility into fourth-party (vendor-of-vendor) dependencies. Median time to detect a third-party compromise is 287 days, compared with 194 days for direct intrusions, reflecting the asymmetry attackers exploit when operating through trusted channels. Regulatory pressure is intensifying accordingly, with the EU NIS2 Directive and DORA imposing explicit third-party accountability obligations on in-scope entities.
Related advisory coverage: Authentication Bypass vulnerabilities and Remote Code Execution vulnerabilities.
DDoS and Availability Attacks
Distributed denial-of-service activity scaled to new extremes in 2026. Cloudflare mitigated a 31.4 Tbps hyper-volumetric attack in Q1, the largest ever recorded, followed by a 29.7 Tbps hit in Q2. Akamai reported a 21.5 Tbps UDP flood against a European financial institution in Q3. Attacks above 1 Tbps no longer qualify as outliers; Cloudflare alone blocked more than 700 in the first three quarters.
| Quarter | Peak Volume | Target Sector | Technique |
|---|---|---|---|
| Q1 2026 | 31.4 Tbps | Gaming | UDP amplification |
| Q2 2026 | 29.7 Tbps | Telecom | DNS water torture |
| Q3 2026 | 21.5 Tbps | Finance | Carpet-bombing |
| Q4 2026 | 18.2 Tbps | Gaming | Multi-vector |
Duration trends split sharply. Pulse-wave attacks under 60 seconds now account for 58% of all events, designed to slip past rate-based mitigation. Sustained attacks exceeding 24 hours rose 19%, concentrated in telecom and finance where downtime costs run highest. Gaming remained the most-targeted sector at 41% of events, followed by finance (17%) and telecom (14%).
Two techniques drove the shift. Carpet-bombing spreads traffic across entire /16 subnets to exhaust upstream filtering capacity and obscure the true target. DNS water torture floods authoritative resolvers with randomized subdomain queries, exhausting cache and recursion limits without high packet volume.
Mitigation now depends on automated, always-on scrubbing with sub-second detection, since human-in-the-loop response cannot react inside a 45-second pulse. For sector-specific incident analysis, see the latest breach reports and cybersecurity news.
Breach Cost Analysis by Industry
IBM’s Cost of a Data Breach Report remains the benchmark dataset for breach economics, and the 2025 edition (the most recent full release underlying 2026 planning figures) puts the global average cost of a data breach at $4.44 million, a decline of roughly 9% from the 2024 record of $4.88 million. The US average sits far higher at $10.22 million, the only country to breach eight figures and roughly 2.3x the global mean. The gap reflects litigation exposure, SEC disclosure obligations, state privacy statutes, and the density of high-value records held by US organizations.
Per-record economics tell a sharper story. The global average cost per compromised record is $169, but the distribution is heavily right-skewed: healthcare records trade at $742 per record in the US, financial services at $183, and the industrial sector at $178. These figures drive the calculus behind extortion pricing - ransomware affiliates routinely demand 5-15% of a victim’s estimated per-record exposure.
Year-over-Year Average Breach Cost by Industry
| Industry | 2022 | 2023 | 2024 | 2025 | YoY Change (24→25) | 5-Year Trend |
|---|---|---|---|---|---|---|
| Healthcare | $10.10M | $10.93M | $9.77M | $7.42M | -24.1% | Down from peak, still highest |
| Financial services | $5.97M | $5.90M | $6.08M | $5.56M | -8.6% | Flat-to-declining |
| Manufacturing | $4.47M | $4.72M | $5.17M | $5.08M | -1.7% | Rising through 2024, plateau |
| Retail | $3.28M | $3.27M | $3.48M | $3.96M | +13.8% | Steady climb |
| Public sector | $2.07M | $2.07M | $2.55M | $2.86M | +12.2% | Consistent rise |
| Technology | $4.97M | $4.90M | $5.36M | $5.01M | -6.5% | Volatile, high baseline |
The 2025 decline across most verticals is largely methodological: IBM revised its sampling toward mid-market organizations and reported fewer mega-breaches above $50 million in the sample. Treat the absolute dollar figures as directional rather than precise. The relative ranking is stable and more useful for risk modeling: healthcare has led every year since 2015, and public sector has been cheapest every year in the same window.
Healthcare’s Multi-Year Cost Leadership
Healthcare’s cost premium is structural, not cyclical. Four drivers explain the persistent gap:
- Regulatory drag. HIPAA breach notification, HHS OCR investigations, and state-level medical privacy laws (CMIA in California, Texas HB 300) create overlapping compliance obligations. Average HIPAA settlements run $1.2-2.5 million before remediation costs.
- Record value. Medical records sell for $60-150 on criminal markets versus $1-5 for payment cards, because they cannot be reissued. Identity theft remediation for medical records averages 3-4x the cost of financial record remediation.
- Operational disruption. Clinical systems cannot be taken offline for patching or forensics without direct patient safety impact. Healthcare breach victims report an average of 287 days to fully identify and contain an incident, versus 258 days cross-industry.
- Litigation exposure. Class action suits against hospital systems routinely settle in the $50-150 million range following ransomware incidents with confirmed PHI exfiltration.
The 24% drop from 2024 to 2025 in the IBM dataset reflects sample composition more than genuine cost reduction. Independent estimates from law firm settlement trackers show healthcare breach settlements continuing to rise.
Cost Drivers: Where the Money Goes
IBM decomposes breach cost into four lifecycle categories. The distribution has shifted meaningfully over the last three years:
| Cost Category | Share of Total (2022) | Share of Total (2025) | Primary Drivers |
|---|---|---|---|
| Detection and escalation | 29% | 27% | Forensics, IR retainer fees, threat hunting, overtime |
| Lost business | 38% | 40% | Customer churn, downtime, reputation damage, SLA penalties |
| Post-breach response | 23% | 22% | Legal fees, credit monitoring, system hardening, PR |
| Notification | 10% | 11% | Regulatory filings, individual notices, call center operations |
Lost business is now the single largest component and the most variable across industries. Retail and consumer technology firms report churn rates of 3-7% following disclosed breaches, translating to $8-40 million in lost lifetime value for mid-size firms. B2B manufacturers and public sector entities see near-zero churn but absorb cost through delayed procurement cycles and increased audit burden.
Detection and escalation costs are rising fastest in absolute terms. Median time to identify a breach remains 194 days, and every additional day of dwell time adds an estimated $5,600 to total cost. Organizations with dedicated security operations centers and EDR coverage report identification times 60-90 days faster and total costs roughly 22% lower.
Regulatory Fines and Disclosure Rules
Fines remain a minority of total cost - typically 3-8% - but they drive the tail of the distribution. GDPR penalties under Article 83(5) cap at €20 million or 4% of global annual turnover. Cumulative GDPR fines passed €5.9 billion by end of 2025, with the largest single penalty (Meta, €1.2 billion, 2023) exceeding the entire annual breach cost of most affected organizations.
The SEC’s 2023 cybersecurity disclosure rules (Item 1.05 of Form 8-K) require material incident disclosure within four business days of materiality determination. The rule has reshaped cost dynamics for US public companies: average share price decline in the 30 days following a materiality-triggering 8-K filing is 4.2%, and derivative shareholder suits have followed roughly 40% of such filings. For a company with $10 billion market cap, a 4.2% decline represents $420 million in lost market value - dwarfing direct breach costs.
The Cyber Incident Reporting for Critical Infrastructure Act (CIRCIA) final rule, expected to take effect in 2026, will require covered entities to report significant incidents to CISA within 72 hours and ransom payments within 24 hours. Early compliance cost estimates run $200,000-$1.5 million per organization annually, with the reporting itself creating new discovery and litigation exposure.
For organizations modeling breach risk, the actionable takeaway is that direct remediation costs are increasingly a minority of total exposure. Litigation, churn, and market reaction dominate the cost profile, and these are driven by disclosure timing and regulatory posture as much as by the technical severity of the incident. Organizations tracking latest breach reports can benchmark their own exposure against sector peers, while threat intelligence feeds inform which adversary groups target which verticals - a critical input to industry-specific cost modeling.
Regional Analysis
Attack volume and breach economics diverge sharply by region. IBM’s 2025 Cost of a Data Breach study puts the global average at $4.44 million, but regional spreads exceed 300%, driven by regulatory exposure, incident response maturity, and attacker prioritization. The table below summarizes the 2025-2026 picture.
| Region | Avg. Breach Cost (USD) | Incident Growth (YoY) | Top Attack Vector |
|---|---|---|---|
| North America | $5.9M - $9.4M (US) / $5.1M (Canada) | +18% | Phishing / credential theft |
| Europe | $4.1M - $4.8M | +22% | Ransomware (double extortion) |
| APAC | $3.2M - $3.9M | +31% | Ransomware / web exploitation |
| Middle East | $3.5M - $4.2M | +27% | Web application attacks / BEC |
| Latin America | $2.2M - $2.8M | +24% | Banking trojans / ransomware |
North America: The Cost Premium
The United States remains the most expensive place on earth to suffer a breach. IBM’s 2025 figures put the US average at $10.22 million, a figure inflated by class-action litigation, SEC disclosure obligations, and a dense third-party ecosystem that widens blast radius. Healthcare and financial services anchor the top of the distribution. Canada runs roughly 45% below US costs, though incident volume growth is comparable.
Ransomware operators treat North American mid-market manufacturing and healthcare as primary revenue targets. The latest breach reports show a consistent pattern: initial access via phishing or edge appliance exploitation, 4-9 days of dwell time, then exfiltration before encryption. Volt Typhoon and related China-nexus actors have shifted toward pre-positioning in US critical infrastructure - water utilities, energy, and transportation - with the explicit goal of maintaining persistent access rather than immediate monetization. CISA’s 2025 advisories confirm living-off-the-land tradecraft that defeats signature-based detection.
Europe: Regulation as Both Shield and Liability
European breach costs average $4.1M to $4.8M, but the distribution is bimodal: GDPR-exposed organizations with mature programs report lower per-record costs, while first-time offenders face compounding fines. NIS2 transposition deadlines through 2025-2026 have expanded the regulated population from roughly 15,000 entities to over 160,000 across the EU, pulling thousands of mid-sized firms into mandatory incident reporting and supply chain security obligations for the first time.
DORA (Digital Operational Resilience Act), fully applicable since January 2025, imposes 4-hour initial incident notification on financial entities and their ICT third-party providers. ENISA’s 2025 Threat Landscape report flags ransomware as the dominant threat to EU entities, accounting for 44% of observed incidents, with manufacturing and public administration most affected. Interpol’s African Cyberthreat Assessment notes Europe-facing fraud infrastructure increasingly routed through West African proxies.
APAC: Fastest Growth, Weakest Baseline
Asia-Pacific records the fastest incident growth at +31% YoY. Ransomware against Japanese, Australian, and Southeast Asian manufacturing and logistics has accelerated, with groups like Akira and LockBit successors targeting OT-adjacent systems. Web exploitation dominates initial access - unpatched VPN appliances and public-facing APIs remain the primary entry point, tying directly to the Path Traversal vulnerabilities and Remote Code Execution vulnerabilities classes that dominate regional CVE exploitation.
Interpol’s 2025 assessment highlights Southeast Asia as a hub for scam-center operations and ransomware affiliate recruitment. Australia’s mandatory ransomware reporting regime, in force since 2024, has improved visibility but not reduced volume. India and Indonesia show the steepest growth in web application attacks.
Middle East: State-Aligned Targeting
Middle Eastern incident growth of +27% is heavily shaped by geopolitical targeting. Iranian actors (MuddyWater, APT35) and Israel-linked operations trade blows across financial, energy, and government sectors. Web application attacks and business email compromise dominate, with SQL Injection vulnerabilities and Authentication Bypass vulnerabilities recurring in regional incident data. Saudi and UAE investment in SOC maturity has reduced dwell time, but third-party exposure remains a weak point.
Latin America: Financial Sector Under Siege
Latin America’s lower average cost ($2.2M-$2.8M) reflects both smaller deal sizes and less mature disclosure. Banking trojans (Grandoreiro, Mekotio) and ransomware against Brazilian and Mexican financial institutions drive volume. Interpol flags the region as a growing ransomware affiliate market, with initial access brokers selling Latin American credentials at a discount relative to North American targets.
Geopolitical Targeting Patterns
APT28 (Fancy Bear) continues to prioritize European NATO members and Ukraine-adjacent logistics, with credential phishing and router exploitation. Volt Typhoon’s US focus and APT28’s European focus illustrate a broader trend: state-aligned actors select regions by strategic value, while criminal groups select by ability to pay. The overlap - criminal affiliates leasing access from state-aligned infrastructure - is the defining regional risk of 2026.
Sector Spotlights: Healthcare, Finance, Manufacturing, and Public Sector
Four sectors account for a disproportionate share of 2026 incident volume and loss. Each subsection below pairs a quick-reference table with the statistics that matter most.
Healthcare
Ransomware remains the defining threat. The Change Healthcare breach set the scale benchmark: roughly 190 million individuals affected, with recovery costs estimated near $2.9 billion. 2026 data shows the playbook has not changed, only scaled - attackers now routinely exfiltrate before encrypting, doubling extortion leverage.
| Metric | 2026 Figure |
|---|---|
| Confirmed incidents (global) | ~1,240 |
| Average total cost per breach | $11.2M |
| Top vectors | Ransomware (58%), third-party/supply chain (19%), phishing (14%) |
| Regulatory exposure | HIPAA, HHS 405(d), SEC disclosure, GDPR |
- Healthcare has led all sectors in average breach cost for 14 consecutive years.
- Median time from initial access to encryption fell to under 4 hours in observed ransomware cases.
- Over 60% of incidents involved business associates or vendors, not the covered entity directly.
Finance
Finance faces a different mix: business email compromise (BEC), DDoS extortion, and payment fraud dominate, with ransomware less prevalent but more disruptive when it lands. BEC losses reported to the FBI’s IC3 continue to climb, with wire-fraud and vendor-impersonation schemes the most common variants.
| Metric | 2026 Figure |
|---|---|
| Confirmed incidents (global) | ~980 |
| Average total cost per breach | $6.4M |
| Top vectors | BEC/phishing (41%), credential attacks (22%), DDoS extortion (17%) |
| Regulatory exposure | PCI DSS, GLBA, DORA (EU), SEC, PSD2 |
- DDoS extortion campaigns now pair volumetric floods with data-theft threats, pressuring institutions to pay without a ransom note tied to encryption.
- Financial institutions saw the highest share of credential-stuffing attempts, driven by infostealer malware markets.
- Regulatory reporting windows (DORA’s 4-hour initial notification) compress incident response timelines.
Manufacturing
OT and ICS environments are the front line. Dragos year-over-year data shows a steady rise in ransomware groups specifically targeting industrial operations, with the goal of halting production rather than stealing patient or customer records. IP theft and design-data exfiltration remain persistent, often undetected for months.
| Metric | 2026 Figure |
|---|---|
| Confirmed incidents (global) | ~1,090 |
| Average total cost per breach | $5.8M |
| Top vectors | Ransomware against OT-adjacent IT (44%), IP theft/espionage (21%), phishing (18%) |
| Regulatory exposure | NIS2 (EU), TSA directives (US), IEC 62443, sector-specific rules |
- Ransomware incidents affecting industrial firms rose roughly 30% year over year, per Dragos tracking.
- Attackers increasingly pivot from IT into OT via exposed remote-access and VPN appliances.
- Mean time to detect IP-theft intrusions exceeded 180 days in reported cases.
Public Sector
Municipalities, school districts, and government agencies remain high-value, low-budget targets. Ransomware against local government and education continues at scale, and data exfiltration of citizen and student records adds regulatory and reputational fallout.
| Metric | 2026 Figure |
|---|---|
| Confirmed incidents (global) | ~1,310 |
| Average total cost per breach | $4.1M |
| Top vectors | Ransomware (52%), data exfiltration (23%), phishing (15%) |
| Regulatory exposure | FERPA, state breach laws, GDPR, CISA reporting rules |
- Public-sector incidents outnumber every other sector in raw count, driven by school districts and small municipalities.
- Ransom demands against public entities averaged lower than enterprise figures, but recovery downtime stretched longer due to budget and staffing constraints.
- Exfiltration-only attacks (no encryption) rose sharply, often tied to hacktivist and data-leak-site operations.
For sector-specific incident breakdowns and emerging campaigns, see the latest breach reports and threat intelligence.
Emerging Trends: AI, Cloud, and Identity Attacks
Three vectors account for most of the year-over-year delta in incident volume: AI-assisted attack generation, cloud control-plane misconfiguration, and identity compromise. Each lowers the cost or skill floor for intrusion, and each compounds the others.
AI-Accelerated Attack Generation
Generative models have collapsed the time and language proficiency required for convincing social engineering. Phishing remains the top initial access vector in Microsoft’s and Mandiant’s annual telemetry, and 2025-2026 data shows why: large language models produce grammatically native, context-aware lures at scale, removing the tell-tale errors that trained users once flagged. Deepfake voice and video are now routine in business email compromise and vendor fraud, with finance and HR teams the most frequent targets. On the reconnaissance side, automated agents enumerate exposed services, fingerprint software versions, and map attack surfaces faster than manual operators, feeding directly into Remote Code Execution vulnerabilities and SSRF vulnerabilities exploitation pipelines.
| AI Attack Use | Observed Impact |
|---|---|
| Phishing/ lure generation | Higher click and credential-submission rates; native-language lures across regions |
| Deepfake voice/video | BEC and vendor fraud; wire-transfer authorization bypass |
| Automated recon | Faster time-to-exploit after disclosure |
| Polymorphic malware | Evasion of signature-based detection |
Cloud Misconfiguration
Exposed storage buckets, over-permissive IAM roles, and unauthenticated control-plane endpoints remain the dominant cloud exposure class. The recurring pattern is privilege sprawl: identities granted broad permissions for convenience become the pivot for lateral movement once any single credential leaks. Public cloud incidents increasingly combine a misconfiguration with a stolen token rather than standing alone.
Identity as the Dominant Initial Access Method
Microsoft and Mandiant both report that valid accounts are now the single most common initial access technique, ahead of exploitation of public-facing applications. Stolen credentials, session tokens, and MFA bypass drive the shift. MFA fatigue (push bombing) and adversary-in-the-middle phishing kits that relay real-time tokens remain effective against push-based and SMS-based factors.
Infostealers are the supply chain feeding this market. Commodity families (RedLine, Lumma, Vidar, and successors) harvest browser-stored credentials, cookies, and session tokens at volume, and the resulting infostealer logs are sold in bulk. Pricing data illustrates the economics:
| Credential Type | Typical Market Price |
|---|---|
| Bulk infostealer log (per record) | Under $1 to a few dollars |
| Corporate VPN/SSO credential | $50-$500+ depending on access |
| Valid account with MFA-bypass session | Premium tier, often bundled |
Cheap, abundant logs make credential stuffing and session hijacking economically trivial, and stolen sessions that skip MFA entirely are the reason identity now outranks exploitation as an entry point.
The Defensive Counterpoint
Defensive AI adoption is climbing in parallel. Security operations teams report growing use of ML-based anomaly detection, automated triage, and AI-assisted phishing analysis, and identity providers are pushing phishing-resistant FIDO2/passkeys to retire push-based MFA. The gap remains execution speed: attackers weaponize new techniques in days, while enterprise rollout of passwordless authentication and conditional access still runs in quarters. Organizations tracking exposure should prioritize phishing-resistant MFA, least-privilege IAM, and continuous monitoring of credential-leak sources, and should review Authentication Bypass vulnerabilities and Privilege Escalation vulnerabilities as the identity perimeter tightens.
Key Takeaways and Defensive Recommendations
The 2026 dataset points to a threat landscape where initial access is cheap, dwell time is compressed, and the highest-leverage controls are identity and edge infrastructure. Seven findings should drive budget and program priorities for the next planning cycle.
| # | Finding | 2026 figure | Defensive priority |
|---|---|---|---|
| 1 | Phishing remains the top initial access vector | 41% of confirmed intrusions | Phishing-resistant MFA, not just TOTP |
| 2 | Ransomware payment rates | 34% of victims paid, down from 46% in 2024 | Assume extortion without encryption; protect backups |
| 3 | Median breach cost | $5.1M, up 9% year over year | Invest in detection and response, not just prevention |
| 4 | Supply chain involvement | 29% of incidents touched a third party | Vendor risk tiering with contractual SLAs |
| 5 | Identity-based attacks | 61% of incidents involved credential abuse | Session token protection, conditional access |
| 6 | Edge device exploitation | Median time-to-exploit under 48 hours | Patch SLAs measured in hours for internet-facing systems |
| 7 | Detection gap | Median dwell time 11 days | Tabletop exercises and 24/7 triage coverage |
Prioritized Recommendations
1. Close the MFA gap on every externally reachable identity. Organizations reporting above 95% phishing-resistant MFA coverage saw 71% fewer successful account takeovers. Prioritize Authentication Bypass vulnerabilities in legacy SSO and VPN concentrators, where MFA is often enforced inconsistently.
2. Enforce a 72-hour patch SLA for edge devices. Firewalls, VPN gateways, and email security appliances accounted for 23% of exploited entry points. Given sub-48-hour exploitation windows, monthly patch cycles are effectively unpatched. Track Remote Code Execution vulnerabilities affecting perimeter products as a standing operational metric.
3. Build a tiered third-party risk program. With 29% of incidents involving a supplier, tier vendors by data access and enforce security requirements in contracts, including breach notification timelines and evidence of patching. Review the latest breach reports to model realistic supplier failure scenarios.
4. Protect session tokens and service accounts. Token theft bypasses MFA entirely. Inventory non-human identities, rotate service account credentials, and monitor for anomalous session reuse.
5. Run quarterly tabletop exercises tied to real incidents. Organizations with tested response plans contained incidents a median of 38% faster. Use current threat intelligence to build scenarios reflecting active ransomware and extortion tactics.
6. Instrument for detection, not just prevention. With an 11-day median dwell time, EDR coverage gaps and unmonitored cloud control planes are the largest blind spots.
Data Limitations and What to Watch in 2027
Self-reported breach cost figures skew toward large enterprises and undercount small business incidents. Ransomware payment data depends on voluntary disclosure and likely understates actual payments. Sector attribution varies across sources due to differing classification methods.
For 2027, watch three signals: the maturation of AI-assisted social engineering against voice and video channels, regulatory pressure on ransom payment disclosure, and whether identity provider compromises become a distinct category in incident reporting. Track developments in cybersecurity news as the year unfolds.
Related Resources
Explore 150+ real SOC analyst interview questions for 2026. This database covers technical, behavioral, and scenario-based questions to help you ace your next cybersecurity interview.
Master a complete malware analysis workflow in 2026: static and dynamic analysis techniques, tools, and step-by-step processes to safely dissect malicious code.
Learn how the MITRE ATT&CK framework maps adversary tactics and techniques, and discover practical use cases for enterprise security and SOC teams.
Explore common cloud misconfigurations in AWS, Azure, and GCP that lead to data breaches. Learn how to identify and prevent these critical security flaws.