Web3 security incidents totaled $1.31 billion across 344 incidents in the first half of 2026, according to CertiK's Hack3D: H1 2026 report published July 8. The headline figure represents a 46.8% decline from H1 2025's $2.47 billion — but that comparison is misleading. H1 2025 was dominated by a ...
"The weakest link has moved from code to keys and people." — Ronghui Gu, CEO, CertiK
Web3 security incidents totaled $1.31 billion across 344 incidents in the first half of 2026, according to CertiK's Hack3D: H1 2026 report published July 8. The headline figure represents a 46.8% decline from H1 2025's $2.47 billion — but that comparison is misleading. H1 2025 was dominated by a single $1.45 billion Bybit wallet compromise that accounted for 58.7% of the period's losses. Stripped of that outlier, H1 2026 losses are approximately 28% higher on a like-for-like basis.
The data reveals a structural shift in attack methodology. Smart contract code vulnerabilities, long the primary attack surface, have been supplanted by wallet compromise and social engineering as the costliest vectors. Two incidents — the Kelp DAO bridge exploit ($291 million) and the Drift Protocol breach ($285 million) — accounted for 44% of all H1 losses. Neither exploited a bug in smart contract code. Both targeted operational infrastructure: key management systems, RPC nodes, and human access controls. North Korea's Lazarus Group and its subgroups were attributed with $643 million in theft during H1 2026 — approximately 49% of all losses — confirming state-sponsored actors as the dominant structural threat to Web3 capital.
A parallel trend compounds the risk: physical coercion attacks against crypto holders surged 33% year-over-year, with 52 verified incidents and $124.1 million in exposure during H1 2026. Home invasions, not kidnappings, are now the primary vector. France alone accounted for 63.5% of all global wrench attack cases.
CertiK's data breaks H1 2026 losses into three primary categories:
| Attack Vector | Losses | Incidents | Avg. Loss per Incident | |---|---|---|---| | Wallet Compromise | $445M | 33 | $13.5M | | Phishing | $366M | 63 | $5.8M | | Code Vulnerability | $152M | 204 | $745K | | Total | $1.31B | 344 | $3.8M |
Wallet compromise generated $445 million from just 33 incidents — an average of $13.5 million per event and the highest per-incident yield of any category. The vector targets key management infrastructure, multisig governance, and privileged access controls rather than on-chain logic.
Phishing accounted for $366 million across 63 incidents but exhibited a declining trend. Volume fell 52.3% and losses fell 10.8% versus H1 2025. Concentration was extreme: the top four phishing incidents represented approximately 85% of all phishing losses, including a single January 2026 incident worth roughly $285 million.
Code vulnerability exploits remained the most frequent category at 204 incidents but generated the lowest total losses at $152 million. A concerning sub-trend: legacy contract exploits — attacks on older, unpatched contracts — rose from 7 in October 2025 to 18 in May 2026, according to CertiK data. As DeFi's code surface area ages, unmaintained contracts present a growing tail risk.
Adjusted net losses — after accounting for frozen and recovered funds — stood at approximately $1.2 billion, implying a recovery rate of roughly 8.4% ($110 million). This is consistent with the industry's historically low post-incident recovery performance.
April 2026 produced $651 million in losses across 61 incidents, making it the single costliest month for Web3 security since the Bybit incident in February 2025. According to CertiK data, April saw only three days without a recorded hacking incident.
The month's losses were dominated by two events: Kelp DAO on April 18 ($291 million) and Drift Protocol on April 1 ($285 million). Together, these two incidents accounted for 88% of April's total and 44% of all H1 2026 losses.
The concentration of losses in a narrow window raises questions about attacker coordination. Both incidents have been linked to North Korean state-sponsored groups, according to TRM Labs and Chainalysis, suggesting coordinated campaign timing rather than coincidental clustering.
On April 18, 2026, attackers drained 116,500 rsETH — worth approximately $291-293 million — from Kelp DAO's LayerZero-powered cross-chain bridge. The exploit targeted infrastructure configuration, not smart contract code.
Root cause: Kelp DAO operated a 1-of-1 Decentralized Verifier Network (DVN) configuration. A single verification node was responsible for validating cross-chain messages before releasing funds. The attackers compromised internal RPC nodes and launched DDoS attacks against external nodes, feeding false data to the sole verifier. The Ethereum-side contract released funds based on a phantom token "burn" on the source chain that never occurred.
The attack had cascading effects. The rsETH token represents approximately 18% of its 630,000-token circulating supply. Less than 24 hours after the exploit, Aave users reported difficulty withdrawing funds tied to rsETH, triggering a secondary liquidity crunch. Aave froze markets tied to the affected token.
According to Chainalysis, the attack was attributed to TraderTraitor, a Lazarus Group subunit specializing in cross-chain infrastructure compromise. The 1-of-1 verifier setup represented a single point of failure that is standard practice for many smaller bridge deployments but inadequate for securing $291 million in assets.
On April 1, 2026, Drift Protocol — one of the largest perpetual futures platforms on Solana — lost $285 million in approximately 12 minutes. The exploit ranks as the largest in Solana's DeFi history and the second-largest Solana ecosystem loss after the $326 million Wormhole bridge hack of 2022.
Attack sequence: According to TRM Labs, the attack was the culmination of a six-month social engineering operation that began in the fall of 2025. Staging activity on-chain was first observed on March 11, 2026 — nearly three weeks before execution. Attackers manufactured fake tokens, compromised an administrative key, manipulated oracle price feeds, and drained protocol vaults.
The DRIFT token fell more than 40% during the incident. The attack wiped out more than half of Drift's total value locked.
TRM Labs attributed the theft with medium confidence to UNC4736, a North Korean state-sponsored hacking group. The attribution was based on infrastructure overlap with prior Lazarus operations and laundering patterns consistent with DPRK-linked crypto processing.
Drift Protocol subsequently announced a recovery framework involving issuance of recovery tokens pegged to verified user losses and froze approximately $3.36 million in USDC. A 10% public bounty was offered for additional recovery.
DPRK-linked actors stole $643 million in crypto during H1 2026, according to TRM Labs — representing approximately 49% of all crypto stolen in the period and two-thirds of total attributed theft. The April campaign alone accounted for $577 million across the Kelp DAO and Drift Protocol incidents.
The cumulative scale of North Korean crypto theft continues to compound. According to TRM Labs and Chainalysis data, DPRK-linked actors stole approximately $2.02 billion in 2025 — a 51% year-over-year increase from 2024. All-time cumulative theft attributed to DPRK-linked groups now stands at approximately $6.75 billion.
The operational pattern has shifted. Rather than targeting smart contract logic, Lazarus Group subunits — including TraderTraitor, AppleJeus, and UNC4736 — focus on social engineering, supply chain compromise, and infrastructure infiltration. The Drift Protocol attack exemplifies this approach: a six-month social engineering campaign preceded a 12-minute extraction window.
This pattern has implications for institutional Web3 adoption. According to CoinDesk reporting from May 2026, DeFi security vulnerabilities remain the single largest blocker cited by traditional financial institutions evaluating blockchain-based settlement and trading infrastructure. The frequency and sophistication of state-sponsored attacks directly increases the risk premium that institutional capital demands before deploying on-chain.
CertiK's companion Wrench Attacks Report, published July 22, 2026, documented 52 verified physical coercion incidents in H1 2026 — a 33% increase from 39 incidents in H1 2025. Financial exposure reached $124.1 million, nearly 12 times the $10.5 million recorded in H1 2025.
A "wrench attack" encompasses any incident in which victims are physically coerced — through kidnapping, home invasion, armed robbery, or assault — into surrendering cryptocurrency, private keys, or wallet credentials.
Geographic concentration: Europe accounted for 39 of the 52 global incidents (75%). France alone recorded 33 cases — 63.5% of the worldwide total.
Method shift: Home invasions surged from one incident in H1 2025 to 20 in H1 2026, overtaking kidnappings as the most common attack method and representing 41% of all cases. CertiK notes this method "bypasses on-chain security controls entirely."
Notable cases:
The geographic concentration in France is noteworthy. Public blockchain data, combined with social media exposure of crypto wealth, creates a target-rich environment for physical attackers in jurisdictions where crypto ownership is relatively common but physical security measures lag.
CertiK projected that 2026 will close with approximately 130 wrench attacks and hundreds of millions in losses if current trends persist. The firm announced institutional response measures including real-time threat intelligence for regulators, a pro-bono forensic testimony task force, and annual security briefings to INTERPOL and UNODC-affiliated agencies by December 2027.
CertiK CEO Ronghui Gu has characterized the current security landscape as "an unfair game" in which well-funded attackers outspend constrained defenders. According to CertiK analysis, AI tools are accelerating attacker capabilities in three areas: vulnerability discovery across forked codebases, replication of successful attack patterns across protocols, and social engineering at scale.
As smart contract auditing standards improve, the attack surface is migrating. Gu noted that attackers are increasingly targeting supply chain, operational security, and infrastructure layers — areas where audit coverage is thin and defender tooling is immature.
The Web3 security industry is responding with a shift from point-in-time audits toward continuous monitoring. Firms including CertiK, Halborn, OpenZeppelin, Trail of Bits, and Sherlock are expanding into connected security systems that combine audits, on-chain monitoring, researcher networks, and post-launch incident response into unified workflows. Whether this transition can match the pace of attacker evolution remains to be demonstrated.
Post-incident fund recovery in H1 2026 remained low. Of the $1.31 billion in gross losses, approximately $110 million — 8.4% — was frozen or recovered, yielding adjusted net losses of approximately $1.2 billion.
Recovery performance varies by incident. In the Drift Protocol case, the protocol froze $3.36 million in USDC and announced a recovery token framework — representing roughly 1.2% of the $285 million lost. No substantial recovery has been reported for the Kelp DAO exploit.
The low recovery rate reflects two structural factors. First, DPRK-linked laundering networks employ rapid chain-hopping and mixing services that make fund tracing progressively harder with each passing hour. Second, legal complexities around asset freezes — including jurisdictional questions and the risk of litigation from affected users — slow institutional response times.
The H1 2026 security data confirms a structural shift in Web3's threat landscape. The primary attack surface is no longer code — it is people, keys, and operational infrastructure. Smart contract audits, while necessary, are insufficient against social engineering campaigns that unfold over months before executing in minutes.
For the Web3 sector, this shift has direct economic implications. The $1.31 billion in H1 losses represents a de facto tax on the ecosystem — capital destroyed that could otherwise compound as TVL, liquidity depth, or protocol revenue. Protocols that generate $10-50 million in annual fee revenue can see years of accumulated value erased in a single incident.
The concentration of losses in DPRK-attributed operations — 49% of H1 total — indicates that Web3 security is not merely a technical challenge but a geopolitical one. Private-sector audit firms and on-chain monitoring tools are structurally mismatched against nation-state attackers with six-month operational timelines and zero legal constraints.
The parallel rise in physical coercion attacks adds a dimension that blockchain architecture cannot address. No multisig threshold or formal verification method can protect against a home invasion. The 12x increase in wrench attack financial exposure — from $10.5 million in H1 2025 to $124.1 million in H1 2026 — suggests that as crypto wealth becomes more visible, the risk profile for individual holders is changing in ways the industry has not adequately addressed.