2026 OKX Web3 Security Half-Year Report

Jointly produced by OKX Web3 Security Team, SlowMist, and OtterSec

Preface

In the first half of 2026, if you only looked at publicly reported losses, they decreased by nearly 60% compared to last year, suggesting the crypto industry is becoming safer. However, in reality, the decline in losses is almost entirely due to the absence of a single, massive event from the same period last year, not a decrease in hacker activity itself. The truth is quite the opposite: the frequency of attacks is increasing, methods are evolving, and attackers' targets are shifting from "code" to "people."

In the past, many projects prioritized contract audits and on-chain activity monitoring. But in the first half of this year, the biggest losses often stemmed not from the contracts themselves. It could be a blindly signed administrative transaction, a leaked cloud key, a developer's computer infected with malware, or even a seemingly realistic video conference.

For OKX, security is not just a slogan in a report, but a real risk that needs to be dealt with every day: malicious addresses, phishing domains, high-risk signatures, abnormal approvals, cross-chain risks, scam tokens, and social engineering attacks are constantly emerging. This report aims to do two things: first, clarify the truly alarming attack trends of the first half of the year; second, share some of OKX's practices in wallets, risk control, cross-chain, and Agent security as a catalyst for discussion within the industry.

Chapter 1 · Security Landscape in H1 2026: Dispersed Incidents, Concentrated Losses

Declining Losses Do Not Mean Fewer Attacks

There's a phenomenon in the first half of this year that is easily misinterpreted: while the total amount stolen decreased, the number of security incidents actually increased. According to incomplete statistics from SlowMist's hacked archive database, there were 182 publicly reported security incidents in H1, resulting in approximately $956 million in losses. Compared to 121 incidents and approximately $2.373 billion in losses in the same period of 2025, the number of incidents increased by about 50% year-over-year, while financial losses decreased by about 60% year-over-year. The apparent decrease in loss amount is mainly due to a single ultra-high-value incident in the same period last year. The absence of an equally extreme event this year does not mean attackers have stopped. After removing outliers, the comparable losses this year have actually increased, not decreased. Attack activity hasn't weakened; it has shifted towards more frequent and dispersed strikes.

For comparison, the largest publicly disclosed security incidents in H1 2025 were as follows:

[Table]

First, large losses are increasingly occurring outside of contract code. The most severe losses in H1 did not come from typical contract vulnerabilities like reentrancy or precision errors, but from operational-level breaches – induced blind signing of administrative transactions, compromised single validator nodes, and stolen cloud signing keys. By incident count, contract and logic vulnerabilities remain the primary cause (85 incidents). However, in terms of loss amount, supply chain attacks topped the list at approximately $298 million, followed by contract vulnerabilities (approximately $152 million) and private key leaks (approximately $130 million). This precisely illustrates that an audit report alone is far from sufficient for a project. Even if a contract has no obvious on-chain logic flaws, as long as there's a single point of risk in the signing process, cloud keys, cross-chain verification, or operational systems, attackers can bypass the contract itself and infiltrate through the weakest link. The boundaries of security have long expanded from "is the code secure?" to "who can sign, where are the keys stored, does verification rely on a single point, and is the operation trustworthy?"

Second, ordinary users are becoming the primary target. As attacking protocols becomes more costly, attackers are turning to users. Phishing websites, malicious browser extensions, search engine ad poisoning, fake customer support, fake job offers, malicious meeting software, clipboard hijacking, and fake 2FA verification were all recurring user-side attack methods in H1. These attacks don't necessarily rely on advanced technology; they exploit users' habits and trust in everyday operations – trusting the first search result, trusting high ratings in app stores, trusting links sent from a friend's account, trusting the "real person" in a video conference, trusting a page that says "security verification." AI makes this cheaper and more realistic: attackers can mass-produce phishing content, forge identities, clone voices, and create deepfake videos, polishing once crude scams enough to fool even experienced users. The success of an attack increasingly depends not on whether the user "understands the technology," but on whether the attacker can precisely hit the blind spots of human trust.

Third, while flourishing, AI Agents are gradually becoming new prey for hackers. As Agents evolve from "being able to converse" to "being able to execute" – calling tools, reading contexts, controlling assets, initiating transactions – the greater their capabilities, the more severe the consequences when compromised. In the past, prompt injection was more of a concept in model security, with the worst-case scenario being the model saying something inappropriate. But when an Agent can sign transactions and transfer funds, a single malicious instruction disguised as normal input can directly translate into a real on-chain loss. Attacks targeting this "cognition-execution" chain are becoming as dangerous as attacks on private keys.

Chapter 2 · Project Perspective: The Biggest Losses Often Occur Outside the Contract

For project teams, the most important lesson from H1 2026 is that problems in operational areas can be just as severe, or even more severe, than contract vulnerabilities.

The following events, compiled by OtterSec based on loss amount, are the major project security incidents of H1 2026. They cover several common high-value attack surfaces: cross-chain bridges, validator infrastructure, signing processes, cloud keys, developer devices, oracles, and access controls.

[Table]

KelpDAO: Attackers Targeted the Verification Path, Not the Contract

KelpDAO represents the largest single loss in H1. The most alarming aspect is that attackers didn't directly attack the contract logic; they attacked the verification path for cross-chain messages. According to OtterSec's post-mortem, the attackers compromised LayerZero's internal RPC node while simultaneously launching a DDoS attack against honest external nodes. Consequently, the single DVN the bridge relied on primarily saw data forged by the attackers and signed off on a withdrawal request that lacked a corresponding real burn. Approximately 116,500 rsETH were transferred out, with about $75 million later frozen.

A single DVN has long been recognized as a high-risk configuration. But in the past, this was more of a theoretical risk reminder in architectural discussions. After the KelpDAO incident, it materialized into a real loss of nearly $300 million. The clear message for project teams is: do not let a single validator, RPC, price source, or signing path determine whether funds can leave. As long as there's only one verification point on a critical path, attackers will prioritize studying it.

Cross-chain bridges and oracles especially need multi-source verification, redundant nodes, outflow rate limits, anomaly monitoring, and emergency pause mechanisms. Otherwise, even if the contract itself has no obvious vulnerabilities, attackers can still drain funds through off-chain infrastructure.

Drift: Attackers Waited Six Months for a Single Signature

If KelpDAO exposed infrastructure fragility, Drift exposed the fragility of the signing process. This wasn't a spontaneous phishing attempt but a long-term social engineering operation. Attackers spent about six months building relationships, waiting for a multisig signer to approve a critical management transaction without fully understanding its implications. The most alarming aspect here was the abuse of Solana's durable-nonce mechanism. The attackers induced relevant personnel to pre-sign multisig authorization transactions. At the time of signing, these transactions appeared to have no immediate impact. However, after Drift subsequently adjusted the multisig threshold, the attackers broadcast the already-obtained signed transactions, draining over 50% of the TVL in a very short time.

This type of attack reminds projects that a transaction appearing "harmless" at the moment of signing doesn't mean it cannot be exploited later. Blind signing, pre-signing, and unparsable management transactions should all be considered high-risk operations. Critical transactions must have clear parsing, simulation, and review processes. For multisig teams, the signing process itself should be protected as a core asset.

OKX Practice: From Transaction Parsing to Signature Risk Protection

From OKX's perspective, protecting against signature risks cannot be limited to simply determining if "the transaction was initiated by the user." More critical is whether the user truly understands the consequences of a transaction before signing. For high-risk scenarios like Solana durable nonce, account ownership changes, and nonce account initialization, OKX has established capabilities in identification, alerting, interception, and isolation across multiple risk rules such as solana_assign_account_owner, solana_init_nonce_account, and nonce_account_risk. In H1 2026, these rules intercepted or alerted on over 4 million related high-risk operations, protecting assets totaling approximately $526 million, helping users identify and avoid such risks before signing.

Concurrently, OKX continues to advance its transaction parsing capabilities. Our goal is to make on-chain transactions more transparent and readable, striving for "What You Sign Is What You Get": users should see not a string of incomprehensible calldata or instructions, but what the transaction actually intends to do, which assets it will affect, what permissions it grants, and whether any abnormal risks exist. To date, OKX has parsed and matched over 50,000 on-chain methods, helping users better understand the transactions they are about to sign.

Chapter 3 · User Perspective: Trust Becomes the Sharpest Weapon

User-side risks continued to rise in H1. Many attacks no longer start with a strange link, but from entry points where users are familiar and let their guard down: app stores, search results, friend accounts, meeting software, recruitment processes, and customer service emails.

Phishing Starts Wearing the Shells of Real Platforms

Phishing remains the primary method for stealing user assets, but its forms are upgrading.

One common method is malicious browser extensions. Attackers often mimic well-known wallet tools, replicating brand names, icons, and page copy. They then use fake reviews and downloads to make the extension look like a legitimate product. Users seeing it on the official app store easily lower their guard, thinking it's the plugin they know. These extensions often employ a "local shell, cloud poison" strategy. This strategy allows them to not directly contain malicious logic, making it easier to pass the store's static review. The truly dangerous phishing page is delivered in real-time from a remote server, allowing attackers to swap pages, change domains, and even display different content to different users at any time. Once you enter your seed phrase or private key, your asset control is gone.

Another type is search engine ad phishing. Attackers buy ad slots for popular keywords, placing fake official websites at the top of search results. In a typical case from H1, a user searching for a development tool on a new computer clicked the top ad and followed on-screen instructions to execute a piece of "installation command" in their terminal. This command actually deployed a clipboard hijacker trojan, allowing the attacker to tamper with the user's pages. Later, when the user attempted to transfer about $20,000, the recipient address was automatically replaced, sending the funds to the attacker. The difficulty in defending against this type of attack is that the user wasn't doing anything obviously dangerous. They were just searching for an official website, downloading a tool, and copying a command – actions that are part of daily routine.

Social Engineering: Most Active and Most Destructive

The key to social engineering isn't technology; it's making the victim let their guard down at the crucial moment.

The most common method is to impersonate someone the victim already knows. In one real case, a victim received an event invitation from a long-trusted friend. The 'friend' insisted on downloading a specific meeting software. Although hesitant, the victim installed it because they trusted the friend. Hours later, their wallet was emptied. It was later discovered that the friend's account had been hijacked by the attacker. High-impact KOLs are also heavily targeted; attackers create high-simulating accounts and use the public figure's credibility to lure fans into fake events, airdrops, and investment groups. For the average user, the difficulty isn't judging whether a stranger is trustworthy, but judging whether an account that "looks like an acquaintance or celebrity" has been hijacked or imitated.

Job recruitment and interview scams are also becoming more targeted. Attackers first make contact through "technical interviews," "operations interviews," or "volunteer interview" scenarios, asking victims to share their screen, open their wallet, and demonstrate their DeFi experience. On the surface, this looks like an interview process; in reality, the attacker is recording wallet addresses, holdings, frequently used protocols, and operational habits. In one real case we observed, the attacker learned about the victim's recent interaction protocols and preferences through an interview. They then forged an airdrop page tailored to a protocol the victim had actually used and sent a highly customized phishing message, ultimately defrauding them of approximately $88,000.

Two Newly Emerging Techniques to Watch

The first type is the fake "2FA Security Verification" scam.

Attackers send emails disguised as official wallet communications, using spoofed domains that differ by just one character and creating urgency with countdown timers, prompting users to enter their seed phrase to "complete verification." It needs to be stressed repeatedly: any page asking for your seed phrase for verification, authentication, recovery, or upgrade is a scam. Your seed phrase is not a verification code; it IS control of your assets. No legitimate wallet will ever ask for it via a webpage for any reason.

The second type is business process fraud.

These attacks don't look like phishing; they look like normal work. Attackers use scenarios like "confirming company legal name," "external audit," "token vesting confirmation," or "supplementary cooperation materials" as bait, delivering malicious attachments disguised as Word files, PDFs, or collaborative documents. Once opened, the malicious program disguises itself as a system update, prompting the user to enter their system password and requesting permissions for the camera, screen recording, keyboard logging, etc. The target of these attacks isn't necessarily just the individual's personal wallet. Often, the real goal is access to the work terminal, browser sessions, password manager, cloud service permissions, and the project's internal systems.

OKX Practice: Positioning User Protection at Devices and Access Points

Looking at the evolution of user-side attacks, relying solely on on-chain protection is no longer enough. Many losses don't start with an on-chain transaction but much earlier: the user downloads a malicious app, installs a disguised plugin, clicks on a phishing site, or continues signing and transferring funds on a compromised device. Therefore, OKX is moving user protection forward from on-chain transactions to devices, applications, and access points. OKX has launched a Security Scan Assistant to help users identify risky applications hidden on their devices, reducing the risk of asset loss from malware, disguised applications, remote control tools, or clipboard hijackers. As of the writing date, OKX has completed over 200,000 risk detections, identified over 60,000 high-risk applications, and guided users to uninstall or handle them. It currently covers Windows and Android systems, with a MacOS version coming soon.

Additionally, for phishing websites and malicious DApps, OKX performs risk identification and alerts users at key scenarios like visiting a potentially risky URL, connecting a wallet, or initiating an interaction. As of the writing date, OKX has intercepted over 7 million visits to risky websites, helping users avoid phishing risks before they enter seed phrases, connect wallets, or sign transactions. For users, the best security warning isn't a notification after a loss occurs. It's an extra step of interception before the risk actually hits the chain. OKX aims to intercept more attacks before signing and transactions through device risk detection, URL risk identification, DApp risk alerts, and on-chain transaction parsing.

Chapter 4 · How AI is Reshaping Attacks: From Content Forgery to "Synthetic Reality"

AI's transformation of the security landscape is the most profound structural change of H1. Its impact is multifaceted. On one hand, AI makes it easier for attackers to generate phishing emails, fake websites, fake customer service scripts, and malicious code. On the other hand, it makes voices, videos, identities, and community environments easier to forge. Going further, when AI Agents themselves begin to hold funds or control transaction tools, they become new attack targets.

New Battlefield: When Agents Gain Access to Funds and Transaction Execution

At OKX, we see AI Agents not just as a new source of risk, but potentially a crucial gateway for Web3 to reach more users. OKX is advancing the Agentic Wallet, allowing users to understand strategies, manage on-chain operations, and execute complex DeFi interactions through an Agent, subject to authorization and confirmation.

By integrating vetted DeFi project plugins, the Agentic Wallet can consolidate multi-step on-chain operations like Swaps, lending, yield management, and cross-chain transfers into a more natural interaction flow. For many ordinary users, this lowers the barrier to understanding and using on-chain finance, making Web3 services feel closer to familiar product experiences. However, precisely because the Agent begins to handle assets, permissions, and transaction execution