On June 25, 2026, a supply-chain attack injected malicious JavaScript into the Polymarket frontend, draining approximately $2.94 million in PUSD from at least 11 user wallets in a single morning. The platform's on-chain smart contracts were never breached. The weak point was a compromised third-p...
"We've contained it and removed the affected dependency. We're contacting impacted users and refunding them in full." — Polymarket Traders, Official Statement (June 25, 2026)
On June 25, 2026, a supply-chain attack injected malicious JavaScript into the Polymarket frontend, draining approximately $2.94 million in PUSD from at least 11 user wallets in a single morning. The platform's on-chain smart contracts were never breached. The weak point was a compromised third-party vendor whose code was served directly to users' browsers.
This incident is the latest in a pattern that has come to define crypto security in 2026: the attack surface has migrated off-chain. According to data compiled by Chainalysis, off-chain attack vectors — compromised credentials, social engineering, and supply-chain manipulation — accounted for 76% of all crypto hack losses in 2025, totaling $2.2 billion of the $3.4 billion stolen. Through the first half of 2026, the trend has accelerated, with frontend hijacks, DNS takeovers, and dependency poisoning replacing smart contract exploits as the primary theft mechanism.
At approximately 10:00 UTC on June 25, attackers compromised a third-party dependency used by Polymarket's frontend. Malicious JavaScript was injected into the code served to a subset of users. When victims connected their wallets, the script generated fraudulent approval or signature requests. Once signed, the attacker's contracts drained PUSD — Polymarket's ERC-20 collateral token backed 1:1 by USDC on Polygon.
On-chain investigator Specter traced approximately $2.94 million in PUSD drained from at least 11 wallets. The attacker bridged the funds from Polygon to Ethereum and swapped them into approximately 1,893 ETH, consolidating proceeds into a single address. Blockchain analytics platform Bubblemaps confirmed the damage was "largely contained" to fewer than 15 accounts.
Polymarket confirmed the breach within 15 minutes of the first public report and removed the affected dependency. The platform pledged full refunds to impacted users but did not name the compromised vendor.
This was Polymarket's second security incident in two months. In May 2026, an internal operations wallet key was compromised, draining approximately $700,000 — though user funds were not affected in that instance.
The incidents hit a platform processing substantial volume. Polymarket recorded $10.57 billion in monthly trading volume in March 2026, crossing the $10 billion threshold for the first time. Monthly unique wallets nearly tripled in the six months to February, reaching 840,000, according to TRM Labs.
The Polymarket drain is one entry in a growing ledger of off-chain exploits targeting crypto infrastructure in 2026:
| Date | Target | Vector | Loss | |------|--------|--------|------| | April 14 | CoW Swap | DNS hijack via social engineering of .fi registrar | $1.2M | | May 2026 | Polymarket | Internal ops wallet key compromise | ~$700K | | May 2026 | 600+ npm/PyPI packages | Mini Shai-Hulud worm (CI/CD supply chain) | Undisclosed | | May 2026 | 34 packages, 384 versions | TrapDoor malware targeting crypto/AI developers | Undisclosed | | June 20 | jaredfromsubway.eth MEV bot | Approval trap via fake token contracts | $7.5M | | June 25 | Polymarket | Third-party vendor frontend injection | $2.94M |
The CoW Swap case is instructive. Attackers impersonated a senior CoW DAO contributor, submitting falsified identification documents to Finland's Communications Regulatory Authority (Traficom), which operates the .fi top-level domain registry. When registrar Gandi failed to respond to the dispute, the attacker obtained domain control and redirected swap.cow.fi to a pixel-perfect phishing clone. At least $1.2 million was drained, including 219 ETH from a single wallet. Core smart contracts remained intact.
The Mini Shai-Hulud worm, documented in MetaMask's May 2026 Crypto Security Report, poisoned over 600 packages across npm and PyPI, affecting repositories associated with OpenAI, TanStack, and Mistral AI. The campaign targeted CI/CD pipelines directly — a vector that, if successfully deployed against a DeFi protocol's build system, could inject malicious code at the compilation stage.
The data is unambiguous. According to Chainalysis's 2026 Crypto Crime Report:
The Verizon 2025 Data Breach Investigations Report found that third-party involvement in breaches doubled from 15% to 30% in a single year — the largest single-year shift ever recorded. IBM's 2025 Cost of a Data Breach Report placed the average cost of a supply-chain compromise at $4.91 million, with an average identification-to-containment lifecycle of 267 days — the longest of any breach vector tracked.
Through mid-2026, the FBI's Internet Crime Report logged over $11 billion in crypto-related losses reported by Americans, prompting the bipartisan Federal Cryptocurrency Theft Enforcement and Coordination Act introduced on June 11 by Representatives Lance Gooden (R-TX) and Josh Gottheimer (D-NJ). The bill proposes a DOJ-housed task force to coordinate crypto-theft investigations across the FBI, DHS, and Treasury.
In the open-source supply chain specifically, the Sonatype 2026 State of the Software Supply Chain Report cataloged more than 454,600 new malicious packages in 2025 across npm, PyPI, Maven Central, NuGet, and Hugging Face — a 75% year-over-year increase. Over 99% of malicious open-source packages now reside on npm.
The off-chain attack surface in crypto can be categorized into three primary vectors:
1. Frontend Injection (Polymarket, Ledger Connect Kit) Attackers compromise a third-party JavaScript dependency that is loaded by the target application's frontend. When users interact with the compromised interface, the injected code generates malicious transaction approval requests. The smart contracts function as designed — the exploit exists entirely in the browser layer.
The 2023 Ledger Connect Kit attack demonstrated this vector at scale, affecting SushiSwap, Revoke.cash, Balancer, and Zapper simultaneously through a single compromised library, with over $600,000 stolen. Polymarket's June 2026 incident follows the same pattern with a different entry point.
2. DNS/Domain Hijacking (CoW Swap) Attackers seize control of a protocol's domain through social engineering of registrars or exploitation of domain management infrastructure. Users are redirected to pixel-perfect clones of legitimate interfaces. This vector bypasses all on-chain security measures and most client-side wallet protections, as the malicious site appears authentic.
3. Build Pipeline Poisoning (Mini Shai-Hulud, TrapDoor) Attackers inject malicious code into developer dependencies at the package-manager level. If a compromised package enters a protocol's build pipeline, malicious code can be compiled directly into production releases. This represents the highest-severity supply-chain vector, as it can compromise smart contract deployments themselves — not just frontends.
MetaMask's May 2026 report also documented the first known AI prompt injection exploit against a live crypto wallet, where an attacker used prompt injection to trick an AI agent into transferring $204,000 — a new vector category that may expand as AI-powered wallet interfaces proliferate.
The economic calculus of supply-chain attacks favors the attacker. A compromised frontend dependency on a high-volume platform provides access to every user who loads the page during the attack window. On Polymarket, which processes over $300 million daily in peak periods, even a 15-minute window of exposure can reach thousands of wallet connections.
The relatively small number of affected wallets in the Polymarket case (11-15) suggests either targeted filtering by the malicious script or rapid containment limiting exposure. In contrast, a smart contract exploit on a protocol of similar scale would typically drain the entire liquidity pool in a single transaction.
This creates a paradox: supply-chain attacks yield smaller per-incident losses but are cheaper to execute, harder to detect, and more repeatable. The Ledger Connect Kit attack compromised multiple protocols simultaneously through a single point of failure. The Mini Shai-Hulud worm poisoned 600+ packages in a single campaign.
Total DeFi and crypto losses exceeded $750 million through mid-April 2026 alone, with two single attacks — Kelp DAO's LayerZero bridge ($292 million) and Drift Protocol ($285 million) — accounting for $577 million. But the aggregate damage from dozens of smaller frontend and supply-chain compromises, many unreported, may rival these headline figures.
The migration of the primary attack surface from on-chain to off-chain has structural implications for the industry:
Audit coverage gaps. Smart contract audits, which consume the majority of protocol security budgets, do not cover frontend code, third-party dependencies, DNS configurations, or build pipelines. A protocol can hold multiple audit certifications while remaining fully exposed to the vectors responsible for 76% of losses.
Vendor concentration risk. Crypto frontends share a common dependency tree. Popular libraries — wallet connectors, UI frameworks, analytics packages — create single points of failure. The Ledger Connect Kit incident demonstrated that one compromised library can simultaneously affect dozens of independent protocols.
Incident response asymmetry. Polymarket detected and contained its breach within 15 minutes. CoW Swap's DNS hijack ran for several hours. Smart contract exploits are typically detected in seconds via on-chain monitoring. The detection gap for off-chain vectors remains significantly wider.
Regulatory pressure. The proposed Federal Cryptocurrency Theft Enforcement and Coordination Act responds directly to the shift in attack patterns. The bill's scope is limited to criminal investigative coordination — it explicitly excludes market regulation — but signals that enforcement agencies recognize off-chain exploitation as the dominant vector.
The Polymarket incident is a data point, not an anomaly. The crypto industry spent the 2020-2024 period hardening smart contracts through formal verification, bug bounties, and audit frameworks. That effort succeeded: pure smart contract exploits are declining as a share of total losses. But the attack surface did not shrink — it relocated. The code that runs in users' browsers, the packages that build production releases, and the DNS infrastructure that routes traffic to frontends now represent the dominant pathway for theft.
The economic structure of the problem is clear. Smart contract security is expensive, well-understood, and improving. Off-chain security — dependency management, vendor vetting, DNS hardening, build-pipeline integrity — is fragmented, underfunded, and worsening. Until security budgets reflect where losses actually occur, the gap between perceived and actual risk will continue to widen.