TL;DR

Web3 hacks have moved in eras: from contract logic bugs to bridge failures and CEX cold-wallet compromises.
Each landmark exploit exposed a different wrong assumption about trust, governance, or composability.
Reusing old mental models (Web2 ops, “battle-tested” code, centralized security) keeps creating new failure modes.
Studying these eras is the fastest way to design better threat models for whatever you are building today.

1. 2016-The DAO and the birth of on-chain failure

In early 2016, The DAO raised more than 11.5 million ETH (about 150 million dollars at the time) from over 11,000 participants, becoming one of the largest crowdfunding events in history.

The goal was simple and radical: an investor-directed, on-chain venture fund with no traditional management. The contracts that powered it were complex, ambitious and, crucially, optimistic about how external calls behaved.

The reentrancy bug that allowed an attacker to recursively drain around 3.6 million ETH from The DAO is now textbook. But at the time, it defined an era. Smart contracts were shown to be brittle not because Ethereum’s cryptography failed, but because contract authors mis-modeled what an “account” could do when you sent it funds.

This first era of Web3 exploits taught three painful lessons: contracts are social contracts, external calls are adversarial by default, and governance has to decide what to do when “code is law” collides with “this will destroy the ecosystem.”

2. 2017-Multisig drama and the Parity wallet incidents

If The DAO showed that application logic could fail, 2017 proved that even the tools people used to hold funds were dangerous.

The Parity multisig wallet bugs came in two waves. In July 2017, a vulnerability in Parity’s multisig contract allowed an attacker to steal over 150,000 ETH, worth roughly 30 million dollars at the time. Later that year, a separate bug led to hundreds of millions of dollars’ worth of ETH being frozen and rendered inaccessible when a user accidentally became the owner of the library contract and then killed it.

The second “hack” was not even a traditional exploit. It was the side-effect of treating shared library contracts as if they were inert infrastructure. The era of multisig drama highlighted that in Web3, the boundary between “product code” and “infrastructure code” is blurry. Bugs in wallet templates or shared libraries can be just as destructive as bugs in the protocols they secure.

It also showed that “security tools” themselves must be threat-modeled, versioned, and governed like any other critical dependency.

3. 2018-2020-Exchange and hot-wallet heists

Before DeFi took over headlines, the biggest numbers were stolen from centralized exchanges and custodians. In January 2018, Tokyo-based exchange Coincheck lost about 530 million dollars in NEM tokens from a compromised hot wallet. Mt. Gox’s multi-year loss of roughly 500 million dollars in bitcoin (revealed in 2014) still defined public perception of crypto risk.

These incidents were less about smart contract bugs and more about operational security: key management, hot-wallet exposure, and the difficulty of running traditional custodial infrastructure in a new asset class. But they set the stage for the perception that “crypto hacks” are nearly always about broken platforms, not broken users.

From a Web3-native perspective, this era is a reminder that centralization bottlenecks remain attractive targets even in a decentralized ecosystem. No matter how elegant your on-chain logic is, a sloppy CEX or custodian can still become the weakest link.

4. 2021-Interoperability euphoria and the Poly Network exploit

By 2021, the conversation had shifted to interoperability. Capital was fragmented across chains; bridges and routing protocols promised to connect everything. Poly Network was one of those projects, facilitating cross-chain transfers between Ethereum, Binance Smart Chain and Polygon.

On August 10, 2021, an attacker exploited a flaw in Poly Network’s cross-chain logic to move around 610 million dollars’ worth of assets to their own addresses, making it one of the largest DeFi incidents by mark-to-market value.

Poly Network on X (formerly Twitter): "Important Notice:We are sorry to announce that #PolyNetwork was attacked on @BinanceChain @ethereum and @0xPolygon Assets had been transferred to hacker's following addresses:ETH: 0xC8a65Fadf0e0dDAf421F28FEAb69Bf6E2E589963BSC: 0x0D6e286A7cfD25E0c01fEe9756765D8033B32C71 / X"

Important Notice:We are sorry to announce that #PolyNetwork was attacked on @BinanceChain @ethereum and @0xPolygon Assets had been transferred to hacker's following addresses:ETH: 0xC8a65Fadf0e0dDAf421F28FEAb69Bf6E2E589963BSC: 0x0D6e286A7cfD25E0c01fEe9756765D8033B32C71

In a surreal twist, the attacker later returned the assets after negotiations and public messaging from the team. But the point had been made: when you move from single-chain contracts to state that spans multiple systems, subtle authorization mistakes become catastrophic.

The Poly era showed that “interoperability” is not just a feature; it is a new class of trust. Every bridge must answer a brutal question: who, or what, is actually allowed to say that value moved?

5. 2022-Bridges as systemic risk: Ronin and Wormhole

If 2021 warned us about interop, 2022 proved that bridges were now systemic infrastructure.

In February 2022, a vulnerability in the Wormhole bridge allowed an attacker to mint 120,000 wrapped ETH on Solana without depositing ETH on Ethereum, draining over 320 million dollars’ worth of value.

A month later, the Ronin bridge that connected Axie Infinity’s sidechain to Ethereum was exploited. Attackers stole 173,600 ETH and 25.5 million USDC, totaling about 540–615 million dollars depending on price at the time, in what became one of the largest crypto heists on record. Subsequent analysis linked the operation to North Korea’s Lazarus Group and revealed that a fake job offer had helped compromise a key employee.

This era marked a shift from “clever DeFi traders exploiting math” to highly resourced, state-backed actors targeting infrastructure that sat between ecosystems. Bridges stopped being just smart contracts and started being geopolitically relevant targets.

For protocol teams, the message was clear: every time you rely on a bridge, you are importing its entire security model, including its human weaknesses.

6. 2023-Composability, flash loans, and language bugs

By 2023, DeFi was deeply composable. Lending protocols stacked on top of AMMs, yield optimizers wrapped both, and bundlers orchestrated complex sequences with a single click. Attackers did the same.

Euler Finance, a permissionless lending protocol on Ethereum, lost about 197–199 million dollars in March 2023 through a sophisticated flash-loan-powered attack that abused donation and liquidation logic in its markets. The incident demonstrated how small configuration choices and complex interactions can be chained into a nine-figure exploit.

That same year, several pools on Curve Finance were drained for roughly 69–70 million dollars due to a bug in certain versions of Vyper, the smart contract language some pools used. For a few chaotic days, the fear was that any protocol using those compiler versions might be silently vulnerable.

This era pushed home a hard truth: composability is a double-edged sword. Every integration, language choice, and configuration knob can amplify risk. The more contracts rely on each other’s invariants, the more systemic the impact of a single mistake.

7. 2024–2025-Cross-chain chaos and institutional-scale thefts

Even as tooling and audits improved, the headline numbers did not magically shrink. In 2024 alone, over 2 billion dollars were stolen across the crypto ecosystem, marking the fourth straight year with more than 1 billion in losses.

Bridge and DeFi exploits remained central, but a new pattern emerged: institutional-scale thefts targeting large centralized players and cross-chain infrastructure. By early 2025, the Bybit cold-wallet hack alone accounted for around 1.5 billion dollars in stolen ETH, making it the single largest known crypto heist to date.

At the same time, research put cumulative bridge losses since 2020 at nearly 2 billion dollars, representing close to 40 percent of all Web3 value hacked. The “cross-chain chaos” era is defined by attackers viewing the ecosystem as one connected graph, not isolated chains. They steal where funds are most concentrated and controls are least mature, regardless of whether that’s a DeFi protocol, a bridge, a custodian, or a mining pool.

The lesson is uncomfortable: as Web3 grows up and attracts larger pools of capital, it becomes a target not just for clever individuals, but for organized, well-funded groups that are happy to invest months in reconnaissance for a single, enormous payout.

Putting the eras together

Seen in sequence, these seven eras are less about individual hacks and more about evolving blind spots:

In 2016, we underestimated how adversarial contract calls could be.
In 2017, we underestimated the risk of the tools we used to “add security.”
In 2018–2020, we underestimated how fragile centralized infrastructure remained.
In 2021, we underestimated the complexity of cross-chain logic.
In 2022, we underestimated how attractive bridges would be to state-backed attackers.
In 2023, we underestimated how small composability and language bugs could scale into systemic failures.
In 2024–2025, we underestimated how big and how targeted Web3 thefts could become once institutional-scale capital was in play.

The common thread is not “hackers got smarter.” It is that each new wave of innovation created assumptions that were never truly threat-modeled before real money arrived.

If you are designing or defending a protocol today, the practical takeaway is simple: treat these eras as a checklist of assumptions to revisit. Governance, wallets, exchanges, cross-chain messaging, composability, compilers, custodians and bridges have all failed at scale before. None of them get a free pass just because they are “battle-tested.”

How Exploitless uses this history

At Exploitless, we treat hack history as raw material for better design, not just war stories.

When we work with teams, we use incidents like The DAO, Parity, Poly Network, Wormhole, Ronin, Euler and the recent bridge and exchange heists as reference points for concrete questions:

Where in your architecture do you repeat patterns that failed in those incidents?
Which external systems (bridges, oracles, custodians) you depend on would deserve their own postmortem if they failed?
How exactly would you detect and respond if your “era-equivalent” bug were exploited tomorrow?

Our goal is not to scare teams with numbers, but to help them build protocols that would be boring to attack. If you want support turning these historical lessons into a practical threat model and runtime security program for your project, that is precisely the work we do at Exploitless.

References

. Background on The DAO crowdsale size, hack mechanics and consequences Wikipedia and historical analyses.(Wikipedia)
. Parity multisig theft and later wallet freeze-security write-ups and news coverage.(OpenZeppelin)
. Major centralized exchange hacks, including Coincheck and Mt. Gox industry summaries and news reports.(Reuters)
. Poly Network exploit amounts and timeline-incident reports and forensic articles.(Wikipedia)
. Wormhole and Ronin bridge hacks-on-chain analyses and coverage of Lazarus Group involvement.(Chainalysis)
. Euler Finance and Curve/Vyper exploits-postmortems and analytics from blockchain security firms.(Chainalysis)
. Aggregate statistics on bridge hacks and 2024–2025 losses, including Bybit and broader DeFi numbers.(Chainlink)

Learn More at Exploitless

Seven Eras of Web3 Hacks: From The DAO to Cross-Chain Chaos