Yearly security audit · 2025
Hash XR, January 2026. Same yearly series as before — this issue is the 2025 record.
Executive summary. Markets moved sharply higher in 2025 versus 2024. In our count, though, attackers were less busy: fewer incidents than in 2024.
About US $1.47 billion left in 145 notable incidents.
Terms used in this report, before the numbers:
CCBS means “Centralized Crypto or Blockchain Service”: a product or service around crypto or chains, run by a conventional firm — a centralized exchange such as Binance, or an issuer such as Tether.
Flash loans are a tool attackers lean on against EVM contracts. The AAVE team [1] shipped the pattern: borrow any available liquidity with no collateral, so long as it returns inside the same block [2]. The usual path is to pull ERC-20s [3] and hit a DeFi app. The caller writes a contract that borrows, uses the funds, and repays principal plus interest and fees in that one transaction.
A cross-chain bridge links independent networks so value, data, or messages can move from one chain to another.
As each chain grows its own apps and assets, demand to move value across them rises — and so does bridge volume. That is why bridges keep drawing fire.
This report tallies notable 2025 incidents, reads their root causes, and lists practices we would actually use.
We took 145 notable 2025 incidents and broke them down by target and by cause.
Exploited value in 2025 was US $1.47 billion. Tradingview put the crypto market near US $3.26 trillion. The stolen slice was 0.05% of that cap.
We group the incidents into four target types:
2. Blockchains
3. DApps 4. Cross-chain bridges. A CCBS incident is an attack on a centralized crypto or chain service that knocks the service down or drains assets it holds.
A chain incident is an attack — from inside, outside, or both — on a mainnet, side chain, or layer 2 that knocks operations off, or a failure of software or hardware that lets attackers abuse consensus for profit.
A dApp incident is when daily operation breaks or is attacked, so users and assets the app holds become fair game.
A bridge incident is an attack that drains assets the bridge holds or stops the exchange between the linked chains.
145 incidents in all. The figure below splits them by target.
dApp incidents are more than 89.66% of the set. Of 145: 7 CCBS, 6 chain, 0 bridges, 130 dApps.
Chain incidents split three ways: i. mainnets ii. side chains iii. layer 2s.
A mainnet (layer 1) is its own network: protocol, consensus, validators. Its validators finalize its own transactions, data, and blocks. Bitcoin and Ethereum are the usual examples.
A side chain is a separate chain that runs beside a mainnet, with its own consensus and validators, linked — often by a two-way peg [4].
A layer 2 is a protocol or network that borrows security and finality from a base chain [5]. The job is scale: faster, cheaper transactions than L1. Since 2021 the Ethereum L2 set has grown quickly.
Side chains and L2s both exist to scale a mainnet. The split: a side chain need not lean on that mainnet for security or finality; an L2 does.
Six chain incidents in 2025. The figure splits mainnet, side chain, and L2.
Mainnet 66.67% (4), L2 33.33% (2). No notable side-chain incident sat in this set.
Of 130 dApp incidents: 5 rugs, 38 exploit-adjacent, 87 direct hits. A hit can land on the front end, the server, or the contracts. Those 87 split as: i. front end ii. server
iii. the contracts
Front-end incidents abuse ordinary client bugs to take account data and personal details, then the user’s assets.
Server-side incidents abuse ordinary backend bugs so on-chain / off-chain traffic can be hijacked and user assets taken.
Contract incidents are design or implementation bugs used to pull assets from users.
The figure below splits front end, server, and contracts.
Contracts 85.06%, server 0%, front end 14.94%. Of 87: 13 front end, 74 contracts.
Loss by slice: front end US $95.77 million; contracts US $553.26 million.
Contracts were the main problem. Typical 2025 contract bugs: logic holes, re-entrancy, price games, leaked keys, flash loans, and the rest of the usual set.
The 74 direct contract hits, split by bug class:
Logic bugs were the most common class, then price manipulation. Logic here means missing parameter checks, missing access checks, and similar gaps. 40 projects hit by logic bugs; 12 by price games.
Loss by bug class:
Logic bugs still cost the most: 40 incidents, US $443.28 million — 80.12% of this slice. Re-entrancy was second: 2 incidents, US $27.5 million (4.97%). Price manipulation was third: 12 incidents, US $25.05 million (4.53%).
By cause: i. external attackers ii. rugs
iii. Misc.
The split looks like this.
External attacks 76.55% (111); rugs 3.45% (5).
Loss by cause:
Attacks took 98.99% of the dollars (US $1.46 billion); rugs 0.19% (US $2.77 million). In 2025, outside attackers were the main threat to the ecosystem.
Where attackers aimed:
Hits: dApps 88.07% (96), CCBS 6.42% (7), chains 5.5% (6).
Loss on those targets:
Dollars: dApps 54.54% (US $732.03 million), CCBS 45.01% (US $604.1 million), chains 0.45% (US $6 million).
All 2025 rugs we counted were dApps: 5 events, US $2.77 million — far below attack losses.
dApps took the most hits in 2025 and the most money: US $732.03 million, 54.54% of attacker-driven loss. That is a structural problem for the apps that are already live.
Outside attackers stayed the main threat: over 76.55% of events and over 98.99% of dollars — well ahead of rugs and the rest.
A dApp is a front end, a server, and contracts. Attackers can hit any mix. In 2025 the contracts still absorbed the large majority of both hits and dollars versus UI or server.
Every rug in this 2025 set was a dApp.
Inside contract incidents, logic bugs led both the count and the dollars — ahead of every other subclass.
Below are practices for builders and users facing the same 2025 failure modes, and for a faster joint response. Use as many as your resources allow.
Note: “builders” here means people who ship chains, dApps, or other crypto systems. “Users” means anyone who operates, governs, or trades in those systems.
Bridge teams should treat security as a first-class job now that cross-chain volume is up. A bridge is on-chain and off-chain; the off-chain half is the softer target. Playbooks have to cover that half cleanly.
L2s saw few hits and small losses this year — do not read that as safety. More L2s are coming; the research has to stay ahead of them.
Move admin rights off a single key — onto a multi-sig or a DAO — for assets and other critical
operations. That step is not optional.
Once a contract bug is in view — re-entrancy, missing access checks, a broken price function — attackers reach for a flash loan to size the hit. Those classes of bug should be first on a contract author’s list.
More attackers used Discord and Twitter for phishing through 2025; 2026 will look the same. Users lost real money. Teams should treat social-account security as part of the product, not an afterthought.
Portfolios now sit on more than one chain, so bridge use is up. Every cross-chain move goes through a bridge, and bridges are a favorite target. Check the bridge’s record before you send.
Contract safety still matters when you use a dApp — and so does the UI. Treat odd prompts, copy, and behavior as hostile until proven otherwise.
Before you commit, look for an audit report and actually read it.
Keep idle funds in a cold wallet or a multi-sig. Treat hot wallets as a smaller working set, and keep the machine they live on locked down.
Be slow with a dApp whose team is anonymous or unproven — that is a common rug shape. Be equally slow with an exchange that has no track record or no third-party flow history.
[1] Aave. https://aave.com/
[2] Flash loans. https://aave.com/flash-loans/
[3] ERC-20 token standard. https://ethereum.org/en/developers/docs/standards/tokens/erc-20/
[4] Sidechains. https://ethereum.org/en/developers/docs/scaling/sidechains/
[5] Layer-2. https://academy.binance.com/en/glossary/layer-2