Skip to news
← News
Analysis

B18 TimeBank loss shows logic flaws still drain DeFi

Follow us on Telegram

A day that cost about $715,000 looks minor beside this year's crypto theft total, yet it exposes two routes attackers still use with effect: flaws inside a protocol's own rules, and malicious code placed in web tools that users trust. B18 TimeBank led three cases, and its method, not its size, is the warning for anyone with funds in a decentralised finance service.

Losses to hacks and exploits by month ($ billion)
The latest month is incomplete.Dec 25Feb 26Apr 26Jun 26Aug 266.50Oct 260.06Nov 26Data: DefiLlama hacks database.
Losses by type of attack this year ($ million)
Frontend & Infrastructure: 6,515 (74%)Key Compromise: 722 (8.2%)Bridge & Cross-Chain: 701 (8.0%)Access Control: 371 (4.2%)Token & Share Accounting: 254 (2.9%)Other: 234 (2.7%)74%
  • Frontend & Infrastructure6,51574%
  • Key Compromise7228.2%
  • Bridge & Cross-Chain7018.0%
  • Access Control3714.2%
  • Token & Share Accounting2542.9%
  • Other2342.7%
Data: DefiLlama hacks database.

DefiLlama's public database of hacks, exploits and thefts, covering incidents dated to 17 November 2026, records the three cases across two days. Two are dated before the previous report and were added later, a reminder that the daily picture is often completed in arrears. Read together, they pose a clear question: why do the largest and smallest losses keep arriving through the ordinary machinery of a service, rather than through a direct break of the underlying blockchain?

Two protocols undone by their own logic

A protocol logic attack means the code followed a path its designers did not intend, allowing a withdrawal the rules should have blocked. B18 TimeBank, a DeFi protocol on Base, was exploited on 17 November through its level-reward voucher redemption mechanism. The database account says signed reward vouchers triggered a withdrawal of approximately $404,540 in USDC from the settlement contract in a single identified transaction. The point for a user is narrow and practical. A signed voucher can carry authority, so what a wallet is asked to sign matters as much as where funds are deposited.

MORE Markets was hit a day earlier on Flow EVM, losing approximately $308,000 in WFLOW from its lending reserve. According to the account, the attacker used flash loans, which are borrowed and repaid within one transaction sequence, to open e-mode positions involving ankrFLOW and WFLOW, then repeatedly self-liquidated those positions. The result was unbacked WFLOW debt left in the protocol. Two different products, on two different networks, failed at the level of their own accounting and permissions. Neither account describes a stolen password. Each describes rules that could be turned against the pool that other users had supplied.

A web script that reached far beyond crypto

The third case was smaller in money and wider in reach. Optimizely, listed as OPTIMJACK, was recorded on Bitcoin, Litecoin, TRON, Ethereum and BSC, with about $2,260 taken. This was classed as Frontend & Infrastructure, a supply chain attack: instead of attacking a protocol directly, an attacker inserted malicious JavaScript into scripts from a web experimentation platform used by 178 organizations, including Ledger, HSBC, BBC and Zoom.

Incidents in this report
ProjectNetworkTakenType of attack
B18 TimeBankBase$405,000Protocol Logic
MORE MarketsFlow$308,000Protocol Logic
Optimizely (OPTIMJACK)Bitcoin$2,260Frontend & Infrastructure
Data: DefiLlama hacks database.

The injected code replaced wallet addresses with addresses controlled by the attacker, redirecting suspected victim payments. The account also says the malware attempted to take payment-card information, cookies and screenshots, and delivered exploit kits aimed at iOS devices. Optimizely removed the code and patched the vulnerability, while the full extent of data theft and further losses remained unknown. That unknown is the central fact here. The recorded crypto amount is tiny, but the distribution route reached organisations well outside crypto, so the figure measures identified payments, not the possible exposure of everyone who loaded the affected scripts. For a user, the check is equally concrete: a wallet address shown on a web page is not trustworthy simply because the page itself looks familiar.

A small day inside a much larger year

The year's record puts the three cases in scale and also changes the reading of them. From 1 January to 17 November, the database counts 354 incidents and about $8.80 billion taken where the amount is known. In the same part of 2025 there were 126 incidents and about $2.64 billion. The rise is therefore not only in total value. The number of reported incidents has risen sharply as well, even though one event dominates the money.

That event is Binance CEX on 29 October, at about $6.36 billion, also classed as Frontend & Infrastructure. It explains why losses by attack type put that category first for the year, at about $6.51 billion, far ahead of Key Compromise at about $722.08 million. The recent period shows the same concentration. In the last thirty days to 17 November there were 48 incidents and about $6.45 billion taken; in the thirty days before them, 32 incidents and about $517.19 million. Remove the single largest case from the picture and the year still shows more incidents than in 2025, but the money no longer tells a story of uniform growth. It tells a story of frequent smaller failures punctuated by one web infrastructure loss large enough to move the annual total on its own.

Recovery is the exception, not the pattern

Funds returned this year, where the database records them, amount to about $8.25 million. Set against the annual total, recovery is marginal, and that is the consequence a holder should absorb from this reading. Prevention, in the form of what is signed, which address is used and which web page is trusted, carries almost all of the practical weight, because money that leaves a settlement contract or a lending reserve rarely comes back in any material proportion.

DefiLlama lists incidents that projects, auditors or investigators reported publicly; amounts are estimates at the time and can be revised, and an incident without a known amount is counted but not added to the totals. That limit cuts both ways. It means the annual total is not a complete census, and it means a quiet day in the database is not proof of a quiet day on-chain. The next readings will test whether protocol logic cases keep appearing at this steady pace without a large loss attached. If they do, the lesson of 17 November will stand: the recurring risk for ordinary users sits in permissions and page scripts they can partly check, while the year's headline total was set by infrastructure they cannot check at all.

Not investment advice.

Source: DefiLlama · Date:

B18 TimeBank loss shows logic flaws still drain DeFi

Read next