Solana Programs - Huma Finance


Prepared by:

Halborn Logo

HALBORN

Last Updated 04/01/2025

Date of Engagement: March 10th, 2025 - March 24th, 2025

Summary

100% of all REPORTED Findings have been addressed

All findings

4

Critical

0

High

0

Medium

0

Low

2

Informational

2


1. Introduction

Huma Finance engaged Halborn to conduct a security assessment on permissionless program beginning on March 10th, 2025 and ending on March 24th, 2025. The security assessment was scoped to the smart contracts provided in the GitHub repository huma-solana-programs, commit hashes, and further details can be found in the Scope section of this report.

The Huma Finance team is releasing the permissionless program, which empowers liquidity providers (LPs) to participate in Huma liquidity pools in a fully compliant, permissionless manner. LPs benefit from attractive double-digit yields along with Huma Feathers as additional rewards.

2. Assessment Summary

Halborn was provided 10 days for the engagement and assigned one full-time security engineer to review the security of the Solana Programs in scope. The engineer is a blockchain and smart contract security expert with advanced smart contract hacking skills, and deep knowledge of multiple blockchain protocols.

The purpose of the assessment is to:

    • Identify potential security issues within the codebase.

    • Validate that the lenders have access to participate in the protocol by depositing their assets

    • Check that the platform allows lenders to request redemptions of their tokens in a permissionless fashion

    • Verify that the funds are correctly managed so they can only be accessible by the correct entities


In summary, Halborn identified some improvements to reduce the likelihood and impact of risks, which should be addressed by the Huma Finance team. The main ones were the following: 

    • Implement a functionality to refresh the assets of the different pool modes.

    • Verify that the modes provided to the entry points are not duplicated.



Most of the findings were addressed, and the corresponding fixes have been merged into the branches listed below. The final commits reflect the changes that solved the issues:

    • 4217cfb901a60e7a8a4166673b6884a0c7ada423 on develop branch

    • 7eb9e6c97b74edeffdd2fd7cf53f750067fb19bc on main branch



3. Test Approach and Methodology

Halborn performed a combination of manual review and security testing based on scripts to balance efficiency, timeliness, practicality, and accuracy in regard to the scope of this assessment. While manual testing is recommended to uncover flaws in logic, process, and implementation; automated testing techniques help enhance coverage of the code and can quickly identify items that do not follow the security best practices. The following phases and associated tools were used during the assessment:

    • Research into architecture and purpose.

    • Differences analysis using GitLens to have a proper view of the differences between the mentioned commits

    • Graphing out functionality and programs logic/connectivity/functions along with state change


4. Caveats

    • Following a requirement from the Huma Finance team to prevent lenders from canceling redemption requests, a new commit (216f890) was added to the scope of this audit. The corresponding functionality was reviewed and confirmed to operate as intended.

    • Additionally, Huma Finance introduced a security feature aimed at improving the process for closing lender accounts. This enhancement was reviewed and confirmed to implement the intended security upgrade correctly. The corresponding functionality is included in Pull Request #208 of the huma-solana-programs repository.


5. RISK METHODOLOGY

Every vulnerability and issue observed by Halborn is ranked based on two sets of Metrics and a Severity Coefficient. This system is inspired by the industry standard Common Vulnerability Scoring System.
The two Metric sets are: Exploitability and Impact. Exploitability captures the ease and technical means by which vulnerabilities can be exploited and Impact describes the consequences of a successful exploit.
The Severity Coefficients is designed to further refine the accuracy of the ranking with two factors: Reversibility and Scope. These capture the impact of the vulnerability on the environment as well as the number of users and smart contracts affected.
The final score is a value between 0-10 rounded up to 1 decimal place and 10 corresponding to the highest security risk. This provides an objective and accurate rating of the severity of security vulnerabilities in smart contracts.
The system is designed to assist in identifying and prioritizing vulnerabilities based on their level of risk to address the most critical issues in a timely manner.

5.1 EXPLOITABILITY

Attack Origin (AO):
Captures whether the attack requires compromising a specific account.
Attack Cost (AC):
Captures the cost of exploiting the vulnerability incurred by the attacker relative to sending a single transaction on the relevant blockchain. Includes but is not limited to financial and computational cost.
Attack Complexity (AX):
Describes the conditions beyond the attacker’s control that must exist in order to exploit the vulnerability. Includes but is not limited to macro situation, available third-party liquidity and regulatory challenges.
Metrics:
EXPLOITABILITY METRIC (mem_e)METRIC VALUENUMERICAL VALUE
Attack Origin (AO)Arbitrary (AO:A)
Specific (AO:S)
1
0.2
Attack Cost (AC)Low (AC:L)
Medium (AC:M)
High (AC:H)
1
0.67
0.33
Attack Complexity (AX)Low (AX:L)
Medium (AX:M)
High (AX:H)
1
0.67
0.33
Exploitability EE is calculated using the following formula:

E=meE = \prod m_e

5.2 IMPACT

Confidentiality (C):
Measures the impact to the confidentiality of the information resources managed by the contract due to a successfully exploited vulnerability. Confidentiality refers to limiting access to authorized users only.
Integrity (I):
Measures the impact to integrity of a successfully exploited vulnerability. Integrity refers to the trustworthiness and veracity of data stored and/or processed on-chain. Integrity impact directly affecting Deposit or Yield records is excluded.
Availability (A):
Measures the impact to the availability of the impacted component resulting from a successfully exploited vulnerability. This metric refers to smart contract features and functionality, not state. Availability impact directly affecting Deposit or Yield is excluded.
Deposit (D):
Measures the impact to the deposits made to the contract by either users or owners.
Yield (Y):
Measures the impact to the yield generated by the contract for either users or owners.
Metrics:
IMPACT METRIC (mIm_I)METRIC VALUENUMERICAL VALUE
Confidentiality (C)None (I:N)
Low (I:L)
Medium (I:M)
High (I:H)
Critical (I:C)
0
0.25
0.5
0.75
1
Integrity (I)None (I:N)
Low (I:L)
Medium (I:M)
High (I:H)
Critical (I:C)
0
0.25
0.5
0.75
1
Availability (A)None (A:N)
Low (A:L)
Medium (A:M)
High (A:H)
Critical (A:C)
0
0.25
0.5
0.75
1
Deposit (D)None (D:N)
Low (D:L)
Medium (D:M)
High (D:H)
Critical (D:C)
0
0.25
0.5
0.75
1
Yield (Y)None (Y:N)
Low (Y:L)
Medium (Y:M)
High (Y:H)
Critical (Y:C)
0
0.25
0.5
0.75
1
Impact II is calculated using the following formula:

I=max(mI)+mImax(mI)4I = max(m_I) + \frac{\sum{m_I} - max(m_I)}{4}

5.3 SEVERITY COEFFICIENT

Reversibility (R):
Describes the share of the exploited vulnerability effects that can be reversed. For upgradeable contracts, assume the contract private key is available.
Scope (S):
Captures whether a vulnerability in one vulnerable contract impacts resources in other contracts.
Metrics:
SEVERITY COEFFICIENT (CC)COEFFICIENT VALUENUMERICAL VALUE
Reversibility (rr)None (R:N)
Partial (R:P)
Full (R:F)
1
0.5
0.25
Scope (ss)Changed (S:C)
Unchanged (S:U)
1.25
1
Severity Coefficient CC is obtained by the following product:

C=rsC = rs

The Vulnerability Severity Score SS is obtained by:

S=min(10,EIC10)S = min(10, EIC * 10)

The score is rounded up to 1 decimal places.
SeverityScore Value Range
Critical9 - 10
High7 - 8.9
Medium4.5 - 6.9
Low2 - 4.4
Informational0 - 1.9

6. SCOPE

Files and Repository
(a) Repository: huma-solana-programs
(c) Items in scope:
  • permissionless/src/mode/instructions/cancel_redemption_request.rs
  • Only changes related with disallowing a lender to cancel a redemption request are considered in scope. Every other change in this commit is not in scope
Out-of-Scope: Third party dependencies and economic attacks.
Files and Repository
(a) Repository: huma-solana-programs
(c) Items in scope:
  • programs/permissionless/src/mode/instructions/add_redemption_request.rs
  • programs/permissionless/src/pool/instructions/process_redemption_request.rs
Out-of-Scope: Third party dependencies and economic attacks.
Files and Repository
(a) Repository: huma-solana-programs
(b) Assessed Commit ID: c19bddf
(c) Items in scope:
  • programs/permissionless/src/common/common_utils.rs
  • programs/permissionless/src/common/constants.rs
  • programs/permissionless/src/common/errors.rs
↓ Expand ↓
Out-of-Scope: programs/permissionless/src/mode/instructions/cancel_redemption_request.rs, programs/permissionless/src/mode/instructions/add_redemption_request.rs, programs/permissionless/src/pool/instructions/process_redemption_request.rs, third party dependencies and economic attacks.
Remediation Commit ID:
Out-of-Scope: New features/implementations after the remediation commit IDs.

7. Assessment Summary & Findings Overview

Critical

0

High

0

Medium

0

Low

2

Informational

2

Security analysisRisk levelRemediation Date
Lack of a Direct Entry Point for Refreshing Pool Mode AssetsLowSolved - 03/18/2025
Duplicate ModeConfig Accounts Can Cause Incorrect Yield Calculations in Pre-ClosureLowSolved - 03/15/2025
Lack of Two-Step Ownership Transfer in Pool Ownership and Treasury UpdatesInformationalAcknowledged - 03/25/2025
Improving Readability in Mode Addition LogicInformationalAcknowledged - 03/25/2025

8. Findings & Tech Details

8.1 Lack of a Direct Entry Point for Refreshing Pool Mode Assets

//

Low

Description

In the Huma Finance Permissionless Program, pool assets need to be updated periodically to ensure that the system correctly reflects accrued interest and liquidity changes. The function refresh_assets_for_all_modes is responsible for updating the asset values across different lending modes within a pool.


Currently, refresh_assets_for_all_modes is not publicly accessible and can only be invoked indirectly through the following entry points:

enter_pre_closure

distribute_loss

pay_back_liquidity


Since these instructions are tied to specific financial events, they do not provide a way to refresh pool assets on demand. This limitation makes it impossible to update pool values at regular intervals unless these financial operations are triggered.


The absence of a dedicated instruction to refresh pool assets means that asset values may become outdated when no qualifying financial event occurs. As shown in the snippet below, refresh_assets_for_all_modes is a private function and is not exposed as a standalone entry point.


programs/permissionless/src/pool/accounts.rs

fn refresh_assets_for_all_modes(&mut self, mode_yield_bps: &[f64]) -> Result<()> {
    assert!(!self.is_pool_closed());

    if !self.is_pool_in_pre_closure() {
        // Refresh assets if the pool is not in pre-closure.
        for (mode_state, yield_bps) in self
            .mode_states
            .iter_mut()
            .zip(mode_yield_bps.iter().cloned())
        {
            mode_state.refresh_assets(yield_bps)?;
        }
    }

    Ok(())
}

This issue impacts price oracles that rely on updated asset values. Since updates occur only when specific instructions are executed, the mode token price could become stale over time. Price staleness affects liquidity pools on decentralized exchanges (DEXes), where accurate pricing is critical.

BVSS
Recommendation

A permissionless entry point should be introduced to allow direct execution of refresh_assets_for_all_modes. This would enable external actors, such as a cron job or off-chain service, to call the function at regular intervals, ensuring that pool assets remain up-to-date.

Remediation Comment

SOLVED: The Huma Finance team solved the issue by implementing a permissionless entry point that allows refreshing the assets for all the modes


Remediation Hash
References
This issue was identified in the c19bddf commit hash.

8.2 Duplicate ModeConfig Accounts Can Cause Incorrect Yield Calculations in Pre-Closure

//

Low

Description

In the Huma Finance Permissionless Program, lending pools support multiple modes, allowing lenders to participate under different configurations. Each mode is represented by a ModeConfig account, which defines specific lending parameters, such as periodic yield rates.


The function get_mode_yield_bps extracts yield basis points (bps) from these ModeConfig accounts and assigns them to a vector, mode_yield_bps, which is later used in critical pool operations. However, the function does not check for duplicate ModeConfig accounts, which can lead to incorrect yield calculations.


programs/permissionless/src/pool/pool_utils.rs

pub(crate) fn get_mode_yield_bps(
    remaining_account_infos: &[AccountInfo],
    pool_state: &PoolState,
    pool_config_key: Pubkey,
    program_id: &Pubkey,
) -> Result<Vec<f64>> {
    let num_mode = pool_state.mode_config_keys.len();
    require!(
        remaining_account_infos.len() >= num_mode,
        Error::TooFewModes
    );

    let mode_config_account_infos = &remaining_account_infos[0..num_mode];
    let mut mode_yield_bps: Vec<f64> = vec![0.0; num_mode];

    for account_info in mode_config_account_infos.iter() {
        let (mode_config, mode_index) = preconditions::only_valid_mode_config(
            pool_state,
            pool_config_key,
            account_info,
            program_id,
        )?;
        mode_yield_bps[mode_index] = mode_config.periodic_apy_bps;
    }

    Ok(mode_yield_bps)
}

A duplicated ModeConfig results in one of the mode_yield_bps entries being 0.0, leading to incorrect asset and cumulative yield calculations in pre-closure mode.


When mode_yield_bps is passed to enter_pre_closure, the function refresh_assets_for_all_modes uses the incorrect values, preventing some assets from being refreshed.


programs/permissionless/src/pool/instructions/enter_pre_closure.rs

pub fn enter_pre_closure(&mut self, mode_yield_bps: &[f64]) -> Result<()> {
    self.refresh_assets_for_all_modes(mode_yield_bps)?;
    self.status = PoolStatus::PreClosure;

    Ok(())
}

Since mode_yield_bps is incorrect, any mode with a yield of 0.0 will not refresh its assets correctly in refresh_assets_for_all_modes.

programs/permissionless/src/pool/accounts.rs

fn refresh_assets_for_all_modes(&mut self, mode_yield_bps: &[f64]) -> Result<()> {
    assert!(!self.is_pool_closed());

    if !self.is_pool_in_pre_closure() {
        for (mode_state, yield_bps) in self
            .mode_states
            .iter_mut()
            .zip(mode_yield_bps.iter().cloned())
        {
            mode_state.refresh_assets(yield_bps)?;
        }
    }

    Ok(())
}

Finally, refresh_assets depends on a correct yield_bps value to update self.assets and self.cumulative_yields. Since a 0.0 yield prevents updates, cumulative_yields will remain incorrect permanently.


programs/permissionless/src/pool/accounts.rs

pub fn refresh_assets(&mut self, yield_bps: f64) -> Result<u128> {
    let current_ts = Clock::get()?.unix_timestamp as u64;
    let current_assets = self.assets;
    if self.assets_refreshed_at == 0 {
        self.assets_refreshed_at = current_ts;
        return Ok(current_assets);
    }

    let seconds_elapsed = current_ts - self.assets_refreshed_at;
    if seconds_elapsed > 0 {
        if yield_bps > 0.0 {
            let yield_rate_per_second =
                yield_bps / (SECONDS_IN_A_YEAR * HUNDRED_PERCENT_BPS as u64) as f64;
            self.assets = (current_assets as f64
                * (1_f64 + yield_rate_per_second).powi(seconds_elapsed as i32))
                as u128;
            self.cumulative_yields += self.assets - current_assets;
        }
        self.assets_refreshed_at += seconds_elapsed;
    }

    Ok(self.assets)
}

The risk of this issue is relatively low, as self.assets is updated in multiple parts of the code, ensuring that the total asset calculations remain mostly accurate.


However, self.cumulative_yields will not be updated once the pool enters pre-closure, leading to inconsistencies in yield tracking. Since self.cumulative_yields is used exclusively for calculating the cumulative yield of each mode, the affected mode’s cumulative yield will be incorrectly updated, potentially leading to discrepancies in reported earnings. While this does not immediately impact functionality, it could result in data inconsistency, especially in historical yield tracking and analytics.

BVSS
Recommendation

To prevent duplicate ModeConfig accounts, update get_mode_yield_bps to enforce uniqueness before processing them.

Remediation Comment

SOLVED: The Huma Finance team solved the issue by modifying get_mode_yield_bps.

A boolean array has been introduced, initialized to false, to track processed modes. During iteration, the corresponding index is set to true for each mode. After the loop completes, a validation check ensures all expected entries are set, preventing duplicate modes and ensuring proper updates.

Remediation Hash
References
This issue was identified in the c19bddf commit hash.

8.3 Lack of Two-Step Ownership Transfer in Pool Ownership and Treasury Updates

//

Informational

Description

In the Huma Finance Permissionless program, change_pool_owner and set_pool_owner_treasury allow modifying critical ownership parameters of a pool. change_pool_owner updates the owner of a pool, while set_pool_owner_treasury updates the address of the treasury that manages the pool’s funds.


Ownership transitions should follow a two-step process: the current owner initiates the transfer, and the new owner explicitly accepts it. This prevents unintended transfers due to compromised keys or administrative errors.


However, as shown in the snippet below, both functions directly assign the new owner or treasury without requiring confirmation from the recipient.


permissionless/src/pool/instructions/update_pool_config.rs

pub(crate) fn change_pool_owner(ctx: Context<UpdatePoolConfig>, new_owner: Pubkey) -> Result<()> {
    let huma_config = preconditions::only_valid_huma_config(ctx.accounts.huma_config.as_ref())?;
    preconditions::only_pool_owner_or_huma_owner(
        &huma_config,
        ctx.accounts.pool_config.as_ref(),
        &ctx.accounts.signer,
    )?;

    let pool_state = ctx.accounts.pool_state.as_ref();
    pool_state.require_pool_not_closed()?;

    let pool_config = ctx.accounts.pool_config.as_mut();
    pool_config.pool_owner = new_owner; // Direct assignment without confirmation

    emit!(PoolOwnerChangedEvent {
        pool_id: pool_config.pool_id,
        owner: pool_config.pool_owner,
    });

    Ok(())
}

permissionless/src/pool/instructions/update_pool_config.rs

pub(crate) fn set_pool_owner_treasury(
    ctx: Context<UpdatePoolConfig>,
    new_treasury: Pubkey,
) -> Result<()> {
    let huma_config = preconditions::only_valid_huma_config(ctx.accounts.huma_config.as_ref())?;
    preconditions::only_pool_owner_or_huma_owner(
        &huma_config,
        ctx.accounts.pool_config.as_ref(),
        &ctx.accounts.signer,
    )?;

    let pool_state = ctx.accounts.pool_state.as_ref();
    pool_state.require_pool_not_closed()?;

    let pool_config = ctx.accounts.pool_config.as_mut();
    pool_config.pool_owner_treasury = new_treasury; // Direct assignment without confirmation

    emit!(PoolOwnerTreasuryChangedEvent {
        pool_id: pool_config.pool_id,
        old_treasury: ctx.accounts.pool_config.pool_owner_treasury,
        new_treasury,
    });

    Ok(())
}

The lack of a two-step ownership transfer mechanism poses a security risk, as the pool ownership or treasury can be reassigned without the new owner’s explicit approval. If an incorrect or malicious address is provided, recovery would be impossible, leading to governance and financial risks.


While only the current pool owner or Huma owner can make these changes, enforcing explicit acceptance by the new owner or treasury manager would mitigate the risk of mistaken or unauthorized transfers.

BVSS
Recommendation

Implement a two-step ownership transfer process by introducing a pending owner field that requires confirmation from the new owner before finalizing the transfer.

programs/permissionless/src/pool/instructions/update_pool_config.rs

pub(crate) fn initiate_pool_owner_transfer(
    ctx: Context<UpdatePoolConfig>,
    pending_owner: Pubkey,
) -> Result<()> {
    let huma_config = preconditions::only_valid_huma_config(ctx.accounts.huma_config.as_ref())?;
    preconditions::only_pool_owner_or_huma_owner(
        &huma_config,
        ctx.accounts.pool_config.as_ref(),
        &ctx.accounts.signer,
    )?;

    let pool_state = ctx.accounts.pool_state.as_ref();
    pool_state.require_pool_not_closed()?;

    let pool_config = ctx.accounts.pool_config.as_mut();
    pool_config.pending_owner = Some(pending_owner); // Store pending owner

    Ok(())
}

programs/permissionless/src/pool/instructions/update_pool_config.rs

pub(crate) fn accept_pool_owner_transfer(ctx: Context<UpdatePoolConfig>) -> Result<()> {
    let pool_config = ctx.accounts.pool_config.as_mut();
    
    require!(
        pool_config.pending_owner == Some(ctx.accounts.signer.key()),
        Error::UnauthorizedTransferAcceptance
    );

    pool_config.pool_owner = pool_config.pending_owner.unwrap();
    pool_config.pending_owner = None; // Clear pending owner

    emit!(PoolOwnerChangedEvent {
        pool_id: pool_config.pool_id,
        owner: pool_config.pool_owner,
    });

    Ok(())
}

By implementing a pending ownership field and requiring explicit confirmation, unauthorized or accidental transfers can be prevented, improving security and governance over the pool.

Remediation Comment

ACKNOWLEDGED: The Huma Finance team acknowledged this finding due to multisig controls for admin accounts will be employed —including the pool owner— and timelocks on all instructions that update the pool configuration will be implemented. This approach ensures that every change, including owner modifications, is reviewed by at least two people before execution, with additional scrutiny likely during the timelock period.


The Huma Finance team believes these measures are sufficient to prevent the accidental changes described in the issue, rendering a two-step implementation unnecessary.

References
This issue was identified in the c19bddf commit hash.

8.4 Improving Readability in Mode Addition Logic

//

Informational

Description

In the Huma Finance Permissionless program, the add_mode function is responsible for adding a new mode to a pool. Modes represent different lending mechanisms that allow lenders to participate under varying configurations. When a new mode is added, its corresponding configuration is stored in mode_config_keys, and a new ModeState is appended to mode_states.

Currently, the function validates that the number of modes does not exceed the maximum allowed by checking if self.mode_config_keys.len() is strictly less than MAX_NUM_MODES.


As shown in the snippet below, the check uses < rather than <= to ensure that the incremented length does not exceed the limit:

programs/permissionless/src/pool/accounts.rs

    pub fn add_mode(&mut self, mode_config_key: Pubkey) -> Result<()> {
        require!(
            !self.mode_config_keys.contains(&mode_config_key),
            Error::DifferentModesRequired
        );
        require!(
            self.mode_config_keys.len() < MAX_NUM_MODES,
            Error::TooManyModes
        );

        self.mode_config_keys.push(mode_config_key);
        self.mode_states.push(ModeState::default());

        Ok(())
    }

Since the check is performed before updating the vector, a more readable approach would be to first append the mode and then validate the length, ensuring clarity in the logic.
This issue does not introduce a functional bug but affects code readability and maintainability. The current approach can lead to confusion for developers unfamiliar with the logic, increasing the risk of misinterpretation when making modifications. Improving readability by restructuring the validation would reduce potential misunderstandings and improve maintainability.

BVSS
Recommendation

Rearrange the logic so that the mode is added first, and then the length validation occurs, improving the readability of the function.


programs/permissionless/src/pool/accounts.rs

    pub fn add_mode(&mut self, mode_config_key: Pubkey) -> Result<()> {
        require!(
            !self.mode_config_keys.contains(&mode_config_key),
            Error::DifferentModesRequired
        );

        self.mode_config_keys.push(mode_config_key);
        self.mode_states.push(ModeState::default());

        require!(
            self.mode_config_keys.len() <= MAX_NUM_MODES,
            Error::TooManyModes
        );

        Ok(())
    }

Remediation Comment

ACKNOWLEDGED: The Huma Finance team acknowledged this finding.

References
This issue was identified in the c19bddf commit hash.

9. Automated Testing

Description

Halborn used automated security scanners to assist with the detection of well-known security issues and vulnerabilities. Among the tools used was cargo-audit, a security scanner for vulnerabilities reported to the RustSec Advisory Database. All vulnerabilities published in https://crates.io are stored in a repository named The RustSec Advisory Database. cargo audit is a human-readable version of the advisory database which performs a scanning on Cargo.lock. Security Detections are only in scope. All vulnerabilities shown here were already disclosed in the above report. However, to better assist the developers maintaining this code, the reviewers are including the output with the dependencies tree, and this is included in the cargo audit output to better know the dependencies affected by unmaintained and vulnerable crates.

Results

ID

package

Short Description

RUSTSEC-2024-0344

curve25519-dalek

Timing variability in curve25519-dalek's Scalar29::sub/Scalar52::sub

RUSTSEC-2022-0093

ed25519-dalek

Double Public Key Signing Function Oracle Attack on ed25519-dalek


Halborn strongly recommends conducting a follow-up assessment of the project either within six months or immediately following any material changes to the codebase, whichever comes first. This approach is crucial for maintaining the project’s integrity and addressing potential vulnerabilities introduced by code modifications.

© Halborn 2025. All rights reserved.