YEAR-END REPORT

Solana Security Ecosystem Review

2025

At a Glance

Exploits

Fewer on-chain smart contract Solana exploits in 2025 (US$ 8mn) down from peak in 2022 (US$ 550mn), even as activity and TVL grew

Audits Analyzed

163 Solana security reviews examined, spanning 1,669 recorded vulnerabilities

Finding Density

Average of 10 issues per audit, with 1.4 High or Critical vulnerabilities in each review

Vulnerability Themes

The most severe issues concentrate in business logic flaws, access control failures, and protocol design weaknesses

A Quick Note

We spend most of the year reviewing Solana programs one code base at a time. Across that work, patterns emerge - how bugs cluster, which security practices help, and where things can still go wrong even after an audit - but those patterns rarely live in one place.

This report is our attempt to put structure around what we are seeing: to quantify what audits actually uncover on Solana, place that evidence next to publicly reported incidents on-chain, and connect both to the design and operational choices development teams were making.

The aim is to give builders, reviewers and ecosystem participants a shared reference point for how Solana security is evolving and where attention is likely to matter most in the next year.

Security Reviews

Results

We analyzed 163 Solana security audits drawn from a mix of publicly released reports and anonymized Sec3 review engagements. Together these reviews produced 1,733 findings, of which 1,669 qualified as vulnerability-level issues.

10.3
Findings Per Review
7
Median Findings
1 to 112
Range Per Review

99.4% of audits identified vulnerabilities

162 of 163 reviews identified at least one vulnerability

Severity Distribution

Distribution of 1,669 Vulnerabilities

Informational33.9%
Low32.2%
Medium20.2%
High8.4%
Critical5.3%
76%Medium+

of reviews contained at least one medium-or-higher issue

51%High+

of reviews contained at least one high-or-critical issue

23%Critical+

of reviews contained at least one critical issue

What Vulnerabilities Dominate

Among findings with clear classifications (approximately 70% of the total dataset):

All Classified Findings

Business Logic
38.5%
Input Validation & Data Hygiene
25%
Access Control & Authorization
19%
Data Integrity & Arithmetic
8.9%
Denial of Service & Liveness
8.5%

High + Critical Only

Business Logic
36.9%
Input Validation & Data Hygiene
27.9%
Access Control & Authorization
20.7%
Data Integrity & Arithmetic
8.9%
Denial of Service & Liveness
5.6%

Serious issues overwhelmingly stem from Business Logic, Permissions, and Validation Errors rather than low level arithmetic or liveness problems. Top 3 categories increased from 82.5% → 85.5% of all severe findings.

Top Vulnerability Patterns

We further clustered findings into specific vulnerability types. Excluding the generic "Other Mixed" bucket, the 10 most frequent patterns across all severities:

#VulnerabilityFindingsDetails
1Business Logic or Economic Flaws413Incorrect fee/reward accrual, misconfigured liquidation rules, accounting drift over time
2Input Validation and Sanitation Issues326Unbounded user parameters, missing oracle freshness checks, unverified account ownership
3Arithmetic & Rounding Errors93Fixed-point math errors, exploitable rounding at scale, release-build calculation discrepancies
4Programming Errors80Incorrect control flow, module assumption mismatches, incomplete state transitions
5Denial of Service or Resource Exhaustion79Unbounded loops/structures allowing single-account progress blocking
6Access Control Issues (General)67Missing or incomplete permission checks
7Missing Account Owner Checks61Failure to verify account ownership before state modifications
8Missing Signer or Authority Checks36Operations executable without proper authorization
9PDA Derivation Errors32Incorrect program-derived address computation or validation
10Lamport and Rent Handling Issues29Improper balance management or rent exemption handling

Among High-and-Critical issues, Access Control Problems dominate. Grouping "Access control (general)", "Missing signer/authority check", "Missing account owner check", "PDA derivation issues", and other access control related patterns together, they account for over 20% of all high-and-critical findings.

Snapshot of Solana in 2025

DeFi TVL

$9.3B USDas of Nov 2025

Source: defillama.com/chain/solana

Stablecoin TVL

$15.0B USDas of Nov 2025

Source: defillama.com/stablecoins/solana

Higher TVL and accelerating stablecoin growth indicate an ecosystem supporting more applications, more users, and more complex financial activity. As value concentration and protocol complexity increase, the security surface expands. This broader context makes the decline in large scale exploits particularly notable and strengthens the case that security practices across the ecosystem have improved.

How Framework Choices Shape Vulnerability Patterns

Most Solana programs use Rust SDK, Anchor, or Pinocchio. Each approach changes where validation happens and which bugs are most common.

FrameworkGuardrailTypical Vulnerability ProfileProsConsBest Used When
Rust SDKNone by defaultMissing checks, fragile dispatch, incorrect PDA derivationsFull control, maximum flexibilityEasy to miss basics, lots of boilerplateYou need total customization
AnchorBuilt into the frameworkProtocol logic issues, edge cases, incentive designSafe defaults, standardized account validationMore abstraction, larger binariesMost new protocols, standard workflows
PinocchioMust be rebuilt manuallyMissing owner/signer checks, offset or aliasing errorsFastest, most compact, lowest compute unitManual validation everywhere, unsafe if rushedHighly optimized systems with disciplined engineering

The frameworks aren't inherently safe or unsafe. The difference is simply where validation happens: in the framework (Anchor) or in the team's engineering discipline (Pinocchio).

Practical Guide

Before the Audit

How Strong Teams Prepare

Plan the audit

  • Engage auditors early and lock dates
  • Define scope: programs that hold value
  • Create dedicated audit branch
  • Budget time post review for fixes

Make the system easy to understand

  • Overview: purpose, assets, users
  • Roles & permissions table
  • Account & PDA map
  • Core flow diagrams

Ship reviewable code and tests

  • Clear structure, no dead code
  • Inline comments for critical logic
  • High coverage unit & integration tests
  • Document known limitations

Run automated tools first

  • Linters & static analyzers
  • Formal verification when relevant
  • Security scanners (Soteria, etc.)
  • Fix obvious findings before kickoff

During the Review

Collaboration & Communication

Stay engaged

  • Rapid response on auditor questions
  • Clarify intent vs. actual behavior
  • Acknowledge & track all findings
  • Don't assume; confirm understanding

Prioritize findings

  • High/Critical first, info later
  • Discuss severity if you disagree
  • Validate exploit paths together
  • Document mitigation strategy

Track changes properly

  • Keep fixes on audit branch
  • Tag each commit to a finding
  • Request re-review for complex fixes
  • Don't deploy until cleared

Where Risk Is Moving

Fewer failures come from simple bugs in isolated programs - more risk now sits in how systems, automation, and people interact

Stacked systems

Restaking, leverage, looping, multi-protocol strategies. Shared collateral creates correlated failures. Need stack level stress tests.

Yield bearing stablecoins

Supply has jumped into low double digit billions. Risk is in the portfolio behind each dollar, not just the peg.

Vibe coded contracts

AI generated code shipped with thin specs. Forks tweak logic assuming changes are safe, breaking critical guardrails.

Agents and prompt injection

Agents act on natural language with keys and payment rails. Malicious prompts can steer them to misuse safe contracts.

Wallets and the human layer

Wallet drains and social engineering drive most losses. A single permit can empty accounts even when contracts are clean.

The Security at Scale

Independent security reviews of Solana programs consistently identify vulnerabilities - primarily in business logic, input validation, and access control. Yet on-chain exploits declined 98% from their 2022 peak in 2025.

Multiple factors likely contributed: adoption of frameworks like Anchor that provide built-in security guardrails, more experienced teams, better testing practices, and widespread adoption of pre-launch security review. As the ecosystem continues to grow in complexity, sustained attention to security - across tooling, testing, and review processes - will remain important.

Dataset Scope and Constraints

This dataset is informative but not exhaustive. Key constraints:

Audits do not cover all deployed code

Findings represent only projects that engaged auditors and agreed to publish or share results. Unaudited code, informal reviews, and unpublished work fall outside the sample. Read these numbers as "what auditors typically find when they look," not "what exists on-chain overall"

Automated classification is imperfect

Despite LLM assistance and manual sampling, some findings inevitably land in the wrong specific type. We maintain a large "Other/Mixed" bucket and avoid over-interpreting rare categories

Severity is defined at report time

Severities come from original reviewers. Teams occasionally disagree on edge-case ratings, and some findings shift in importance as protocols evolve or more context emerges. We aggregate severities as reported without re-grading

Per-review statistics are averaged

A small number of very large reviews contribute disproportionately to the total finding count. While the mean is 10.3 findings per review, the median of 7 shows that most audits uncover fewer vulnerabilities. We report both counts and per-review percentages to characterize the full distribution