hypnguyen1209/offensive-claude / skills/vulnerability-analysis
vulnerability-analysis
UnverifiedUse when auditing source code for vulnerabilities — drive CodeQL/Semgrep/Joern to taint untrusted data source-to-sink across injection, memory safety, deserialization/prototype-pollution, secrets/crypto/authz/race, and supply-chain risks
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 11.4 KB |
| references/deserialization-prototype-pollution.md | markdown | 8.3 KB |
| references/injection-source-patterns.md | markdown | 8.6 KB |
| references/memory-safety-c-cpp.md | markdown | 7.7 KB |
| references/secrets-crypto-authz-concurrency.md | markdown | 8.6 KB |
| references/supply-chain-dependency-audit.md | markdown | 7.3 KB |
| references/taint-engines-static-analysis.md | markdown | 11.0 KB |
| references/variant-hunting.md | markdown | 3.4 KB |
| scripts/dep_audit.py | script | 10.9 KB |
| scripts/deser_gadget_scan.py | script | 7.7 KB |
| scripts/evidence_kit.py | script | 13.0 KB |
| scripts/finding_dedup.py | script | 4.2 KB |
| scripts/joern_taint.sc | other | 3.6 KB |
| scripts/merge_runtime_evidence.py | script | 6.1 KB |
| scripts/path_conditions.py | script | 10.6 KB |
| scripts/provenance.py | script | 2.8 KB |
| scripts/sast_runner.py | script | 7.8 KB |
| scripts/secret_crypto_audit.py | script | 8.1 KB |
| scripts/taint_trace.py | script | 14.9 KB |
| scripts/validate_findings.py | script | 22.8 KB |
| scripts/variant_hunt.py | script | 11.8 KB |
| scripts/vocab_gate.py | script | 7.5 KB |
references/memory-safety-c-cpp.md
View on GitHub# Memory-Safety Auditing (C / C++ / unsafe Rust)
Cluster: spotting memory-corruption primitives in native source — buffer
overflow (CWE-787/125), integer overflow→undersized alloc (CWE-190), use-after-
free / double-free (CWE-416/415), and missing NULL checks (CWE-476). The aim is
to find the *primitive* and assess exploitability; turning it into a working
exploit is `exploit-development`. Backing tools: `scripts/joern_taint.sc`,
`scripts/taint_trace.py` (sink census).
## Theory / Mechanism
Native code has no runtime bounds checking. A vulnerability arises when a length,
index, or pointer derived from untrusted input is used in a memory operation
without validation. The recurring root causes, all live in 2024-2026 CVEs:
1. **Integer overflow in size calculation → undersized buffer.** `n * sizeof(x)`
wraps, `malloc` succeeds small, the subsequent write overflows. (CWE-190 →
CWE-787.)
2. **Use-after-free.** A pointer is freed on one path, dereferenced on another
(often during error handling or in object trees). (CWE-416.)
3. **Off-by-one / unbounded copy.** `strcpy`/`memcpy`/`sprintf` with a length
that exceeds the destination. (CWE-787/120.)
4. **NULL deref.** A fallible allocator/lookup return used without a check.
(CWE-476.)
## Audit Patterns
### Integer overflow → heap overflow (the libxml2 CVE-2025-6021 shape)
```c
// VULN: integer overflow in buffer-size arithmetic, then write past end.
size_t len = lenA + lenB + 1; // can wrap if lenA/lenB attacker-influenced
char *buf = (char*)xmlMalloc(len); // allocates a too-small buffer on wrap
memcpy(buf, a, lenA); memcpy(buf+lenA, b, lenB); // overflow
```
This is exactly **CVE-2025-6021** in `xmlBuildQName`: *integer overflows in buffer
size calculations can lead to a stack-based buffer overflow* (CWE-190, CVSS 7.5).
Audit rule: any `malloc(a * b)` / `a + b + k` feeding an allocation where the
operands trace to input and there is no overflow guard
(`if (a > SIZE_MAX / b) ...` or use of `reallocarray`/`__builtin_mul_overflow`).
### Unbounded / mismatched copy
```c
char dst[64];
strcpy(dst, src); // VULN if strlen(src) >= 64 (CWE-787)
sprintf(dst, "%s", user); // VULN
memcpy(dst, src, hdr->len); // VULN if hdr->len attacker-controlled, no clamp
// SAFE: snprintf(dst, sizeof dst, "%s", user); + explicit length clamp on memcpy
```
### Use-after-free (the libxml2 schema/XPath CVE shape)
```c
node = lookup(doc);
xmlFree(node); // freed on one path...
...
process(node->children); // ...dereferenced later (CWE-416)
```
**CVE-2024-56171** (UAF in `xmlSchemaIDCFillNodeTables`/`xmlSchemaBubbleIDCNodeTables`,
CVSS 7.8) and **CVE-2025-49794** (UAF parsing XPath under schematron, CWE-825,
CVSS 9.1) are this class — both cluster in `xmlschemas.c`. Audit rule: for each
`free(p)`/`xmlFree(p)`, check every subsequent use of `p` on reachable paths;
flag any deref without an intervening `p = NULL` or reassignment.
### Missing NULL / under-read
```c
p = malloc(n); memcpy(p, src, n); // VULN: malloc may return NULL (CWE-476)
// SAFE: if (!p) return -ENOMEM;
```
**CVE-2025-27113** (NULL deref in libxml2 `xmlPatMatch`) and **CVE-2025-32415**
(heap under-read in `xmlSchemaIDCFillNodeTables`) round out the recent cluster.
## Find it with the engines
CodeQL ships strong C/C++ memory queries; Joern shines where the project doesn't
build cleanly. The sink census + Joern taint script handle both:
```bash
# Sink census of dangerous native APIs (then manual review):
grep -rnE '\b(strcpy|strcat|sprintf|gets|memcpy|memmove|alloca)\b' --include=*.c --include=*.cpp ./src
# Joern: untrusted length reaching a write sink
joern-parse ./src --output cpg.bin
joern --script scripts/joern_taint.sc --param cpgPath=cpg.bin --param outPath=joern.json
```
CodeQL for the integer-overflow-to-alloc class:
```ql
import cpp
import semmle.code.cpp.dataflow.new.TaintTracking
// Sketch: arithmetic on a tainted operand flowing into an allocation size arg.
from FunctionCall alloc, Expr size
where alloc.getTarget().hasName(["malloc","xmlMalloc","calloc"])
and size = alloc.getArgument(0)
and size.getAChild*() instanceof MulExpr
select alloc, "Allocation size from multiplication; check for integer overflow."
```
Joern UAF pass (free then later use of the same symbol):
```scala
importCpg("cpg.bin")
cpg.call.name("free|xmlFree").argument(1).code.l // candidate freed pointers; cross-ref later uses
```
### Dynamic confirmation (when you can build/run)
```bash
# Compile with sanitizers and replay the suspected input.
clang -fsanitize=address,undefined -g harness.c -o h && ./h crash.bin # ASan/UBSan
# Fuzz the parsing entrypoint to confirm reachability + get a crashing input.
clang -fsanitize=address,fuzzer harness.c -o fuzz && ./fuzz -max_len=4096 corpus/
```
ASan turns a UAF/overflow from "looks suspicious" into a confirmed, attributed
crash with a stack trace — the gold standard for the exploitability gate.
## Modern 2024-2026 Variants
- **libxml2 memory-safety cluster** (the canonical teaching set this year):
CVE-2025-6021 (int overflow→stack overflow, `xmlBuildQName`),
CVE-2024-56171 & CVE-2025-49794 (UAF in schema/XPath),
CVE-2025-24928 (stack overflow in `xmlSnprintfElements`, DTD validation),
CVE-2025-12863 (UAF in `xmlSetTreeDoc`), CVE-2025-27113 (NULL deref),
CVE-2025-32415 (heap under-read). Fixes across 2.12.10 / 2.13.6–8 / 2.14.x.
- **CPG+LLM discovery**: *codebadger* (Joern MCP) reported finding an unreported
libtiff buffer overflow and auto-patching the libxml2 integer overflow
(CVE-2025-6021) — the 2025 frontier for native bug hunting at scale.
- **`unsafe` Rust** is the new native surface: audit `unsafe { ... }` blocks for
`from_raw_parts`, `get_unchecked`, manual `ptr::copy`, and FFI boundaries where
Rust's guarantees don't hold. Pattern is identical (length/pointer from input
into an unchecked memory op).
## Detection
Memory-corruption *exploitation* telemetry (runtime, blue-team pairing):
```yaml
# Sigma: service crash / abnormal exit indicative of memory corruption attempt
title: Native Service Crash (Possible Memory-Corruption Exploit)
logsource: { product: linux, service: auditd }
detection:
sel:
type: ANOM_ABEND # kernel abnormal-end / SIGSEGV/SIGABRT
condition: sel
level: medium
```
Other signals: repeated SIGSEGV/SIGABRT in service logs, ASan/UBSan reports in CI
fuzzing, glibc `malloc(): corrupted`/`free(): double free` messages, EDR alerts
on `W^X` violation or new RWX regions in the target process. Prevention controls
that change exploitability: ASLR/PIE, stack canaries, FORTIFY_SOURCE, CFI/CET —
note their presence per binary in the finding (it moves CVSS).
## OPSEC
- Pure source/CPG review is silent. The moment you *fuzz a live service* you
generate crash dumps, core files, and log noise — fuzz a local copy in a
container, not production.
- Cores/ASan logs contain memory contents (possibly secrets/PII) — store in the
engagement workspace, purge on teardown.
- A crash is not yet a finding: distinguish a DoS-only NULL deref from a
controllable write primitive when scoring; note mitigations present.
## References
- libxml2 CVE-2025-6021 (CWE-190): https://www.cvedetails.com/cve/CVE-2025-6021/
- libxml2 CVE-2025-12863 (UAF) analysis: https://zeropath.com/blog/cve-2025-12863-libxml2-use-after-free-summary
- libxml2 CVE-2025-49794 (UAF, CWE-825): https://www.sentinelone.com/vulnerability-database/cve-2025-49794/
- IBM advisory bundling CVE-2024-56171 / CVE-2025-24928 / CVE-2025-27113 / CVE-2025-32415:
https://www.ibm.com/support/pages/security-bulletin-vulnerabilities-libxml2-library-cve-2024-56171-cve-2025-24928-affect-power-hmc
- Joern CPG memory analysis: https://docs.joern.io/code-property-graph/
- CWE-787 / CWE-190 / CWE-416 / CWE-476 (MITRE CWE database).
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |