Skip to content
AppSec tool stack r...
 
Notifications
Clear all

AppSec tool stack recommendations from security engineers

7 Posts
7 Users
0 Reactions
82 Views
(@cloud_ops_amy)
Honorable Member
Joined: 8 months ago
Posts: 453
Topic starter   [#22016]

Hey everyone! I've been refining our AppSec pipeline at work and wanted to share what we're using, hoping to get your thoughts and recommendations too. We're a mid-sized team running mostly containerized apps on AWS with a GitHub Actions CI/CD pipeline.

Our main goals were shifting security left without drowning devs in alerts, and having clear ownership (security team writes the rules, devs fix their own code). Here's our current stack:

**Pre-Commit & SAST:**
- **Trivy** for container and filesystem vulnerability scanning. We run it in CI against our Dockerfiles and built images. The SBOM generation is a nice bonus.
- **Semgrep** for static analysis. We found it easier to write custom rules for our frameworks compared to some heavier tools. The performance is great.
- **Gitleaks** as a pre-commit hook and in CI for secret scanning. It's been incredibly effective at catching accidental commits.

**SCA & Dependencies:**
- **Dependabot** (native to GitHub) for automated dependency PRs. We've tuned it to only flag high/critical severity for certain repos.
- We evaluate **Renovate** periodically for more configurability, but Dependabot's simplicity wins for now.

**Infrastructure as Code:**
- **Checkov** for scanning our Terraform plans. It catches a lot of misconfigured S3 buckets and IAM policies early.
```hcl
# Example of a rule it caught: an S3 bucket without versioning enabled
resource "aws_s3_bucket" "example" {
bucket = "my-app-logs-bucket"
# checkov:skip=CKV_AWS_21: "Ensure all data stored in the S3 bucket have versioning enabled"
# We actually had to justify this skip in a PR comment!
}
```

**Runtime & Observability:**
- **Falco** on our EKS clusters for runtime threat detection. Alerts go to our security Slack channel.
- Centralized logging (Amazon OpenSearch) with dashboards for security events.

I'm particularly curious about:
1. **DAST tools** – we're looking at OWASP ZAP in the pipeline, but are there lighter-weight options you love?
2. **Software Bill of Materials (SBOM)** – how are you managing and distributing these beyond just generating them?
3. **Container signing** – are you using Cosign, Notary, or something else for supply chain integrity?

What's in your stack that you'd swear by, especially for cloud-native setups? Any tools you tried and abandoned?

-- Amy


Cloud cost nerd. No, I don't use Reserved Instances.


   
Quote
(@annaw)
Reputable Member
Joined: 4 months ago
Posts: 310
 

Love this breakdown, especially your point about clear ownership - security writes rules, devs fix code. That's the model that actually sticks.

We tried Renovate for a while too. The configurability is fantastic, but we found that same simplicity trap. Our devs just wanted the PR, not another config file to manage. Dependabot's "set and forget" vibe won out, even if we miss some of the grouping options.

Curious, have you hit any scaling issues with Semgrep on your larger monorepos? We saw some slowdown that forced us to get clever with path filtering.



   
ReplyQuote
(@frankd)
Reputable Member
Joined: 4 months ago
Posts: 313
 

Yeah, the configurability versus adoption trade-off is real. We've actually had the opposite experience, funnily enough. Our team really took to Renovate once we gave them a single, well-documented default config in the repo root, and locked down the major version updates to security-only in the auto-merge rules. It became truly "set and forget" for them, while giving us security folks the control we needed under the hood.

On your Semgrep scaling question, absolutely. Path filtering became mandatory. We also had to switch from running it against the whole codebase on every push to a scheduled nightly scan for the full monorepo, with PR scans only targeting changed directories. The performance hit on the CI runners was getting unsustainable. Have you looked at their new 'osemgrep' engine? We're trialing it now and the speed improvement on a first pass is noticeable, though not magic.


buyer beware, but buy smart


   
ReplyQuote
(@crm_hopper_2025_new)
Honorable Member
Joined: 5 months ago
Posts: 365
 

Solid foundation, especially the focus on ownership. You mention running Trivy against built images, but what's your threshold for failure? We set ours too aggressively at first and blocked deployments for medium-severity CVEs in dev-stage base images that were purely theoretical. Had to back it off to critical/high only for non-prod environments.

Also, be wary of Gitleaks fatigue if you have a lot of legacy code. The initial scan will light up like a Christmas tree with old AWS keys in commit history. We had to run it with a `--no-verify` flag on the first pass and just generate a report for manual cleanup, otherwise it just halted all new commits.



   
ReplyQuote
(@grafana_guy_night)
Honorable Member
Joined: 8 months ago
Posts: 427
 

Totally feel you on the Trivy threshold. I'm just starting out with it and made the same mistake - blocking my test pipeline on a medium CVE in a dev image that had no reachable code path. Felt silly. Critical/high only for non-prod seems like the sane default.

That Gitleaks tip is gold, thanks! We have a ton of old repos. Starting with `--no-verify` for a baseline report is a much gentler way to onboard everyone. Did you find a good way to track the manual cleanup, or just tickets?



   
ReplyQuote
(@budget_minded_buyer)
Reputable Member
Joined: 7 months ago
Posts: 313
 

Sane default, maybe. But wait until you price out the alternative.

That critical-only policy is a great way to rack up a massive "vulnerability debt" in your lower environments. The medium-severity CVEs you're ignoring in dev will still be there when you promote the image to staging, and then prod. Suddenly you're facing a last-minute, high-pressure fix of a library you could have updated months ago.

It's not just about the severity score, it's about the *timeline and cost* of fixing it later under duress.


always ask for a multi-year discount


   
ReplyQuote
(@amelia2)
Reputable Member
Joined: 4 months ago
Posts: 261
 

Good stack. We use a similar setup but added a policy-as-code gate with Conftest for our infrastructure security. Trivy finds the vuln, Conftest decides if it blocks the PR based on environment and severity. Stops the "too many false positives" problem cold.

For secret scanning, you mentioned pre-commit and CI. Make sure you're also scanning your container images with Trivy for any baked-in secrets. Gitleaks won't catch those.


Ship it, but test it first


   
ReplyQuote