Skip to content

Add rules for BRATA#78

Closed
pulorsok wants to merge 1 commit into
ev-flow:masterfrom
pulorsok:add-brata-rules
Closed

Add rules for BRATA#78
pulorsok wants to merge 1 commit into
ev-flow:masterfrom
pulorsok:add-brata-rules

Conversation

@pulorsok
Copy link
Copy Markdown
Member

close #77

* Adds rule #00276 "Acquire a wake lock to keep the device awake",
  generated from analysis of BRATA samples.  The rule pairs
  Context.getSystemService() (to obtain the PowerManager) with
  PowerManager.WakeLock.acquire().  BRATA is a remote-access trojan
  whose background service holds a wake lock so the OS does not
  suspend the device while it is being remotely controlled; Quark
  observed this at Level 5 inside the malware's own service
  (e.g. Lcom/vera/GeneralService) across several samples.  Wake
  locks are a dual-use API, so after training this rule carries a
  near-zero score (-0.14) — it is a weak supporting signal, not a
  standalone indicator.
* Re-tunes the existing 275 rules' scores from the score-adjustment
  training pass run with the BRATA sample set added to the corpus
  (recall=1.0 over 432 malware samples, precision=0.77).  272 rules
  changed score; 4 were unchanged.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
@pulorsok
Copy link
Copy Markdown
Member Author

Closing — analysis showed BRATA does not need a new rule.

The candidate rule (#00276, "Acquire a wake lock to keep the device awake") provides no detection value:

  • L5 fires on 15.7% of malware vs 16.3% of benign samples (ratio 0.96:1 — benign is actually slightly higher), so it cannot discriminate malware from benign.
  • After score-adjustment training it carries a near-zero/negative score (-0.14).
  • Training already reaches recall = 432/432 (every malware sample high-risk) WITHOUT this rule contributing.

BRATA's detectable malicious behaviors are already covered by existing rules (screen capture #00210, contact reads #00187/#00212/#00215, security-software discovery #00265). The rule pool is effectively saturated for this family, so no new rule is warranted.

The BRATA analysis report (quark-engine PR #917) stands on the existing rules and is unaffected.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Add Rules to detect BRATA Family

1 participant