Fathom — Specialized Cybersecurity Analysis Model

Mixtral-8x7B-Instruct-v0.1 + 10× LoRA adapters (rank=32, bf16)
Primary adapter: unified-v2 (general cybersecurity + malware analysis)
9 expert adapters for domain-specific routing (static/dynamic analysis, network, forensics, threat intel, etc.)

Fathom turns raw sandbox reports (CAPE, Joe Sandbox, etc.) into high-quality ATT&CK-mapped malware analysis. It outperforms general-purpose models on cybersecurity tasks while remaining fully open-source and runnable on a single AMD MI300X / A100 80GB.

Model Overview

Base: Mixtral-8x7B-Instruct-v0.1 (full bf16, no quantization)
Training: Direct PEFT+TRL
Adapters: 1 unified + 9 expert LoRA adapters (all rank=32, α=64)
Hardware: AMD MI300X (205.8 GB VRAM) — full bf16 training
Key Innovation: Evidence extraction layer + structured behavioral prompts → 9× improvement in real ATT&CK mapping

Designed for:

Malware analysts & threat hunters
SOC / DFIR teams
CAPE / sandbox report enrichment
Automated ATT&CK technique extraction

Benchmark Results

All results use the real Fathom pipeline ([INST] chat template + 8192 context + structured evidence from CAPE extraction layer v3). Greedy decoding, bf16.

1. General Cybersecurity Knowledge (vs. Closed & Open Models)

Benchmark	Fathom unified-v2	GPT-4 (ref)	GPT-3.5 (ref)	Base Mixtral-8x7B	Llama-2-70B (ref)
CyberMetric-80	91.25%	~87%	~67%	82.5%	~57%
MMLU Computer Security	79.0%	~82%	~65%	—	~54%
MMLU Security Studies	64.0%	~74%	~60%	—	~48%
TruthfulQA MC1	65.0%

Visual bar comparison (CyberMetric-80):

Fathom unified-v2     ████████████████████ 91.25%
GPT-4                 ██████████████████   ~87%
Base Mixtral          █████████████████    82.5%
GPT-3.5               ██████████████       ~67%
Llama-2-70B           ████████████         ~57%

2. Expert Adapter Comparison (CyberMetric-80)

Adapter	Score	Specialty
`unified-v2`	91.25%	All-domain baseline
`expert-e8-analyst`	91.25%	Analyst Q&A & reporting
`expert-e3-network`	90.00%	Network traffic / C2 analysis
`expert-e4-forensics`	90.00%	Memory & disk forensics
`expert-e6-detection`	88.75%	Detection engineering
`expert-e7-reports`	88.75%	Structured report generation
`expert-e2-dynamic`	85.00%	Behavioral / sandbox analysis
`expert-e1-static`	83.75%	Static PE + evasion detection
`expert-e9-cot`	87.50%	Chain-of-thought reasoning
`expert-e5-threatintel`	81.25%	Threat intel & actor profiling

3. Core Contribution: Real ATT&CK Mapping Accuracy

Progression table (same model weights, only input pipeline improved):

Configuration	Exact F1	Parent F1	Improvement
Raw API list (naive)	0.083	0.095	—
Structured prompt (manual)	0.370	0.429	+0.334
Real Fathom evidence layer	0.534	0.508	+0.413
Real pipeline + full context fix	0.868	0.841	+0.746

This proves the architecture (evidence extraction + structured prompts) matters more than additional fine-tuning.

4. Real Malware Analysis — CAPE Pipeline ( malscore 10/10 samples)

Sample	Family	GT T-codes	Predicted T-codes	Exact F1	Parent F1	Family ID
12	Emotet	T1012, T1071, T1071.004, T1083	T1012, T1055, T1071, T1071.004, T1083	0.889	0.857	100% conf
15	Formbook	T1012, T1055, T1071, T1071.004, T1083	T1003, T1012, T1027.002, T1055, T1059, T1071, T1071.004, T1083, T1497	0.714	0.667	85% conf
16	Dridex	T1012, T1055, T1071, T1071.004, T1083	T1012, T1055, T1071, T1071.004, T1083	1.000	1.000	68% conf
Average (n=3)				0.868	0.841	—

This is the original Run-7 breakdown. The n=12 run below re-scores samples 12/15/16 as part of its validation gate; small per-sample differences (e.g. sample 15) vs. this table reflect run-to-run generation variance, not a data conflict — both runs used the identical pipeline, prompt, and generation config.

*n=12 expansion (preliminary):* the same pipeline and generation config, run on 12 real CAPE-detonated samples across 5 malware families (Emotet, Formbook, Dridex, Conti, Agent Tesla). Exact F1 0.408, Parent F1 0.514 — lower than the n=3 headline above, as expected: those three were unusually clean samples, and a broader sample set is a more credible signal. Ground truth for 9 of the 12 samples is auto-derived from sandbox behavior and not yet hand-verified, so this number is not yet final. Full per-sample results, raw predictions, and logs: journal-artifacts/evaluation/ in the source repo.

5. Additional Benchmarks

ATT&CK Mapping MCQ (30 handcrafted questions): 80%
MMLU Machine Learning: 60%
MMLU Electrical Engineering: 64%
Rigorous ground-truth F1 (23 test cases): Exact = 0.184, Parent = 0.344 (synthetic); real CAPE = 0.841 after pipeline fixes

5. Key Discovery: Mal-API-2019 Analysis

We evaluated Fathom on the public Mal-API-2019 dataset (Catak & Yazı, arXiv:1905.01999) — 7,107 API call sequences from Cuckoo Sandbox.

Variant	Accuracy	Macro F1
Raw API sequences	12.6%	0.030
Filtered behavioral groups	10.9%	0.052

Insight:

Raw API sequences alone are insufficient for reliable family classification. The dataset contains heavy loader noise and families share nearly identical behavioral APIs. Ground-truth labels come from static AV signatures, not behavioral semantics.

“ In contrast, Fathom’s full evidence extraction pipeline achieves 0.841 Parent F1 on real CAPEv2 reports. This demonstrates that structured behavioral evidence + multi-source context (not raw API text) is the critical enabler for production-grade malware analysis.”

6. Independent Benchmarks (Zero Training-Corpus Overlap)

CyberMetric questions overlap slightly with the training corpus (measured, disclosed, and shown not to inflate scores — see the paper). To settle the question independently, Fathom was also evaluated on two benchmarks that share no data with CyberMetric, MMLU, or the training corpus (verified with an 8-gram shingle-overlap check, <0.3% overlap on every subset scored):

SECURE (Bhusal et al., 2024) — ATT&CK technique and CWE weakness mapping, the same task Fathom is built for, with a published base Mixtral-8x7B baseline for direct comparison:

Subset	Fathom unified-v2	Published base Mixtral-8x7B	Delta
MAET (ATT&CK technique MCQ, n=1072)	87.78%	80.9%	+6.9pp
CWET (CWE weakness MCQ, n=965)	87.88%	83.4%	+4.5pp

SECURE's other two MCQ subsets, KCV and VOOD (CVE-severity boolean prediction — a different task from Fathom's ATT&CK/malware-analysis focus), were scored but excluded from this comparison: the official task grounds each question in an embedded CVE-JSON record long enough to exceed this serving deployment's batched-token limit, which broke that grounding. Scores without it aren't a fair comparison to the published baseline, so they're reported separately in the raw results rather than here.

CyberSOCEval (Meta CyberSecEval 4) — multi-select reasoning over real sandbox detonation logs and CTI reports, no published Mixtral baseline exists yet:

Subset	Exact-set accuracy	Mean Jaccard overlap
Malware-analysis (n=609)	10.18%	0.415
Threat-intel reasoning (n=588)	27.04%	0.555

For context, the paper's conservative 20-case synthetic-suite lower bound is Parent F1 0.344 — the independent-benchmark numbers above sit alongside that as additional, non-overlapping evidence, not a replacement for it.

Raw predictions and logs: journal-artifacts/evaluation/.

How to Use

Loading the unified model (recommended for most users)

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model_name = "mistralai/Mixtral-8x7B-Instruct-v0.1"
adapter = "umer07/fathom-mixtral"   # unified-v2 at root

tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype=torch.bfloat16,
    device_map="auto",
    trust_remote_code=True
)
model = PeftModel.from_pretrained(model, adapter, adapter_name="unified-v2")
model.eval()

Limitations

Sub-technique precision lower than parent techniques (standard across all LLMs)
Family identification improves significantly with KSPN enrichment
Rare/exotic TTPs (UAC bypass, ICMP C2) have low recall
Prompt injection / attribution hallucination remains a base-model weakness (mitigable with system prompt hardening)

Training & Datasets

Unified-v2: 123,912 rows (1 epoch)
Experts: 9 specialized datasets (total > 200k rows after augmentation)
Evasive dataset (NEW): 25,160 obfuscated C++ samples (92 evasion combinations)
ThreatIntel upgrade: 9,532 rows (URLhaus + GTFOBins + MITRE CTI)

Citation

@misc{fathom2026,
  title={Fathom: Expert Cybersecurity Analysis with Mixtral LoRA Adapters},
  author={Abdul Hadi and Muhammad Haseeb and Muhammad Ammar and Sana Aurangzeb},
  year={2026},
  howpublished={\url{https://huggingface.co/umer07/fathom-mixtral}},
}