Skip to main content

ReversingLabs vs. AssemblyLine

AssemblyLine is an open-source file triage framework from the Canadian Centre for Cyber Security. This page compares it with the ReversingLabs Spectra platform: Spectra Analyze, Spectra Detect, Spectra Intelligence, and Spectra Core.

AssemblyLine scores heuristics, while Spectra returns a malicious, suspicious, goodware, or unknown verdict with a threat name. AssemblyLine ships no reputation data, while Spectra Intelligence adds file and network reputation behind every verdict. With AssemblyLine, your team sources signatures, antivirus engines, and sandboxes. ReversingLabs maintains detection content for you.

Analysis and Intelligence​

CriterionReversingLabs SpectraAssemblyLine
VerdictMalicious, suspicious, goodware, unknownHeuristic score; the operator sets the alert threshold
Threat namingPlatform.Type.Family naming standard; 70,000+ families recognizedHeuristics and tags; no family naming in the tested services
Threat intelligenceSpectra Intelligence: file and network reputation from a 400B+ malware and goodware sample repositoryNone bundled; optional external services (for example, VirusTotal, Intezer)
Goodware and trustGoodware via certificates, reputation, and trust factorOptional Safelist service (for example, NSRL) that you maintain
File coverageSpectra Core: unpacks 400+ formats, identifies 4,800+ file typesPer service: each service declares the file types it accepts

Operations​

CriterionReversingLabs SpectraAssemblyLine
Origin and licenseCommercial platform from ReversingLabsCanadian Centre for Cyber Security; MIT license, released in October 2017
DeploymentCloud, on-premises virtual appliances, air-gappedSelf-hosted on Docker or Kubernetes
ScaleSpectra Detect clusters scale from 100K to 100M files per day; files up to 600 GBKubernetes scaling; you size, tune, and operate the cluster
Dynamic analysisReversingLabs Cloud Sandbox built in; Joe Sandbox and VMRay integrationsCAPE service, connected to a CAPE sandbox you deploy
SupportVendor support and maintained detection contentCommunity support channel (Discord)

Head-to-Head Test​

The same 150 real-world files went through both platforms:

  • 50 documents (Office, PDF)
  • 50 executables (PE, .NET, ELF)
  • 50 scripts (JavaScript)

The files were uploaded to a Spectra Analyze instance over about two months. Every file went through that one instance, and AssemblyLine ran on all 150 as Auxiliary Analysis, producing 150 reports.

AssemblyLine ran static and extraction services only, such as Oletools, ViperMonkey, CAPA, PE, FrankenStrings, Extract, and JsJaws. No YARA, antivirus, CAPE, VirusTotal, or Safelist services were enabled.

The ReversingLabs result is the Spectra verdict: the Spectra Analyze final classification, which combines static analysis, Spectra Intelligence reputation, and the ReversingLabs Cloud Sandbox. The ReversingLabs Cloud Sandbox detonated all 100 documents and executables. Scripts weren't detonated.

AssemblyLine scores map to these bands:

ScoreBand
0–299Informational
300–699Suspicious
700–999Highly suspicious
1,000 or higherMalicious

On this page, "flagged" means an AssemblyLine score of 300 or higher.

There's no independent ground truth. The 56 files that the ReversingLabs Cloud Sandbox classified as malicious are the most independent reference, and on this page they're called confirmed malicious. The Spectra verdict uses the sandbox result as an input, so Spectra's 100% on this set is expected.

A sample of 150 files is directional, not a benchmark. To compare the platforms on your own files, run the same test on your samples.

Detection of Confirmed Malicious Files​

PlatformFlaggedMalicious verdictThreat named
ReversingLabs Spectra100% (56 of 56)100% (56 of 56)100% (56 of 56)
AssemblyLine36% (20 of 56)5% (3 of 56)0% (0 of 56)

AssemblyLine scored the other 36 confirmed malicious files as informational. Across the test, Spectra named CVE-2017-11882 and CVE-2017-0199 exploits, AgentTesla, NjRAT, Knotweed, SnakeKeylogger, ValleyRAT, and Mirai.

Neither platform raised an alarm on the 52 goodware files. No goodware file crossed the AssemblyLine suspicious threshold.

Results by File Type​

File typeConfirmed maliciousReversingLabs SpectraAssemblyLine
Documents31100% (31)39% (12)
Executables25100% (25)32% (8)

The largest group of document misses was Office exploit documents (CVE-2017-11882, CVE-2017-0199): 8 of the 19 documents that AssemblyLine scored informational.

Scripts weren't detonated, so there's no behavioral reference for them. Spectra classified 4 of the 50 scripts as malicious, and AssemblyLine flagged 1 of those 4.

Sources of Spectra Verdicts​

Spectra returned 90 malicious verdicts. The deciding source is the classification reason that Spectra Analyze reports for the final verdict.

Deciding sourceFilesFlagged by AssemblyLine
Spectra Intelligence reputation589
ReversingLabs Cloud Sandbox2714
Explainable Machine Learning40
Embedded hyperlink10

ReversingLabs file and threat reputation resolved all 52 goodware files as goodware. AssemblyLine rated all 52 informational, which is a score rather than a statement of trust. A goodware verdict lets a SOC close an alert, while an informational score leaves it open.

MITRE ATT&CK Mapping​

PlatformFiles with ATT&CK techniquesTechnique mappings, total
ReversingLabs Spectra100% (150 of 150)1,410
AssemblyLine16% (24 of 150)101

Spectra combines static analysis on every file (514 mappings) with ReversingLabs Cloud Sandbox runtime behavior on the 100 documents and executables (896 mappings). Counts are summed per file across the test set.

AssemblyLine Strengths​

Teams choose AssemblyLine when they prioritize:

  • No license cost and full source access (MIT)
  • Writing custom analysis services in Python
  • Raw artifact extraction. In this test, it surfaced 717 URLs and 2,428 file hashes that Spectra Core didn't report. These include extracted strings and embedded files that weren't validated as malicious.
  • Few false alarms: no goodware file crossed its suspicious threshold

To reach verdict-grade results with AssemblyLine, you add:

  • YARA rule sets that you write, buy, or curate
  • Antivirus engines and their licenses
  • A CAPE sandbox to deploy and maintain
  • Threat intelligence subscriptions, such as VirusTotal or Intezer
  • Staff to run the cluster and tune scoring