Updated time: 2023-03-08 12:00:55
05A-1100218
2021-04-19
Patent pending
MAMBA, a dynamic behavior analysis system for Windows malware, incorporates ATT&CK knowledge and considers attentions on manipulated resources and malicious activities in the neural network model. MAMBA can scan execution traces to identify potentially malicious activities with
execution codes (i.e., API calls) corresponding to ATT&CK behaviors (TTPs). This helps cybersecurity analysts to comprehend malware’s behavior and software developers to test application security.
The main core techniques of MAMBA involve 1) knowledge collection from ATT&CK, 2) malicious dynamic behavior identification using deep learning, and 3) the identification of associated API calls.
- Cyber threats knowledge integration for gathering the manipulated resources associated with malicious behavior.
- AI-based malware behavior analysis includes identifying malicious behavior from a malware sample's execution trace.
- API call Locating involves aligning high-level behavior to a low-level execution trace.
- Malware forensic analysis: the identified malicious behavior is identified with high-level descriptions of TTPs in ATT&CK as interpretations, linking to low-level execution traces of malware.
- Software security testing: it can detect software security threats when downloaded and presents them with both high-level interpretations and low-level execution traces.
A high-level overview of the MAMBA workflow is shown in Figure 1: this is composed of an extraction phase, a fusion phase, and a threat identification phase. The extraction phase includes technique extraction by extracting knowledge tuples from ATT&CK, and malware execution trace generation from a sandbox. The fusion phase involves the designed resource embedding mechanism maps resources to fixed-size vectors while preserving their semantic properties, and the resource-technique binding neural network to learn the connection between resources and techniques from ATT&CK. The threat identification is a developed neural network model which recognizes malicious behavior as TTPs (tactics, techniques, and procedures) from a malware sample.
Fig2. is MAMBA neural network architecture for threat identification phase. When given malware execution trace with API calls, MAMBA identifies malicious behavior with natural language interpretation. An execution trace is composed of the traces of all processes; each process trace is a sequence of API calls. MAMBA firstly computes the weighted resourced-based API call groups for each process, and then incorporates the binding information from MITRE ATT&CK framework into the malware representation, finally discovers possible malicious behavior of a malware. The detailed procedure includes that 1) API calls are processed by embedding and gated recurrent units (GRUs) layers to obtain sequential hidden vectors, 2) attention mechanisms are applied to highlight the relevance between resources and API calls as well as dependencies among the bindings and API calls, and 3) threat identification yields the compromised techniques.
Fig3. MAMBA infers malicious behavior, TTPs, and locates related API calls for a given execution trace. When given the group attentions and the resource attentions, MAMBA find the dominant resources for discovering TTPs and locating the related API calls in a process. More specifically, a set of manipulated resources is selected based on two criteria: i) the similarity between the resources documented in ATT\&CK and the manipulated resources, and ii) the group attention. The similarity scores reveal the likelihood that a certain resource is being manipulated to implement a TTP. The group attention measures how much information a resource provides, that is, whether it is a common or rare resource across API calls. A large group attention value for a resource indicates that the resource is frequently used among API calls or processes; in contrast, a resource with a small group value means that it is uniquely representative, or is used only by chance. Security analysts use this to select observable resources by setting thresholds for the corresponding similarity scores or the number of highest and lowest attention values.
Once the resources are selected, malicious behavior can be located via the API calls whose resource attention values are larger than the largest attention value minus times the standard deviation.
Meng Chang Chen、Yi-Ting Huang
If you are interested in this technology, please fill in the information on the webpage under the manufacturer option (Request for Technology Licensing, Material Transfer, or Sponsored Research) on the website of this office

MAMBA (MITRE ATT&CK based malicious behavior analysis)