UNVEIL: A Large-Scale, Automated Approach to Detecting Ransomware
0 likes156 views
This document summarizes a research paper on developing an automated approach called UNVEIL to detect ransomware. UNVEIL generates fake user environments and monitors filesystem activity to extract IO access sequences that are characteristic of ransomware families. It was evaluated on 3500 known ransomware samples with 100% precision and recall. During testing on over 148,000 unknown samples, UNVEIL discovered a previously unknown ransomware family called SilentCrypt. The researchers conclude UNVEIL is useful in practice for detecting ransomware.
UNVEIL: A Large-Scale, Automated Approach to Detecting Ransomware
- 1. Northeastern University International Secure Systems Lab A Large-Scale, Automated Approach to Detecting Ransomware Amin Kharraz, Sajjad Arshad, Collin Mulliner, William Robertson, Engin Kirda NEU SECLAB
- 2. NEU SECLAB What is a ransomware attack? Receiving the decryption key2 Paying the ransom fee1 2
- 3. NEU SECLAB 3 A Typical Ransom Note
- 4. NEU SECLAB 4 Attacks on Hospitals
- 5. NEU SECLAB 5
- 6. NEU SECLAB 7
- 7. – CNN Interview with FBI, April 2016 “Cyber-criminals collected $209 million in the first three months of 2016 by extorting businesses and institutions to unlock computer servers.”
- 8. NEU SECLAB 8 How to defend against ransomware attacks? ● Educating end-users - Have a reliable backup policy - Avoid risky online behavior ● Developing detection tools to assist defenders - Providing insight from internal behavior ● Developing protection tools to enhance AV capabilities - Stopping the attack, and keeping the data consistent
- 9. NEU SECLAB 8 How to defend against ransomware attacks? ● Educating end-users - Have a reliable backup policy - Avoid risky online behavior ● Developing detection tools to assist defenders - Providing insight from internal behavior ● Developing protection tools to enhance AV capabilities - Stopping the attack, and keeping the data consistent
- 10. NEU SECLAB 9 Threat Model ● Ransomware can employ any techniques to attack - Inject code into benign processes - Perform encrypted communication - Leverage arbitrary cryptosystems ● We assume that OS kernel, and underlying software and hardware stack are free of malicious code. ● Unveil detects ransomware during dynamic analysis phase, and not at end-user machines. - Complements current dynamic analysis systems - A cloud-based malware analysis service, sample sharing But, how can we detect a ransomware sample?
- 11. NEU SECLAB 10 Achilles’ Heel of Ransomware • Ransomware has to inform victim that attack has taken place • Ransomware has certain behaviors that are predictable – e.g., entropy changes, modal dialogs and background activity, accessing user files • A good sandbox that looks for some of these signs helps here…
- 12. NEU SECLAB 10 UNVEIL: An Early Warning Dynamic Detection System for Ransomware
- 14. NEU SECLAB Approach 12 • Detecting Cryptographic Ransomware: • Generating a fake (and attractive) user environment • Finding a reliable method for monitoring filesystem activity Why do we generate fake user environments? • Making the analysis environment more realistic • Protecting the analysis system from some user environment fingerprinting - A static user environment can be easily detected by a malware
- 15. NEU SECLAB Approach 12 • Detecting Cryptographic Ransomware: • Generating a fake (and attractive) user environment • Finding a reliable method for monitoring filesystem activity Why do we generate fake user environments? • Making the analysis environment more realistic • Protecting the analysis system from bare-user environment fingerprinting - A static user environment can be easily detected by a malware How do we generate fake user environments?
- 16. NEU SECLAB 13 Generating Fake (Honey) Content • Real files with valid headers – Using standard libraries (e.g., python- docx, python-pptx, OpenSSL) – Content that appears meaningful – File names do not look random, and appear realistic • File paths – User’s directory structure is generated randomly, but meaningfully • File attributes – Generate content with different creation, modification, and access times
- 18. User Kernel Rfs = <Time,Pname,Pid,PPid,IRPflag,Arg,Result,BufEntropy> I/O MANAGER UNVEIL
- 19. NEU SECLAB 17 Extracting I/O Access Sequences (1) Overwrites the users’ file with an encrypted version (2) reads, encrypts and deletes files without wiping them from storage (3) reads, creates a new encrypted version, and securely deletes the original files
- 20. NEU SECLAB 18 IO Access Sequences in Multiple Ransomware Families Read Write New File Encrypted version Deleting the Original File New File Encrypted version Secure Deletion
- 21. NEU SECLAB 19 Iteration over files during a CryptoWall attack
- 22. Desktop Locker Ransomware NEU SECLAB 1 Malware run 2 3 20
- 23. Desktop Locker Ransomware NEU SECLAB 20 Dissimilarity Score
- 24. NEU SECLAB 21 Preparing the Analysis Environment ● UNVEIL is deployed on top of Cuckoo Sandbox - UNVEIL supports all versions of Windows platforms. - Our tool is deployed in Kernel. - Bypassing UNVEIL is not technically easy in user-mode. ● Finding active malware is not easy - We modified some parts of Cuckoo to make it more resilient to environmentally sensitive samples - e.g., fake response to some of the environment checks - Other anti-evasion measures to look more realistic - e.g., defining multiple NTFS drives, changing IP address range and MAC addresses
- 25. NEU SECLAB 22 Evaluation 1) Detecting known ransomware samples a) Collecting ~3500 ransomware from public repo, Anubis, two security companies. b) 149 benign executables including ransomware-like behavior c) 348 malware samples from 36 malware families Ransomware FamiliesBenign Applications
- 26. NEU SECLAB 23 Dissimilarity score is different from family to family
- 27. NEU SECLAB Detecting known ransomware samples The threshold value t = 0.32 gives the highest recall with 100% precision 24
- 28. NEU SECLAB Evaluation UNVEIL with unknown samples 25 . . .56 UNVEIL-enabled VMs on 8 Servers Ganeti Cluster ~ 1200 malware samples per day
- 29. NEU SECLAB Evaluation UNVEIL with unknown samples ● We used the same similarity threshold (t = 0.32) for the large scale experiment. ● The incoming samples were acquired from the daily malware feed provided by Anubis from March 18 to February 12, 2016. ● The dataset contained 148,223 distinct samples. 26
- 30. NEU SECLAB Cross-checking with VirusTotal 27 ● Pollution ratio is defined as the ratio of the number of scanners that identified the sample to the number of scanners in VirusTotal
- 32. NEU SECLAB Detection: New Ransomware Family 29
- 33. NEU SECLAB Detection: New Ransomware Family 30 • During our experiments, we discovered a new malware family – We call it “SilentCrypt” – After we reported it, others started detecting it as well – We were not able to find any information about this family online – The ransomware first checks for private files of a user, contacts the C&C server, and starts the attack based on the answer
- 34. NEU SECLAB Detection: New Ransomware Family 31
- 35. NEU SECLAB 32 ● Ransomware is a serious threat ● UNVEIL introduces concrete models to detect Ransomware ● Detecting an unknown family shows that the solutions are useful in practice ● We continue to improve functionality tuned towards detecting ransomware Conclusion