170 likes | 277 Views
Scalable Parallel Intrusion Detection. Fahad Zafar. Advising Faculty: Dr. John Dorband and Dr. Yaacov Yeesha. University of Maryland Baltimore County. Intrusion Detection Systems (IDS). Network IDS
E N D
Scalable Parallel Intrusion Detection Fahad Zafar Advising Faculty: Dr. John Dorband and Dr. YaacovYeesha University of Maryland Baltimore County
Intrusion Detection Systems (IDS) • Network IDS • are placed at a strategic point or points within the network to monitor traffic to and from all devices on the network • Host IDS • monitors the inbound and outbound packets from the device only • Signature based IDS • will monitor packets on the network and compare them against a database of signatures or attributes from known malicious threats • Anomaly based IDS • will monitor network traffic and compare it against an established baseline
Existing Limitations • Network IDS: • Network Speed affected if you analyze all inbound and outbound traffic. • Host IDS: • Slows productivity. • Signature based IDS: • Signature database keeps increasing in size. • Anomaly based IDS: • Training models is hard.
Ping Broadcast Attack • Send an ICMP echo to the network broadcast address with spoofed ip of the server (victim)
Ping broadcast attacks • If you have 81 pcs on the network and your router forwards the request. A single echo request resulted in 81 echo replies, an 81x amplification of Internet traffic.
Points worth a mention • One type of IDS cannot handle all types of attacks • Application IDS cannot handle PING broadcast attacks, but network IDS’ can. • Network rules are needed for dynamic network management • When an attack is identified, write a rule for it.
Our Design • Understandings • Hetrogeneous IDS is the future • Better load balancing and minimum packet loss is a requirement. • Main Characteristics • Isolating different IDS • Traffic specific intrusion detection
Decentralized traffic based Heterogeneous Intrusion Detection eg. SNORT eg. OSSEC HIDS
Novelty • 1. Smart Switch • Block , Fork, Divert traffic. • Small cache for faster throughput. • 2. Decentralized Intrusion Detection • Working with current open source IDS packages • 3. Smart Hashing • Destination specific hashing. • Source specific hashing. • Session specific hashing.
Intrusion Detection Algorithms • Signature Extraction • Detect changes in registry, use of dlls • N-grams to train learning models and detect unknown viruses • Instance-Based Learner, Vector Machines, Decision Trees etc.
A scalable multi-level feature extraction technique to detect malicious executables [5] [5] Mohammad M. Masud & Latifur Khan & Bhavani Thuraisingham A scalable multi-level feature extraction technique to detect malicious executables
We explore multiple paths • Use semantic based searching for malicious code. • Use restricted Regular Expressions for parallel sequence and n-grams for the serial sequence. • Better feature extraction techniques for malicious and benign code.
Future Work: Evolution of Malware • Use metasploit for N-gram analysis • Test our detection techniques • Apply identification technique for encrypted and altered versions of malware code.
Future Work: Detecting a process in execution • Send tagged code and 16K memory dump • Offload work to bluegrit • Fast search according to signature + code sequence Reg-ex. • Reply to server within reasonable time limits
Future Work: Current Progress • Survey Infected Files. • Repository • Look for ways to reduce false negatives and false positives compared to previous approaches.[6] • Parallel scalable detection. [6] Learning to Detect and Classify Malicious Executables in the Wild J. Zico Kolter KOLTER, Marcus A. Maloof