Скачать презентацию Automated Worm Fingerprinting Sumeet Singh Cristian Estan George Скачать презентацию Automated Worm Fingerprinting Sumeet Singh Cristian Estan George

1c02573db1219ed646180425a2ffb19a.ppt

  • Количество слайдов: 88

Automated Worm Fingerprinting Sumeet Singh, Cristian Estan, George Varghese, and Stefan Savage Automated Worm Fingerprinting Sumeet Singh, Cristian Estan, George Varghese, and Stefan Savage

Introduction n n Problem: how to react quickly to worms? Code. Red 2001 ¡ Introduction n n Problem: how to react quickly to worms? Code. Red 2001 ¡ n Infected ~360, 000 hosts within 11 hours Sapphire/Slammer (376 bytes) 2002 ¡ Infected ~75, 000 hosts within 10 minutes

Existing Approaches n Detection ¡ n Ad hoc intrusion detection Characterization ¡ Manual signature Existing Approaches n Detection ¡ n Ad hoc intrusion detection Characterization ¡ Manual signature extraction n Isolates and decompiles a new worm Look for and test unique signatures Can take hours or days

Existing Approaches n Containment ¡ Updates to anti-virus and network filtering products Existing Approaches n Containment ¡ Updates to anti-virus and network filtering products

Earlybird n n Automatically detect and contain new worms Two observations ¡ ¡ Some Earlybird n n Automatically detect and contain new worms Two observations ¡ ¡ Some portion of the content in existing worms is invariant Rare to see the same string recurring from many sources to many destinations

Earlybird n Automatically extract the signature of all known worms ¡ n Also Blaster, Earlybird n Automatically extract the signature of all known worms ¡ n Also Blaster, My. Doom, and Kibuv. B hours or days before any public signatures were distributed Few false positives

Background and Related Work n Almost all IPs were scanned by Slammer < 10 Background and Related Work n Almost all IPs were scanned by Slammer < 10 minutes ¡ Limited only by bandwidth constraints

The SQL Slammer Worm: 30 Minutes After “Release” - Infections doubled every 8. 5 The SQL Slammer Worm: 30 Minutes After “Release” - Infections doubled every 8. 5 seconds - Spread 100 X faster than Code Red - At peak, scanned 55 million hosts per second.

Network Effects Of The SQL Slammer Worm n At the height of infections ¡ Network Effects Of The SQL Slammer Worm n At the height of infections ¡ ¡ ¡ Several ISPs noted significant bandwidth consumption at peering points Average packet loss approached 20% South Korea lost almost all Internet service for period of time Financial ATMs were affected Some airline ticketing systems overwhelmed

Signature-Based Methods n Pretty effective if signatures can be generated quickly ¡ ¡ For Signature-Based Methods n Pretty effective if signatures can be generated quickly ¡ ¡ For Code. Red, 60 minutes For Slammer, 1 – 5 minutes

Worm Detection n Three classes of methods ¡ ¡ ¡ Scan detection Honeypots Behavioral Worm Detection n Three classes of methods ¡ ¡ ¡ Scan detection Honeypots Behavioral techniques

Scan Detection n n Look for unusual frequency and distribution of address scanning Limitations Scan Detection n n Look for unusual frequency and distribution of address scanning Limitations ¡ Not suited to worms that spread in a nonrandom fashion (i. e. emails, IM, P 2 P apps) n n Based on a target list Spread topologically

Scan Detection n More limitations ¡ ¡ Detects infected sites Does not produce a Scan Detection n More limitations ¡ ¡ Detects infected sites Does not produce a signature

Honeypots n Monitored idle hosts with untreated vulnerabilities ¡ n Used to isolate worms Honeypots n Monitored idle hosts with untreated vulnerabilities ¡ n Used to isolate worms Limitations ¡ ¡ Manual extraction of signatures Depend on quick infections

Behavioral Detection n Looks for unusual system call patterns ¡ ¡ n Sending a Behavioral Detection n Looks for unusual system call patterns ¡ ¡ n Sending a packet from the same buffer containing a received packet Can detect slow moving worms Limitations ¡ ¡ Needs application-specific knowledge Cannot infer a large-scale outbreak

Characterization n n Process of analyzing and identifying a new worm Current approaches ¡ Characterization n n Process of analyzing and identifying a new worm Current approaches ¡ ¡ Use a priori vulnerability signatures Automated signature extraction

Vulnerability Signatures n Example ¡ Slammer Worm n n UDP traffic on port 1434 Vulnerability Signatures n Example ¡ Slammer Worm n n UDP traffic on port 1434 that is longer than 100 bytes (buffer overflow) Can be deployed before the outbreak ¡ Can only be applied to well-known vulnerabilities

Some Automated Signature Extraction Techniques n Allows viruses to infect decoy programs ¡ ¡ Some Automated Signature Extraction Techniques n Allows viruses to infect decoy programs ¡ ¡ Extracts the modified regions of the decoy Uses heuristics to identify invariant code strings across infected instances

Some Automated Signature Extraction Techniques n Limitation ¡ Assumes the presence of a virus Some Automated Signature Extraction Techniques n Limitation ¡ Assumes the presence of a virus in a controlled environment

Some Automated Signature Extraction Techniques n Honeycomb ¡ n Autograph ¡ n Find longest Some Automated Signature Extraction Techniques n Honeycomb ¡ n Autograph ¡ n Find longest common subsequences among sets of strings found in messages Uses network-level data to infer worm signatures Limitations ¡ Scale and full distributed deployments

Containment n Mechanism used to deter the spread of an active worm ¡ Host Containment n Mechanism used to deter the spread of an active worm ¡ Host quarantine n ¡ ¡ Via IP ACLs on routers or firewalls String-matching Connection throttling n On all outgoing connections

Host Quarantine n Preventing an infected host from talking to other hosts ¡ Via Host Quarantine n Preventing an infected host from talking to other hosts ¡ Via IP ACLs on routers or firewalls

Defining Worm Behavior n Content invariance ¡ n Content prevalence ¡ n Portions of Defining Worm Behavior n Content invariance ¡ n Content prevalence ¡ n Portions of a worm are invariant (e. g. the decryption routine) Appears frequently on the network Address dispersion ¡ Distribution of destination addresses more uniform to spread fast

Finding Worm Signatures n Traffic pattern is sufficient for detecting worms ¡ ¡ ¡ Finding Worm Signatures n Traffic pattern is sufficient for detecting worms ¡ ¡ ¡ Relatively straightforward Extract all possible substrings Raise an alarm when n Frequency. Counter[substring] > threshold 1 Source. Counter[substring] > threshold 2 Dest. Counter[substring] > threshold 3

Practical Content Sifting n Characteristics ¡ ¡ ¡ Small processing requirements Small memory requirements Practical Content Sifting n Characteristics ¡ ¡ ¡ Small processing requirements Small memory requirements Allows arbitrary deployment strategies

Estimating Content Prevalence n Finding the packet payloads that appear at least x times Estimating Content Prevalence n Finding the packet payloads that appear at least x times among the N packets sent ¡ During a given interval

Estimating Content Prevalence n Table[payload] ¡ n 1 GB table filled in 10 seconds Estimating Content Prevalence n Table[payload] ¡ n 1 GB table filled in 10 seconds Table[hash[payload]] ¡ ¡ ¡ 1 GB table filled in 4 minutes Tracking millions of ants to track a few elephants Collisions. . . false positives

Multistage Filters stream memory Array of counters Hash(Pink) [Singh et al. 2002] Multistage Filters stream memory Array of counters Hash(Pink) [Singh et al. 2002]

Multistage Filters packet memory Array of counters Hash(Green) Multistage Filters packet memory Array of counters Hash(Green)

Multistage Filters packet memory Array of counters Hash(Green) Multistage Filters packet memory Array of counters Hash(Green)

Multistage Filters packet memory Multistage Filters packet memory

Multistage Filters Collisions are OK packet memory Multistage Filters Collisions are OK packet memory

Multistage Filters Reached threshold packet memory packet 1 1 Insert Multistage Filters Reached threshold packet memory packet 1 1 Insert

Multistage Filters packet memory packet 1 1 Multistage Filters packet memory packet 1 1

Multistage Filters packet memory packet 1 1 packet 2 1 Multistage Filters packet memory packet 1 1 packet 2 1

Multistage Filters Stage 1 packet memory packet 1 1 No false negatives! (guaranteed detection) Multistage Filters Stage 1 packet memory packet 1 1 No false negatives! (guaranteed detection) Stage 2

Conservative Updates Gray = all prior packets Conservative Updates Gray = all prior packets

Conservative Updates Redundant Conservative Updates Redundant

Conservative Updates Conservative Updates

Detecting Common Strings n n Cannot afford to detect all substrings Maybe can afford Detecting Common Strings n n Cannot afford to detect all substrings Maybe can afford to detect all strings with a small fixed length

Detecting Common Strings n n Cannot afford to detect all substrings Maybe can afford Detecting Common Strings n n Cannot afford to detect all substrings Maybe can afford to detect all strings with a small fixed length A horse is a horse, of course F 1 = (c 1 p 4 + c 2 p 3 + c 3 p 2 + c 4 p 1 + c 5) mod M

Detecting Common Strings n n Cannot afford to detect all substrings Maybe can afford Detecting Common Strings n n Cannot afford to detect all substrings Maybe can afford to detect all strings with a small fixed length F 2 = (c 2 p 4 + c 3 p 3 + c 4 p 2 + c 5 p 1 + c 6) mod M A horse is a horse, of course F 1 = (c 1 p 4 + c 2 p 3 + c 3 p 2 + c 4 p 1 + c 5) mod M

Detecting Common Strings n n Cannot afford to detect all substrings Maybe can afford Detecting Common Strings n n Cannot afford to detect all substrings Maybe can afford to detect all strings with a small fixed length F 2 = (c 2 p 4 + c 3 p 3 + c 4 p 2 + c 5 p 1 + c 6) mod M = (c 1 p 5 + c 2 p 4 + c 3 p 3 + c 4 p 2 + c 5 p 1 + c 6 - c 1 p 5) mod M = (p. F 1 + c 6 - c 1 p 5) mod M A horse is a horse, of course F 1 = (c 1 p 4 + c 2 p 3 + c 3 p 2 + c 4 p 1 + c 5) mod M

Detecting Common Strings n n Cannot afford to detect all substrings Maybe can afford Detecting Common Strings n n Cannot afford to detect all substrings Maybe can afford to detect all strings with a small fixed length ¡ Still too expensive…

Estimating Address Dispersion n Not sufficient to count the number of source and destination Estimating Address Dispersion n Not sufficient to count the number of source and destination pairs ¡ e. g. send a mail to a mailing list n n n Two sources—mail server and the sender Many destinations Need to count the unique source and destination traffic flows ¡ For each substring

Bitmap counting – direct bitmap Set bits in the bitmap using hash of the Bitmap counting – direct bitmap Set bits in the bitmap using hash of the flow ID of incoming packets HASH(green)=10001001 [Estan et al. 2003]

Bitmap counting – direct bitmap Different flows have different hash values HASH(blue)=00100100 Bitmap counting – direct bitmap Different flows have different hash values HASH(blue)=00100100

Bitmap counting – direct bitmap Packets from the same flow always hash to the Bitmap counting – direct bitmap Packets from the same flow always hash to the same bit HASH(green)=10001001

Bitmap counting – direct bitmap Collisions OK, estimates compensate for them HASH(violet)=10010101 Bitmap counting – direct bitmap Collisions OK, estimates compensate for them HASH(violet)=10010101

Bitmap counting – direct bitmap HASH(orange)=11110011 Bitmap counting – direct bitmap HASH(orange)=11110011

Bitmap counting – direct bitmap HASH(pink)=11100000 Bitmap counting – direct bitmap HASH(pink)=11100000

Bitmap counting – direct bitmap As the bitmap fills up, estimates get inaccurate HASH(yellow)=01100011 Bitmap counting – direct bitmap As the bitmap fills up, estimates get inaccurate HASH(yellow)=01100011

Bitmap counting – direct bitmap Solution: use more bits HASH(green)=10001001 Bitmap counting – direct bitmap Solution: use more bits HASH(green)=10001001

Bitmap counting – direct bitmap Solution: use more bits Problem: memory scales with the Bitmap counting – direct bitmap Solution: use more bits Problem: memory scales with the number of flows HASH(blue)=00100100

Bitmap counting – virtual bitmap Solution: a) store only a portion of the bitmap Bitmap counting – virtual bitmap Solution: a) store only a portion of the bitmap b) multiply estimate by scaling factor

Bitmap counting – virtual bitmap HASH(pink)=11100000 Bitmap counting – virtual bitmap HASH(pink)=11100000

Bitmap counting – virtual bitmap Problem: estimate inaccurate when few flows active HASH(yellow)=01100011 Bitmap counting – virtual bitmap Problem: estimate inaccurate when few flows active HASH(yellow)=01100011

Bitmap counting – multiple bmps Solution: use many bitmaps, each accurate for a different Bitmap counting – multiple bmps Solution: use many bitmaps, each accurate for a different range

Bitmap counting – multiple bmps HASH(pink)=11100000 Bitmap counting – multiple bmps HASH(pink)=11100000

Bitmap counting – multiple bmps HASH(yellow)=01100011 Bitmap counting – multiple bmps HASH(yellow)=01100011

Bitmap counting – multiple bmps Use this bitmap to estimate number of flows Bitmap counting – multiple bmps Use this bitmap to estimate number of flows

Bitmap counting – multiple bmps Use this bitmap to estimate number of flows Bitmap counting – multiple bmps Use this bitmap to estimate number of flows

Bitmap counting – multires. bmp OR OR Problem: must update up to three bitmaps Bitmap counting – multires. bmp OR OR Problem: must update up to three bitmaps per packet Solution: combine bitmaps into one

Bitmap counting – multires. bmp HASH(pink)=11100000 Bitmap counting – multires. bmp HASH(pink)=11100000

Bitmap counting – multires. bmp HASH(yellow)=01100011 Bitmap counting – multires. bmp HASH(yellow)=01100011

Multiresolution Bitmaps n n Still too expensive to scale Scaled bitmap ¡ ¡ Recycles Multiresolution Bitmaps n n Still too expensive to scale Scaled bitmap ¡ ¡ Recycles the hash space with too many bits set Adjusts the scaling factor according n E. g. , 1 bit represents 2 flows as opposed to a single flow

Too CPU-Intensive n A packet with 1, 000 bytes of payload ¡ n Needs Too CPU-Intensive n A packet with 1, 000 bytes of payload ¡ n Needs 960 fingerprints for string length of 40 Prone to Denial-of-Service attacks

CPU Scaling n Obvious approach: sampling - Random sampling may miss many substrings n CPU Scaling n Obvious approach: sampling - Random sampling may miss many substrings n Solution: value sampling ¡ Track only certain substrings n ¡ e. g. last 6 bits of fingerprint are 0 P(not tracking a worm) = P(not tracking any of its substrings)

CPU Scaling n Example ¡ ¡ ¡ Track only substrings with last 6 bits CPU Scaling n Example ¡ ¡ ¡ Track only substrings with last 6 bits = 0 s String length = 40 1, 000 char string n n ¡ 960 substrings 960 fingerprints ‘ 11100… 101010’…‘ 10110… 000000’… Use only ‘xxxxx…. 000000’ as signatures n Probably 960 / 26 = 15 signatures

CPU Scaling n n n P(finding a 100 -byte signature) = 55% P(finding a CPU Scaling n n n P(finding a 100 -byte signature) = 55% P(finding a 200 -byte signature) = 92% P(finding a 400 -byte signature) = 99. 64%

Putting It Together Address Dispersion Table key header src cnt dest cnt payload substring Putting It Together Address Dispersion Table key header src cnt dest cnt payload substring fingerprints AD entry exist? substring fingerprints update counters else update counter key cnt counters > dispersion threshold? report key as suspicious worm cnt > prevalence threshold? create AD entry Content Prevalence Table

Putting It Together n n n Sample frequency: 1/64 String length: 40 Use 4 Putting It Together n n n Sample frequency: 1/64 String length: 40 Use 4 hash functions to update prevalence table ¡ Multistage filter reset every 60 seconds

System Design n Two major components ¡ Sensors n n ¡ Sift through traffic System Design n Two major components ¡ Sensors n n ¡ Sift through traffic for a given address space Report signatures An aggregator n n Coordinates real-time updates Distributes signatures

Implementation and Environment n n Written in C and My. SQL (5, 000 lines) Implementation and Environment n n Written in C and My. SQL (5, 000 lines) rrd-tools library for graphical reporting PHP scripting for administrative control Prototype executes on a 1. 6 Ghz AMD Opteron 242 1 U Server ¡ Linux 2. 6 kernel

Early. Bird n n Processes 1 TB of traffic per day Can keep up Early. Bird n n Processes 1 TB of traffic per day Can keep up with 200 Mbps of continuous traffic

Parameter Tuning n Prevalence threshold: 3 ¡ n Very few signatures repeat Address dispersion Parameter Tuning n Prevalence threshold: 3 ¡ n Very few signatures repeat Address dispersion threshold ¡ ¡ ¡ 30 sources and 30 destinations Reset every few hours Reduces the number of reported signatures down to ~25, 000

Parameter Tuning n Tradeoff between and speed and accuracy ¡ Can detect Slammer in Parameter Tuning n Tradeoff between and speed and accuracy ¡ Can detect Slammer in 1 second as opposed to 5 seconds n With 100 x more reported signatures

Performance n n 200 Mbps Can be pipelined and parallelized for achieve 40 Gbps Performance n n 200 Mbps Can be pipelined and parallelized for achieve 40 Gbps

Memory Consumption n Prevalence table ¡ 4 stages n ¡ n 2 MB total Memory Consumption n Prevalence table ¡ 4 stages n ¡ n 2 MB total Address dispersion table ¡ ¡ n Each with ~500, 000 bins (8 bits/bin) 25 K entries (28 bytes each) < 1 MB Total: < 4 MB

Trace-Based Verification n Two main sources of false positives ¡ 2, 000 common protocol Trace-Based Verification n Two main sources of false positives ¡ 2, 000 common protocol headers n n ¡ ¡ e. g. HTTP, SMTP Whitelisted SPAM e-mails Bit. Torrent n Many-to-many download

False Negatives n n So far none Detected every worm outbreak False Negatives n n So far none Detected every worm outbreak

Inter-Packet Signatures n n An attacker might evade detection by splitting an invariant string Inter-Packet Signatures n n An attacker might evade detection by splitting an invariant string across packets With 7 MB extra, Early. Bird can keep per flow states and fingerprint across packets

Live Experience with Early. Bird n Detected precise signatures ¡ ¡ Code. Red variants Live Experience with Early. Bird n Detected precise signatures ¡ ¡ Code. Red variants My. Doom mail worm Sasser Kibvu. B

Variant Content n Polymorphic viruses ¡ ¡ Semantically equivalent but textually distinct code Invariant Variant Content n Polymorphic viruses ¡ ¡ Semantically equivalent but textually distinct code Invariant decoding routine

Extensions n n Self configuration Slow worms Extensions n n Self configuration Slow worms

Containment n How to handle false positives? ¡ ¡ If too aggressive, Early. Bird Containment n How to handle false positives? ¡ ¡ If too aggressive, Early. Bird becomes a target for Do. S attacks An attacker can fool the system to block a target message

Coordination n Trust of deployed servers Validation Policy Coordination n Trust of deployed servers Validation Policy

Conclusions n Early. Bird is a promising approach ¡ ¡ ¡ n To detect Conclusions n Early. Bird is a promising approach ¡ ¡ ¡ n To detect unknown worms real-time To extract signatures automatically To detect SPAMs with minor changes Wire-speed signature learning is viable