debian/tor.init in the Debian tor0.2.9.11-1~deb9u1 package for Tor was designed to execute aa-exec from the standard system pathname if the apparmor package is installed, but implements this incorrectly (with a wrong assumption that the specific pathname would remain the same forever), which allows attackers to bypass intended AppArmor restrictions by leveraging the silent loss of this protection mechanism. NOTE: this does not affect systems, such as default Debian stretch installations, on which Tor startup relies on a systemd unit file (instead of this tor.init script).

The networkstatusparsevotefromstring function in routerparse.c in Tor before 0.2.2.38 does not properly handle an invalid flavor name, which allows remote attackers to cause a denial of service (out-of-bounds read and daemon crash) via a crafted (1) vote document or (2) consensus document.

routerlist.c in Tor before 0.2.2.38 uses a different amount of time for relay-list iteration depending on which relay is chosen, which might allow remote attackers to obtain sensitive information about relay selection via a timing side-channel attack.

A read from already freed memory and double free flaws were found in the way Tor, a connection-based low-latency anonymous communication system, performed processing of certain failing DNS requests. A remote attacker could issue a specially-crafted DNS request that, when processed would lead to tor executable crash.

Upstream ticket: [1] https://trac.torproject.org/projects/tor/ticket/6480

Relevant patch: [2] https://gitweb.torproject.org/tor.git/commitdiff/62637fa22405278758febb1743da9af562524d4c

References: [3] https://lists.torproject.org/pipermail/tor-announce/2012-August/000086.html [4] https://bugzilla.novell.com/showbug.cgi?id=776642

Tor before 0.2.2.24-alpha continues to use a reachable bridge that was previously configured but is not currently configured, which might allow remote attackers to obtain sensitive information about clients in opportunistic circumstances by monitoring network traffic to the bridge port.

Tor before 0.2.2.34, when configured as a bridge, uses direct DirPort access instead of a Tor TLS connection for a directory fetch, which makes it easier for remote attackers to enumerate bridges by observing DirPort connections.

Tor before 0.2.2.34, when configured as a bridge, sets up circuits through a process different from the process used by a client, which makes it easier for remote attackers to enumerate bridges by observing circuit building.

Tor before 0.2.2.25-alpha, when configured as a relay without the Nickname configuration option, uses the local hostname as the Nickname value, which allows remote attackers to obtain potentially sensitive information by reading this value.

Tor before 0.2.2.34, when configured as a client or bridge, sends a TLS certificate chain as part of an outgoing OR connection, which allows remote relays to bypass intended anonymity properties by reading this chain and then determining the set of entry guards that the client or bridge had selected.

Tor before 0.2.2.34, when configured as a bridge, accepts the CREATE and CREATEFAST values in the Command field of a cell within an OR connection that it initiated, which allows remote relays to enumerate bridges by using these values.

Multiple heap-based buffer overflows in Tor before 0.2.2.35 allow remote attackers to cause a denial of service (memory corruption) or possibly execute arbitrary code by (1) establishing a SOCKS connection to SocksPort or (2) leveraging a SOCKS proxy configuration.

A vulnerability in Tor was reported [1] that could allow a malicious remote attacker to cause a denial of service. This vulnerability is due to a boundary error within the policysummarize() function in src/or/policies.c which can be exploited to crash a Tor directory authority.

Upstream announced the release of 0.2.1.30 [2] that is noted to correct this flaw.

[1] http://secunia.com/advisories/43548/ [2] https://lists.torproject.org/pipermail/tor-announce/2011-February/000000.html

Tor before 0.2.1.29 and 0.2.2.x before 0.2.2.21-alpha does not properly manage key data in memory, which might allow local users to obtain sensitive information by leveraging the ability to read memory that was previously used by a different process.

Tor before 0.2.1.29 and 0.2.2.x before 0.2.2.21-alpha does not properly check the amount of compression in zlib-compressed data, which allows remote attackers to cause a denial of service via a large compression factor.

Tor before 0.2.1.29 and 0.2.2.x before 0.2.2.21-alpha makes calls to Libevent within Libevent log handlers, which might allow remote attackers to cause a denial of service (daemon crash) via vectors that trigger certain log messages.

Tor before 0.2.1.29 and 0.2.2.x before 0.2.2.21-alpha allows remote attackers to cause a denial of service (assertion failure and daemon exit) via blobs that trigger a certain file size, as demonstrated by the cached-descriptors.new file.

Tor before 0.2.1.29 and 0.2.2.x before 0.2.2.21-alpha might allow remote attackers to cause a denial of service (assertion failure and daemon exit) via vectors related to malformed router caches and improper handling of integer values.

The torrealloc function in Tor before 0.2.1.29 and 0.2.2.x before 0.2.2.21-alpha does not validate a certain size value during memory allocation, which might allow remote attackers to cause a denial of service (daemon crash) via unspecified vectors, related to "underflow errors."

Heap-based buffer overflow in Tor before 0.2.1.29 and 0.2.2.x before 0.2.2.21-alpha allows remote attackers to cause a denial of service (memory corruption and application crash) or possibly execute arbitrary code via unspecified vectors.

Heap-based buffer overflow in Tor before 0.2.1.28 and 0.2.2.x before 0.2.2.20-alpha allows remote attackers to cause a denial of service (daemon crash) or possibly execute arbitrary code via unspecified vectors.

Tor before 0.2.1.22, and 0.2.2.x before 0.2.2.7-alpha, when functioning as a bridge directory authority, allows remote attackers to obtain sensitive information about bridge identities and bridge descriptors via a dbg-stability.txt directory query.

Tor 0.2.2.x before 0.2.2.7-alpha, when functioning as a directory mirror, does not prevent logging of the client IP address upon detection of erroneous client behavior, which might make it easier for local users to discover the identities of clients in opportunistic circumstances by reading log files.

Tor before 0.2.1.22, and 0.2.2.x before 0.2.2.7-alpha, uses deprecated identity keys for certain directory authorities, which makes it easier for man-in-the-middle attackers to compromise the anonymity of traffic sources and destinations.

The connectionedgeprocessrelaycellnotopen function in src/or/relay.c in Tor 0.2.x before 0.2.0.35 and 0.1.x before 0.1.2.8-beta allows exit relays to have an unspecified impact by causing controllers to accept DNS responses that redirect to an internal IP address via unknown vectors. NOTE: some of these details are obtained from third party information.

Tor before 0.2.0.35 allows remote attackers to cause a denial of service (application crash) via a malformed router descriptor.

Unspecified vulnerability in Tor before 0.2.0.34 allows directory mirrors to cause a denial of service via unknown vectors.

Tor before 0.2.0.34 treats incomplete IPv4 addresses as valid, which has unknown impact and attack vectors related to "Spec conformance," as demonstrated using 192.168.0.

Unspecified vulnerability in Tor before 0.2.0.34 allows attackers to cause a denial of service (infinite loop) via "corrupt votes."

Unspecified vulnerability in Tor before 0.2.0.34 allows directory mirrors to cause a denial of service (exit node crash) via "malformed input."

Tor 0.2.0.28, and probably 0.2.0.34 and earlier, allows remote attackers, with control of an entry router and an exit router, to confirm that a sender and receiver are communicating via vectors involving (1) replaying, (2) modifying, (3) inserting, or (4) deleting a single cell, and then observing cell recognition errors at the exit router. NOTE: the vendor disputes the significance of this issue, noting that the product's design "accepted end-to-end correlation as an attack that is too expensive to solve."